An MS NOW memo reported by Semafor offers new details on the subscription offering.

On Sept. 9, the network plans to launch a paid membership program “aimed at engaging its existing ‘super fans,’ and offering ways to reach and monetize new audiences that are not everyday cable news viewers,” Semafor’s Max Tani reported on Wednesday. MS NOW president Rebecca Kutler framed the membership program as the first in a suite of planned direct-to-consumer offerings.

MS NOW was renamed after NBCUniversal split off the progressive network and other cable networks into a separate company called Versant. Now, it joins networks from CNN to Fox News to its former sibling NBC News in making digital audiences and monetization a priority as audiences and revenue from TV steadily decline. In 2024, two years after launching and shuttering its streaming platform, CNN put up a paywall ($3.99 a month or $29.99 a year for “basic,” and $6.99/month or $69.99/year for “all access”). NBC News has done the same ($5.99 a month, or $59.99 a year), while Fox News offers subscriptions to its streaming platform Fox Nation for $8.99/month or $71.49/year. CNBC, a part of Versant, is also rolling out its own premium consumer product.

The MS NOW membership will be priced at $7.99 a month, or $79.99 per year, and will “offer perks like members-only content verticals, opportunities for subscribers to interact with the network’s personalities, community features for members to engage with each other, and a 24/7 network feed,” Tani wrote. He added that MS NOW “is leaning more into the parasocial relationship its fans feel with its journalists,” including with live events.

The membership program is based on online survey responses and interviews with 18,000 network viewers, according to Tani. MS NOW told him they found “the network’s fans felt the news media ecosystem was too fragmented, and wanted to get news in one place.” They also found target audiences “wanted online spaces to interact with people they agreed with” and “expressed excitement at the possibility of interacting with the platform’s journalists and personalities.” Kutler noted that the network sees little overlap between its TV viewership and the current audience for its free online content.

Read more in Semafor here.

Backstory, an experimental tool from Google’s AI research lab, Google DeepMind, is helping India Today sort through this stream of slop, scams, and disinformation.

After uploading an image and entering a basic question, Backstory can automatically run checks to determine whether the image is AI-generated, whether it shows signs of manipulation, and where else the image has appeared throughout its lifetime on the internet. Built on top of Google’s Gemini family of large language models (LLMs), Backstory also leans on AI agents to decide which authentication tools it will use and in what order it will use them. The final output is an AI-generated report summarizing what Backstory found out about the image, with citations, as well as a log of the steps it took along the way.

At India Today, Backstory is primarily used to take the first pass on fact-checking a piece of content, speeding up the beginning of the verification workflow.

“On our team, almost everybody uses Backstory in combination with other things. We use it as a quick reference — it gives you some context, it gives you some direction,” said Bal Krishna, who leads the fact-checking team. “It is doing all the work you might have done with five different tools, five different logins, in the same place. That’s the beauty of it.”

In the past, reporters at India Today might have needed to separately run an image through an AI image detector, conduct a reverse image search, and then stitch together a timeline of its historical appearances. Backstory will do all those things at once and, for now, it is entirely free to use.

“The unique value add for Backstory is that there are a number of bites at the apple,” said Zoe Darmé, a product manager for Frontier AI Research, the team at DeepMind building the tool. “A lot of times when we’re talking about looking at an image, trying to understand it in context, people are focusing on one part of the chain. Backstory is really trying to pull it all together into a single user experience.”

Backstory is still only available through Google’s Trusted Testers Program, but India Today’s fact-checkers are among the thousands of journalists, OSINT experts, librarians, and information literacy researchers Google has brought on to trial the tool. The team is currently collecting feedback from this group to make improvements.

That’s already a departure from the many AI authentication tools built without adequate input from journalists, according to a report published in July by Princeton’s Center for Information Technology Policy.

Many other commercial AI detections tools don’t take into account more nuanced forms of manipulation — like divorcing a photograph from its original context — that journalists encounter all the time. A lot of these tools also output percent confidence ratings. For example, they might spit out an 80% likelihood that an image is AI-generated. These ratings are a “black box” with little clarity provided on how ratings are decided or ways to communicate those numbers to readers, according to the report.

Darmé told me she thinks of journalists as Backstory’s core users, but does hope to make the tool more widely available in the future. Even if Backstory launched publicly, she says it will most likely appeal to a niche audience invested in the nuances of image verification, and looking to answer more than the binary question of whether an image is “real” or “fake.”

According to Mike Caulfield, a digital literacy expert and creator of the SIFT method, building for this niche might be even more impactful than building a universal tool for anyone on the internet.

“Information environments work in this very intermediated way, where what actually has the most impact on an information environment is to help the helpers,” Caulfield told me, explaining that vetted information flows out from trusted sources, including journalists and fact-checkers, but also more casual hobbyists he calls the “family fact-checker.”

Caulfield describes Backstory as filling a hole in the market by building for these helpers. He notes that not every beat reporter who needs to verify an image is going to have subscriptions to high-quality AI detection tools, or the expertise to use the dozen of tools in an OSINT specialist’s toolbox.

“There hasn’t really been a tool for something in between the general user and that highly specialized user; between Bob from Peoria [Illinois] — not to pick on Peoria — and Craig Silverman,” said Caulfield, name-checking the disinformation expert and investigative journalist. “This is a tool that really hits that underserved middle.”

As part of its AI detection features, Backstory can run a tool call for SynthID — a way to check for watermarks embedded in images created with Nano Banana, Google’s AI image generation model. Backstory also checks an image’s content credentials, looking for encrypted metadata added by companies who have signed onto the Coalition for Content Provenance and Authenticity (C2PA) standard, including Google, Adobe, and the BBC.

Caulfield, however, says the tool’s ability to trace the history of an image online is often its most useful feature. That feature allows journalists to see an image in its original context, and chart how that context may have changed over time.

“The integration aspect — telling the full story of the image — turns out to be the really time-consuming part,” said Caulfield. “Backstory takes an initial process that would take you 50 minutes, and it can shrink it down to three.”

This feature takes on a challenge that long predates the rise of AI-generated imagery. “If you talk to any fact-checker, I don’t think anybody is going to tell you that they wouldn’t have loved to have this tool in 2016,” he said.

Google DeepMind, like many of the largest AI model makers, has contributed to the more recent explosion of AI-generated imagery on the internet. Nano Banana has been routinely criticized for contributing to disinformation about politicians and public health, and even for its use by online scammers.

Most recently, Google Earth came under fire in late July for integrating Nano Banana into its platform, allowing users to easily manipulate and alter 3D satellite images. Just a day later, Google announced it would roll back the feature and build “stronger guardrails” after examples of false and misleading imagery spread widely online.

Darmé says the decision to build Backstory wasn’t a direct response to these types of criticisms, but it was an attempt to address deteriorating trust in media content on the internet at large.

“Having worked on tools that brought image generation to a large population of users, having listened really closely to feedback from users in terms of what they wanted, there’s a clear user need,” she said, explaining DeepMind’s investment in building an easy-to-use provenance tool. “That might be something slightly different than [feeling] this team has one-to-one personal responsibility to release a tool because we also are in a company that releases Nano Banana.”

Like most tools powered by LLMs, Backstory can’t guarantee accuracy. A disclaimer appears at the bottom of the tool, similar to the one that appears on Google’s general AI assistants: “Backstory can make mistakes, so double-check it.”

A common problem with using LLMs to authenticate images is conflation, when the models mistake one image with another that is visually similar, according to Caulfield. “If you put in an image that can be described in the same way as another image that’s very popular online, it’s not uncommon for the LLM to confuse the two images,” he told me.

This problem surfaced in my own testing. After prompting Backstory to generate a report on the history of my own headshot, it ended up mistaking my image with the headshot of another journalist, Carlos Maza. His headshot came up with the same text search terms, including “man with curly hair,” “black cap,” “smiling.” Backstory output a full report on the history of Maza’s headshot on the Internet, before I fed my name into the tool and it eventually corrected itself.

There are other limitations with the tool. Right now, Backstory can only process images, though the team is exploring video and other modalities, according to Darmé. In its context tracing reports, Backstory can also only reference images that have been indexed on search engines. Even when Backstory claims to have found the origin of an image, it could have had a life beyond the tool’s view.

“It’s possible that something might have originated on TikTok, which is not indexed. It might have originated on WhatsApp or Signal or any other closed group,” said Bal Krishna. “Using Backstory you should keep these limitations in mind. We always do.”

Due to these current shortcomings, Bal Krishna says India Today does not cite Backstory’s AI-generated reports directly in its published fact-checks, and reporters are expected to manually confirm any information they pull from the tool.

Still, he considers Backstory the “first robust tool made by a big tech company that was offered to fact-checkers” for image verification, and is hopeful it will only improve as feedback from testers is folded into the product.

When I asked if Backstory had sped up the daily work of his fact-checkers, he responded, “There is no doubt about it.”

But one publication is over it. If you want to write for The Food Section, you’ll have to pick up the phone.

“We’re switching to a pitch-by-phone system come September,” founder Hanna Raskin announced on Tuesday. “Having been deluged by AI-generated and enhanced story ideas this summer, we’re re-humanizing the process with a dedicated hotline (which has the added benefit of weeding out supposed reporters who aren’t in the habit of ringing up strangers.)”

The Food Section reports on food industries and issues in the American South, with bureaus in Asheville and Charlotte, North Carolina. It was the first solo newsletter to win a James Beard award and the first newsletter to win a National Magazine Award, according to its About page. Raskin currently takes story pitches through a Google Form.

Raskin told me that she’s noticed an increase of AI-generated story pitches since The Food Section upped its pay rates for freelancers on July 17. Raskin estimates that a quarter of recent pitches were “wholly AI, meaning neither the writer nor the proposed story appears to exist in the real world.”

“Like most editors on the receiving end of slop — some of which was disturbingly calculated to appeal to a publication that prides itself on exposing social injustice — I found the situation incredibly depressing,” Raskin said. “What was especially worrisome was I felt myself becoming more inclined to dismiss pitches from writers without sterling portfolios or connections to journalists in my professional circle. That’s hardly compatible with cultivating new voices or encouraging young reporters.”

Raskin plans to keep the pitchline open at designated hours and will see what happens. (“I’m not a fool: I’m sure there is an AI way to game this system too,” she noted. “But my hunch is it will work until it doesn’t.”) She’ll also coordinate with journalists to take pitch calls at times that work for them.

“We understand that many great writers — meaning the kind which aren’t AI — have hourly-wage jobs and other unmoveable obligations,” she said.

Media workers continue to battle the onslaught of AI pitch slop. Last summer, Wired and Business Insider retracted published stories by an “AI freelancer” named “Margaux Blanchard.” In July, Futurism reported that PR firms are using AI publicists to spam journalists with pitches.

Raskin plans to share the phone number in The Food Section’s newsletter next month. You can read the publication’s full pitch guide here.

“I follow Coach,” the seller told Dominic. “$120?”

@theathletic Follow along with @COACHKOE │ SPORTS CARDS at The National Sports Collectiors Convention as he helps a young collector on his journey to find a Caleb Williams card! Sponsored by eBay. #chicagobears #calebwilliams #sportscards #nfl ♬ original sound – The Athletic

The Athletic published the 97-second video of Coach Koe and Dominic to TikTok. It’s one of around 75 videos The New York Times’ sports site has published with sports-focused creators since launching a pilot partnership program last September. Others include explainers, like “How did this 21-year old go from playing college golf to beating Scottie Scheffler less than two months after turning pro?” and another on “one of the most expensive pieces of grass ever sold.”

Ahead of the World Cup this summer, The Athletic also produced videos asking sports creators in different countries about their tournament predictions and favorite soccer moments. And at the start of last year’s NBA season, The Athletic partnered with the social media–based basketball publication @halfpast*noon on a post about breakout stars.

All told, the videos have gotten a combined 50 million views, said Adam Park, The Athletic’s director of video growth. Since last September, The Athletic has also gained 100,000 followers across its different platforms.

The Times has joined forces with creators in the past. Last summer, The Athletic inked a seven-figure deal to license “Pablo Torre Finds Out,” a sports journalism podcast by Pablo Torre. NYT Cooking has also been partnering with independent food creators to diversify its video content and reach new audiences.

Most of The Athletic’s video content is still made in-house with its journalists, but the publication sees creator partnerships as a way to find new audiences. Younger people, in particular, tend to prefer getting their news via video, rather than reading or listening. Park said the existing audience of The Athletic is “quite even across age demographics” but noted the creator-led videos were helping the sports news site work with younger creators that have developed similarly young audiences. The Athletic declined to share details about its audience demographics.

“We have internal creators that host, do podcasts, or social video, but we really wanted to tackle the idea of taking a large-scale newsroom and translating that into more of the human, personal touch of a creator,” Park said. “How do we translate to younger audiences, and make our newsroom accessible to everyone? This felt like the easiest way of doing that.”

Finding the right creators to collaborate with means “a lot of watching TikTok and YouTube,” head of creative development Craig Custance said. Park and then-video growth manager Jordan Pun (who now works as the head of social analytics for MrBeast) compiled lists of creators whose content they thought would mesh well with The Athletic’s brand.

“It was really important to us, as we started connecting with creators, that we found people that wouldn’t feel completely off brand,” Custance said. “The Athletic is not a place of hot takes. If we were to bring in really strong, opinionated creators, it just wouldn’t fit. If we’re connecting with their audience, it would just be a mismatch. We had a lot of conversations about whether the people we’re reaching out to match our journalistic DNA, our storytelling, and our deep analysis.”

The workflow relies on lots of communication between the Times newsroom and the creators, Park said. Sometimes the creators pitch ideas to editors, who will provide feedback and previous stories The Athletic has published on the subject. Other times, editors will flag newly-published stories with creators who might be interested in talking about the reporting in their videos. Creators are paid for their work, though The Athletic declined to provide details or payment rates for the partnerships.

One creator they worked with, Frank Michael Smith, was proactive in finding stories that worked for both his audience and The Athletic, Park said. Smith, who covers several sports and has more than two million TikTok followers, pitches his ideas, scripts the videos, then sends them to The Athletic for editing and a fact-check. Once the edits are in, he produces the videos and files them. The Athletic also has editors who oversee the particular sport or team mentioned in the video review the piece.

Smith then publishes them with The Athletic as a collaborator. Between October 2025 and April 2026, Smith published 24 videos with The Athletic, pulling in 27 million views in total.

“What we built was a blueprint: proof that the creator economy and legacy media aren’t competing,” Smith wrote on LinkedIn in April. “We’re much more powerful together.”

Vertical video — both with journalists and creators — has been a commercial opportunity for the publication. EBay sponsored at least two videos from the Sports Card Convention, including the Coach Koe video. The Athletic’s daily World Cup video recaps, anchored by different journalists, were sponsored by Amazon Prime. Fan interview videos on the street were sponsored by Apple TV. One video, where Smith explains how Spain and Argentina stacked up against each other ahead of the World Cup final, was sponsored by Kalshi. Other creator videos are unsponsored.

The videos aren’t straight advertisements for those companies. Instead, Custance said, the ideas for the videos originated in the newsroom with editorial purpose, and advertisers were approached afterwards.

“We were going to do [the videos] anyway,” he said. “It’s like getting an ad on your story. Now we’re generating revenue with it as well, which is ideal.”

Today, The Athletic usually has at least one creator on its production calendar to produce content, Park said. The Athletic is currently hiring for a video growth manager to lead and expand the creator program.

“We started out as a longform, written company and have evolved constantly since day one,” Custance said. “For us, this is how the audience wants to consume sports news and analysis — on TikTok and Instagram.These [creators] are the people that have found the most effective ways to do it. How can we tap into that and expand our journalism in that way?”

AI providers have leaned into the idea of adding watermarks to the content they generate — sometimes visible to humans, sometimes visible only to a dedicated tool. But adding a watermark to visual media is a much less sticky problem than adding it to text — text that might be edited or removed from its original context. Other than looking for “It’s not X. It’s Y,” how can we know if a set of words we read came from a human or a chatbot? European regulators are forcing Big AI to come up with answers.

Last week, Anthropic announced that it would begin adding invisible watermarks to the text it generates, and on Friday, it explained how it’s going to work. Daring Fireball’s John Gruber has a good, clear piece that explains both the methodology and how it, in his mind, perverts the act of writing. In essence, it involves Claude intentionally changing some of the words it’s generating to less-likely options — picking “guava” instead of “mango,” say, or “auto” instead of “car” — in patterns that Anthropic can then detect. Some people don’t like it; Gruber hates it:

One of my fundamental problems with this is that no two synonyms carry the exact same meaning. “He leaped at the chance” and “He jumped at the opportunity” are very similar sentences expressing the same general sentiment, but they are not the same. The exact words we choose when writing matter. I want any LLM I use to choose the very best, most precise words at every single decision point. An obvious constraint that I accept is time and computation. Within the constraint of executing inference quickly, and at a certain cost per token, I want the best words. This constraint matches human writing. I could surely write a better column by taking longer to write it. I write with a sense of how much care I should put into every word and punctuation choice I make. I take more time with certain paragraphs, sentences, or even individual word choices when my gut feeling says I should.

In other words, these are necessary trade-offs. These factors are all in my interest: speed, cost, quality. Ideally I would like perfect writing, at instantaneous generation speed, at zero cost. None of those things are possible. Computation is not free of charge (and cloud-based LLM inference with leading models is actually expensive). Inference is not instantaneous. And great writing, whether natural or artificial, can only approach perfection.

The idea that anything other than my needs should factor into the generation of text for me is patently offensive.

Whether it’s more or less offensive than any other AI-generated text is an exercise left to the reader.

An animated phone vibrates on a blackened screen as ominous piano music tinkles in the background. When you click, speech bubbles appear. “It’s Thanasis Koukakis,” the person on the other end says. “I’m borrowing a phone…Just don’t try to reach out to me for now. Texts, calls, try to avoid it.”

This is the opening scene of Predatorgate, a new video game based on an investigation into the Predator spyware scandal in Greece. Koukakis was the reporter at the heart of that real-life story who discovered his phone had been infected with the highly invasive spyware. This narrative game is based on the investigation by Inside Story, and shows what it’s like to be in the shoes of a journalist under threat of illegal surveillance. In the game — which takes an estimated 30 minutes to play — Koukakis is a colleague who discovers his phone has been targeted.

Z. Daniel Barnett is a former investigative journalist and voiceover artist who worked on the Predatorgate game project with designer and developer Izzy Fiacco. “A game that deals with the concept of data security and also helps people to think critically about the news they’re reading, and the process that goes into making that news, is super important,” he said. Before this game, Barnett had never really considered how games and the news could be brought together. Predatorgate made him realize how a new format can bring stories about investigative journalism to new audiences.

“I think gamifying people’s understanding of journalistic investigations is really important,” he added, “because we live in a time where people’s media literacy is maybe lower than it’s ever been.”

Predatorgate, which Barnett has called “a love letter to old school journalism,” won the top prize at the Floodlight Investigative Journalism Game Jam in June this year. Run by the Organized Crime and Corruption Reporting Project (OCCRP) and the Gabo Foundation, the idea behind the project is to encourage game developers to transform investigative stories into video games. Two games from the first edition are already being developed into commercial versions.

“Investigative reporting is such an amazing source of storytelling for video games because investigative reporting is full of dilemmas,” said Paul Radu, the co-founder of OCCRP, who is also co-creator of the Floodlight gaming initiative. “There are choices that reporters make all the time, and people get to experience these dilemmas and get to understand what investigative reporting is all about.”

And gamifying investigations, he added, helps investigative reporters reach new, younger audiences. “Video games are such a huge part of how society is shaped,” he said. “It’s how young people learn the world around them.”

An emerging market

More than three billion people globally play some form of video games. For a media industry with falling revenues and growing audience distrust, games — ranging from simple quizzes to fully-fledged downloadable games and augmented reality experiences — have become a new way to engage and inform audiences.

Experts say games increase visits to the outlet, increase the time people spend on a site, and promote loyalty and brand connection. They can also directly and indirectly help with monetizing through boosting subscriptions or drawing advertisers. Conor Dean, founder and CEO of GamesGrid, said in a webinar last year that games are going to become “just as essential as newsletters or podcasts” for publishers that are “losing audiences and advertising revenues.”

In the U.K., the Financial Times launched The Uber Game, where players stepped into the shoes of Uber drivers to see if they would survive the gig economy, a game based on actual interviews. The challenge: figuring out if you will make enough to pay your $1,000 mortgage payment, which is due in a week, amid competing demands on your finances, and moral ones, too, like seeing your children.

In Nigeria, the Centre for Collaborative Investigative Journalism (CCIJ) turned an investigative series about corruption in the 2023 Nigerian presidential elections into Data Vault Canyon, a multi-player Fortnite game.

“We knew that if we are talking about voting, elections, and participation, that young people are not going to be reading 3,000-word stories,” said Nelly Kalu, product and innovation manager at CCIJ. “So how do we get them to be interested and stay engaged? It doesn’t matter to us if they read the stories, as long as they understand the value of protecting their votes.”

Similar concerns exercised the team at the independent digital outlet Cuestión Pública. They have developed four games so far, including one that gamified publicly available data on the country’s previous elections into Game of Votes. “People are like, ‘Oh, this [story] is long and boring,’ or ‘Oh, this is too complex. I don’t understand it,'” said Edier Buitrago, the data editor and tech leader. For two of their games, they drew inspiration from beloved offline games like Guess Who and Cluedo to familiarize people with the large cast of officials, politicians, and families that dominate the country’s political landscape and have been involved in major corruption cases.

Newspapers have long had games sections, whether that’s a daily crossword, Sudoku puzzle, or a jumble. Games ensure habit and familiarity for the daily reader and build a continuing loyalty and relationship with a media outlet.

“Games usually bring highly engaged interactions to the outlet,” said Sisi Wei, chief impact officer at CalMatters, a California-focused nonprofit newsroom. “With an article, you might read it for some amount of time, but with a game, that amount of time goes up significantly.”

As with the Nigerian and Colombian cases, there’s often a practical, news-you-can-use element to the game. But games can also be built around humanitarian issues that don’t necessarily feel connected to a person’s daily life. For instance, Reality Reload, a game studio founded by a journalist-turned-game-designer, created a simulation around the American work visa process for immigrants. Built from interviews, players had to navigate hurdles and practicalities that an H1B visa applicant would.

Similarly, ProPublica created The Waiting Game based on deep reporting of the lives of five asylum seekers and invited readers to experience that for themselves. It turned asylum-seeking from an abstract tragedy into a lived reality, in a way that transcended what a text piece would have done. “That was a powerful way of creating empathy for the people whose stories we were telling,” said Wei who was then at ProPublica and reported the story. “The written word, video, audio, photography, all those mediums are things you consume,” said Wei. “But games, as a medium, are inherently interactive, and if you are invested in a storyline, you are asked to make a choice in order to progress the game.”

Crafting a game takes time and effort and involves artwork, music, sound design, playability, and needs to be finessed through multiple iterations. When you are dealing with real life, and real people there are also other considerations. “Should we use real names? Should we use completely fictitious names?” Pol Grasland-Mongrain remembers discussing with his teammates when designing Influence, a game that was a finalist at the Floodlight jam. Eventually they fictionalized a few names for their game, which puts the player in the shoes of a Mexican journalist.

Choices, conflict, tension, challenges, characters, a process to navigate: these are some of the key features that a news story must have to translate well as a game. “For instance, if I was writing a story about agricultural commodities in Midwest America, that’s not a game, right?” chuckled Barnett. “There’s no dramatic tension there.”

For a longer video game, a narrative arc is crucial, with not just the story itself, but also the process of how the journalist investigated it. For instance, Hunting the Hunter, the previous Floodlight winner, was based on a Ghanaian journalist’s undercover exposé of illegal cocoa smuggling. And the runner-up Dark Money drew on the OCCRP’s money laundering stories.

There’s no set formula that can ensure a good game, and individual tastes vary. But there has to be an element of fun. “If it’s too serious,” said Julie Dupiat, a French designer who worked on Influence, “it is not a game.” At the same time, the game should not trivialize the subject.

Some outlets, for instance, draw a hard line on what kinds of stories should not be gamified, even if they have the narrative potential for it. “Why would you want to make war a game? Why would you want to make abuse a game?” asked Kalu. “That’s a moral choice that you must make as a journalist, an ethical one, to not do that.” Similarly, at the game jam, Radu told developers: “You have one rule: don’t glorify organized crime. That’s it. The rest is up to you.”

Time commitment vs. return on investment

Games can take anything from a few hours to several months to complete, depending on the complexity, style, and cost. Data Vault Canyon, for instance, would have been costly to develop, but CCIJ cut costs by doing it in-house. Their metaverse designer had to retrain herself in different directions and experiment with various free online tools. “Our goal now, as we’re looking towards the future, is to be able to buy the tools that are needed,” said Kalu.

Experts say it’s also easier and cheaper to first try and build browser games that can be played on an outlet’s website itself. Wei advises beginning with simple games that break down practical information for audiences: navigating the school system, for instance, or deciding how to vote. “The news-you-can-use type of game is maybe more practical to start with,” said Wei, “because it gives the reader something very tangible they can get from it, and it’s a less daunting task for you because it’s often easier to communicate the information they need.” At a game jam some years ago, for instance, Wei created Heartsaver to depict how people who needed hospitals were maddeningly far away from them. It took just two days.

Ultimately, one one of the keys is collaboration and finding the right partners. If you are a journalist, “you’ll know what makes a good story, you won’t know what makes a good game,” said Barnett. “And on the flip side I think if a game developer wants to make a journalistic story [into a game], find a reporter to talk to.”

The key, says Kalu, is opening your mind to the possibilities. “We now look at all our stories with the question: is there a potential for productivizing or gamifying this?”

Chicago Public Media is already the parent nonprofit of two existing news sites, WBEZ and, since 2022, the Chicago Sun-Times. For chicago.com, though, the news org is teaming up with New_ Public, a nonprofit working to reimagine online public spaces, to launch a network of online communities for Chicago neighborhoods. Chicago.com (the enviable domain name was donated by the Chicago Community Trust) will host the network, which will be powered by Roundabout, the local community platform incubated by New_ Public over the past year. The partners will launch five neighborhood pilots this fall, and will recruit paid, contract “community stewards” to moderate conversations in each neighborhood.

“This is a community platform, not a news site — it’s very much not a news site,” Chicago Public Media CEO Melissa Bell said. “It’s crucial that we don’t think of this as a publishing platform.”

In a challenging national environment for media, Chicago has a relatively rich nonprofit news scene that includes robust neighborhood-level news coverage. But information access, and civic engagement, varies widely across the city’s neighborhoods, according to a civic information needs census of Chicago conducted earlier this summer in partnership with Chicago Public Media.

For a while now, publishers everywhere have been confronting the bleak reality that, with more and more people turning to social media for their news, news websites — like newspapers before them — are at risk of becoming obsolete. But what if a news website could be something fundamentally different? What if it could beat the platforms at their own game and offer people sick of doomscrolling something better and more useful?

Those are big questions where ‘if’ is the operative word. Building something different requires experimentation that could end in failure. But Chicago Public Media is, at the very least, asking those questions with its vision for what chicago.com could be. It hopes to build a participatory community resource that informs everything from where to volunteer, to plans for the weekend, to how neighbors can meet one another.

This vision for a role a local news org can play — a provider of useful local information (not just news) and a space for community conversation — is attracting energy and attention elsewhere, too. Village Media, the Canadian organization Chicago Public Media has pointed to as one model for what it hopes to build, just received $5 million from the Knight Foundation to expand to 15 communities in the U.S. (Chicago Public Media has also cited Front Porch Forum in Vermont, Zetland Media in Denmark, and the Daily Maverick in South Africa as role models.)

Bell cofounded Vox, a product of a different era of the internet, and had connected with some of the people behind New_ Public and Roundabout in that past life, before joining Chicago Public Media in 2024.

“I think all of us grew up at this time where we were inspired by the opportunity of the internet, and I have seen a lot of the challenges of what has happened to our digital space over the last 15 years,” she said. “When I came to Chicago Public Media, I really wanted to think about not just sustaining and stabilizing the great products that we have with WBEZ and Sun-Times, but really thinking about the future of public media overall. I couldn’t think of better partners than folks who were really focused on how to evolve the next stage of the internet and how we interact in digital spaces.”

Chicago Public Media and Roundabout don’t want to call all the shots on how chicago.com evolves; it’s something they’re explicitly inviting community members to “build with us.” But to understand what the platform’s network of neighborhood communities could look like, it’s helpful to start with the Roundabout pilots that already exist.

Roundabout has a presence in cities including Richmond, Virginia, and Portland, Oregon; the team showed me anonymized versions of its interface. It has many of the hallmarks you picture when you think of a social platform: a feed, posts, reaction buttons. The feed is ordered reverse chronologically, not algorithmically, and can be filtered by topics like arts & culture or pets. There’s a separate, second feed that features posts the community steward has flagged to bring to members’ attention. Posts might be introductions, or events, or requests or offers of help, or steward-posted discussion prompts or guides. There’s also a calendar feature, where users can submit or look for events.

One thing you won’t find: Profiles. The focus is on the shared conversation, not individual members’ digital presence or following, so community members don’t have profile pages.

Reaction buttons on the posts are designed to be “pro-social.” Community members can mark that they’re following a post, or mark it as helpful.

Posts are automatically moderated by an AI-powered tool that checks them against community guidelines, though posters can also flag posts for review by the human steward. Roundabout, and chicago.com, aspire to be spaces for real community conversation, including substantive disagreement. (In an attempt to avoid the toxicity on platforms like Nextdoor, no posts about “suspicious persons” are allowed.)

Stewards aggregate events and other PSAs from Roundabout into a newsletter that non-members can sign up for. Nationally, Roundabout now has more than 6,400 newsletter subscribers and Roundabout accounts in total across its pilots, according to Trei Brundrett, Roundabout’s founder-in-residence at New_ Public (and, like Bell, a Vox alum).

This is a digital platform designed to ultimately bring people together at local in-person events. “Our top priority is that people can find what they need and then take that info out into their community,” Bell said. She noted that corporate-owned social media platforms “are typically trying to keep [users] on there as long as possible, in order to show them as many ads as possible. We do not have that incentive for chicago.com.”

For the past year, artist and researcher Liza Pittard has been one of two stewards for Richmond’s Roundabout community, which includes a little over 1,300 members. (Stewards in Roundabout’s pilot communities are paid by New_ Public.) Many of the people she’s seen join, she said, are new to town and looking for connection. She spends about 15 hours a week on her role as a steward, she estimated, though it varies by week and tends to be more when she’s also organizing an in-person event. Much of that time is spent combing other social platforms for events and picking out the best ones so Roundabout community members don’t have to do the legwork of looking in 20 different places.

A big part of her role is “researching and becoming more aware of what’s happening around town,” she said, “and what kinds of conversations people around town are having outside of Roundabout.”

Brundrett sees community stewards as more than just moderators — almost like a “really great dinner party host,” he explained, whose role is more proactive than reactive. “It’s somebody who is thinking about, ‘hey, what do I want this to be?’ And they’re inviting people in, they’re setting the tone.”

Brundrett described the early joiners of Roundabout as “barn raisers” — people who are already civically engaged in some form, which might include holding a stake in local life as a local business owner. But he’s also excited about Roundabout’s potential to serve people who might be less civically engaged and are just looking for something to do this weekend.

Roundabout’s team sees these communities as places local reporters can be invited into to listen and build trust — places where they can answer people’s questions and learn what they should be reporting on. Reporters, and reporting, are not central to what Roundabout is; they are components that can strengthen the local community conversation that is central.

The primary unit of information, at first, will be “mostly text-based,” Bell expects. But she hopes chicago.com will grow into something that serves people in their preferred information medium, whether reading, listening, or watching (the landing page features short videos).

Chicago Public Media has not yet revealed which five neighborhoods it will launch pilots for, but said it’s looking for those that represent a diverse geographic, economic, and cultural cross-section of the area. “Our approach has been to find the strongest local moderators with the skills to keep community discussions constructive, safe, and civil, while focusing on areas with a deeply rooted sense of local identity and shared history, which research shows is critical to driving active participation,” Bell said.

When chicago.com debuted in July, the site got more than 1,000 email signups within the first two days; a third of those people weren’t already on the WBEZ or Sun-Times lists.

The goal is to keep chicago.com free to access and support it with a mix of memberships, sponsorships, and donations. “Business model innovation is an important piece of this project, and we’re working hard to develop that, but we can’t share many more details at this time,” Bell said. Chicago Public Media’s publications have faced layoffs and buyouts in the past two years, but last month Bell told the Sun-Times financial projections for 2026 suggest the organization will be “close to break-even.”

Bell hopes chicago.com will appeal to Chicago Public Media’s existing audience while also breaking new ground for them. And she hopes creating a social-media-style, participatory information hub corrects a mismatch in available news mediums and people’s actual information habits.

“We want to make sure that we’re providing [people] that information in a format that creates a daily habit like we used to have when you would wake up and get the newspaper from the front step, or turn on the radio and listen to it on your drive into work,” she said.

“If you think about the idea of a newspaper 30 years ago, it was news but it was also local information. We would look for the movie listings in the advertising section…You would have letters to the editor. You would have local businesses telling you what was happening that weekend,” she added. “I think that bundle idea is what we are trying to recreate, but just in a format that makes sense for today.”

Apple has approached publishers to use their news and up-to-date information in the revamped Siri AI, sources told the Journal. The company has reportedly discussed a “nine-figure budget” for the payments.

The next iteration of Apple’s voice-powered assistant was first promised in 2024 and is now expected to be released to the public later this year. (Beta reviewers seem to think it’s an improvement.)

Apple has proposed “a variable compensation plan” to news publishers, according to the Journal. The pay-as-you-go plan would see publishers paid only when Siri AI uses their content. It would be a break from other AI licensing agreements that have involved fixed fees and wide-ranging access to content. Apple currently makes payments to news companies for Apple News+ using a similar model, paying publishers “proportionate to the share of clicks they generate within the platform.”

The raid led to multiple federal lawsuits against the county government, public officials, and the city, accusing them of retaliating against the newspaper for its coverage. Last fall, Marion County agreed to pay $3 million to three journalists and a city councilor, settling some of those lawsuits. And on Tuesday, the third anniversary of the raid, the city of Marion settled with one of those reporters, Phyllis Zorn, agreeing to pay her $850,000, the Kansas Reflector reported.

The costly settlement is a vindication of press freedom at the local level in a year and political climate in which that freedom faces unrelenting stress tests. The raid is covered in the documentary “Seized,” which premiered at Sundance in January.

Other lawsuits against the city remain outstanding. But for Zorn, who retired and moved elsewhere in Kansas after receiving settlement money from the county last fall, the latest settlement ends a yearslong ordeal that has taken a severe toll on her health.

“They picked on the wrong people,” Zorn told the Reflector. “I think they thought they would get away with it because they’re some nothing little town in rural Kansas. But we were a newspaper. We have access to immediately draw the attention of the whole nation.”

Natural disasters have long been flashpoints for disinformation in Japan. Similar falsehoods surfaced online after a major earthquake hit Kumamoto Prefecture last month. But the 2024 Noto earthquake was a call to action for a coalition of Japanese news publishers, advertisers, and technology companies that was already pushing for better tools to fight misinformation.

Their solution: a new web standard called Originator Profile (OP), which helps ordinary users verify who published or created a website.

OP uses cryptographic digital signature technology to create tamper-proof credentials for websites. Each ID card confirms who published the content, whether professional organizations vouch for that publisher, and information about the publisher’s editorial and ethical standards. The exact user interface is not finalized, but a new browser extension released last month displays the information in a pop-up sidebar.

OP isn’t meant to judge whether all of the content on a website is true. It doesn’t authenticate specific photographs or video clips. Instead, it authenticates the site itself — for example, confirming that a news article is genuinely coming from the news organization it claims to be.

“We aren’t trying to dictate whether the information is accurate or not. It’s all about provenance: Where did this information originate?” said Makoto Yoshiike, deputy secretary general of the Originator Profile Collaborative Innovation Partnership (OP-CIP), the group advocating for the standard’s adoption.

OP is far from the first attempt at building new provenance tools in an information ecosystem flooded with AI slop, deepfake scammers, and “LLM poisoning” campaigns. Watermarking tools like Google’s SynthID and cryptographically signed metadata like the C2PA standard aim to tackle similar problems, but so far have mostly focused on addressing the rise of AI-generated imagery and audio.

“OP works within the HTML — it’s focused on the web page itself,” said Yoshiike. “We don’t want a whitelist for good media, separated from bad media. We just want to say it actually came from the authenticated sender of information.”

To develop the technical specs for OP, Dentsu partnered with Jun Murai, a professor at Keio University who is often credited as a founding father of the Japanese internet. He developed the web standards that allow Japanese language characters to travel through email and internet messages, among other innovations.

Next, Dentsu got news publishers on board. The largest daily newspaper in Japan, The Yomiuri Shimbun, was a founding member of OP-CIP. From the news industry’s perspective, there was a clear need to authenticate webpages that might be used as sources, but also a need to protect their brands, at a time when generative AI tools have made it easier than ever to spoof legitimate news publications.

Yomiuri itself has faced this challenge. “I actually caught one red-handed on Facebook,” said Yoshiike, noting that he’s been served fake versions of The Yomiuri Shimbun on his feed. “It’s still happening, individuals mimicking our article pages and luring users to fraudulent websites. Most of it is junk, but it happens every day.”

In December 2022, OP-CIP was officially established with nearly a dozen news publishers on board. The Noto earthquake in 2024, however, expanded the project’s scope. The Japanese national government, realizing the value of the authentication tool for government websites that deliver messages in crisis situations, began funding OP-CIP, and the organization started welcoming local governments into its ranks.

“When disasters strike, there is no end to fraudulent information being disseminated, individuals impersonating local governments or public utilities,” said Tatsuya Kurosaka, a project associate professor at Keio University and secretary general of OP-CIP. “That’s why we are thinking not only about the media context, but also public information more generally as well.”

Today, OP-CIP has built an impressive coalition of supporters. Among them are dozens of members of the Japanese Newspaper Publishers and Editors Association (Nihon Shinbun Kyokai). Major television networks, including NHK, the national public broadcaster, have also signed on. Members of the group from the tech sector include LINE-Yahoo (LY), which operates the most popular news aggregator in Japan and the messaging app LINE, which is used by more than 80% of the country.

While those members publicly signed on to support OP, so far only a handful have fully integrated the technology into their own operations. The Yomiuri Shimbun and The Asahi Shimbun, the two largest newspapers in Japan by circulation, have integrated OP into their content management systems, according to Yoshiike, who also works at The Yomiuri Shimbun as the senior deputy chief officer for the president’s office. Tottori Prefecture in Western Japan has also launched OP on its prefectural government websites. Dentsu is currently rolling out the technology for its ad buys, with hopes that it can help counter the recent sharp rise in ad fraud and AI scam ads across the country.

Yoshiike says that by early 2027 OP-CIP plans to have 50 major organizations operating with the OP standard in place.

Ultimately, though, the group sees Japan as its testing ground. In the years to come, they hope to export the technology and make it an international web standard, with the participation of international news publishers and global technology companies. The World Wide Web Consortium (W3C), which is responsible for existing global internet standards like HTML and CSS, has already created its own OP working group to explore this possibility.

“We’re pushing full speed with implementation in Japanese society to show the W3C community how it works,” Yoshiike said.

Last month, OP-CIP released its first public tool to retrieve and visualize the cryptographically signed profiles on websites. The browser extension, called OP Inspector, is available on Chrome and Firefox. When a user lands on a Yomiuri Shimbun article, for example, the extension activates a pop-up sidebar that shows a badge verifying that Yomiuri Shimbun is the organization operating the webpage and links to the publication’s editorial standards policy and privacy policies.

Right now, the extension is only meant for developers working with the OP standard, though there are plans to fine-tune and release an extension for general internet users down the line. Eventually, OP-CIP hopes it can abandon the need for extensions and apps altogether. One current idea, if the standard is taken up by W3C, is for a badge to appear within a browser URL bar so that users can click it and access the OP without ever downloading an extension.

“My mother is 88 years old. I’m designing OP technology to be simple enough for even my mom to use,” said Kurosaka. “No specific skills are required.”

The project faces clear hurdles as it moves into the next stages of development. One will be adoption by social media platforms that are major vectors for disinformation on the internet. Currently, OP can’t appear on individual social media posts like article links shared on Facebook. It’s only accessible when a user navigates to that article’s webpage, limiting its impact on social media-fueled disinformation.

“I think it is a huge open question,” said Yoshiike. “Our technology side hasn’t caught up with the current social media standards. OP works within the HTML, so it’s not used in the application itself.”

Fully realizing OP’s potential will also require earning the support of tech giants like Google and Apple that hold tremendous influence within W3C. As search increasingly becomes a direct source of news for users through products like AI Overviews and AI Mode, their participation becomes even more essential.

“The last thing standing between us and the users is big tech companies,” said Yoshiike. “We know that it’s not going to come easy.”

As any Nieman Lab reader knows, journalism has seen drastic changes since 2014. One major shift in the last 12 years? About 20% of Americans now turn to news and commentary from influencers, content creators, and others on social media and platforms. Many of those do not self-identify as “journalists,” let alone subscribe to a professional association’s code of ethics, notes the SPJ committee that spearheaded the proposed revisions. The committee also highlighted the “dramatic rise” in the use of generative AI in newsrooms, citing a study that found 56% of journalists now use AI tools at least weekly. As AI-generated content is increasingly published by news organizations, the committee also says this introduced “persistent ethical and practical questions about the use and misuse of technology.”

One striking thing about the revision is that “AI” is actually nowhere to be found in the new code of ethics. In a report accompanying the proposed revisions, the committee writes that the term was sidestepped due to concerns it might “date” the code. (The SPJ omitted “social media” in its 2014 code for similar reasons.) Even so, these technologies motivate many of the changes and additions that were made, according to the report.

For one, the committee has added a whole new tenet to the job: “Identify misinformation and disinformation. Challenge provable falsehoods and avoid uncritically repeating them.” It is an implicit acknowledgement that the work of journalists increasingly revolves around authentication and verification. That’s no doubt in part because of the emergence of social media as a powerful vector for disinformation since 2014. It also calls attention to the responsibility of journalists to combat the recent flood of AI-generated imagery online.

There is another hat tip to this work in a revision that reads, “Journalists should verify the provenance and authenticity of information and sources.” The committee said it added phrases like “verify” and “provenance and authenticity” to add specificity for the “digital and AI age,” and strengthen the language around this work.

The new code also recognizes that journalists can be complicit in the spread of falsehoods and misleading imagery online. “Never publish fabricated content that could be confused with reality,” reads another revised line. Here “content” was added to “better address AI-generated images, graphics, and the misuse of AI in image production,” according to the committee.

The closest the code gets to naming AI outright is the addition of the word “technology” to one of its tenets. “Remember that speed, technology or medium do not excuse inaccuracy or unfairness,” reads the new copy. The committee explained that the addition is meant to address AI, but also “whatever technology may come next.”

Tacked onto that line is also a new sentence: “Journalists are ethically responsible for their work, regardless of tools and technologies used.” Here the committee says it’s conveying, in part, that human journalists are ultimately accountable for what they publish, even if their work is generated or assisted by AI.

The proposal caps a yearlong review process by a blue-ribbon committee. Now, SPJ members will have the opportunity to review, discuss, and comment on the proposal leading up to the SPJ’s national convention in October. There, the association will hold a vote for all members on the new code.

You can read the full proposal with the committee’s comments here.

We wanted to know how this plays out in comments sections. In a paper recently published in Journalism Studies, my coauthor Sheizaf Rafaeli and I analyzed nearly a million comments on 6,400 New York Times stories published between 2014 and 2022. We compared pieces from The New York Times’ data journalism section, The Upshot, to more traditional pieces published during the same period. (I am a computational social scientist who studies how information environments shape who participates in online public discourse. Rafaeli is the president of Shenkar College in Israel.)

Data journalists have sometimes wondered whether loading a story with data can be overwhelming rather than clarifying. We found that giving readers more information changes who comments, and how.

More information, more engagement

For each story, we measured several things: How information-dense it was; the numbers, sources, and charts it used; and how analytical the writing was. Then we looked at the comments on those stories. We tracked how many people commented, whether they were interacting with each other or only reacting to the story, and how much their comments leaned on sources and language that read as analytical versus emotional.

We found that stories combining statistics, external sources, and data visualizations produced a different kind of comment section. Regardless of the topic of these posts, readers tended to reply to each other. That raised another question: What were readers actually saying to each other, and how many of them were showing up to say it?

The answer turned out to hinge on something researchers call “affective style” — how much a comment is driven by feeling rather than reasoning. For example, in a comment on a 2014 Upshot piece about the World Cup, one reader wrote: “Typically, people hate good teams because they win too much and laugh at bad teams. So, if this is true, soccer fans are truly unique.” That’s affective — a feeling dressed up as an observation.

Under a different Upshot piece, this one on the shrinking American middle class, a reader wrote: “From 1950 to 1980 we had thirty years of demand side economy. From 1980 on we’ve had thirty years of supply side economy. Take your pick.” That’s not affective — it’s an argument, laid out for the reader to weigh. Denser, more analytical stories produced far more of this second type of comment.

The catch

The same stories that produced sharper, more analytic conversation also produced less conversation. The more analytically a story was written, and the more statistics it contained, the fewer people commented on it. Emotional stories did the opposite, drawing bigger crowds.

The presence of charts mattered less than you might expect. Stories with more visualizations tended to draw comments that were more analytic and better-sourced, but once we accounted for everything else a dense story was doing, the chart alone was not what moved a reader from reacting to reasoning; rather, that was the statistics and writing style.

Comments on denser stories were more likely to lean on outside sourcing and structured arguments. Researchers have connected this pattern to the digital divide, identifying a “data elite” as the subset of an audience that keeps engaging when others opt out. That tracks with something recent Pew research found about civic participation generally: People sort into distinct groups. As a comments section grows more reasoned, it also becomes more exclusive.

Most people who visit a comment section never post in it, and a section full of careful, evidence-based argument is a different thing to encounter from one that isn’t, whether or not you participate. Producing that kind of journalism costs real money, statisticians, data cleaning, chart-checking, extra editing, and that expense pays for what happens underneath the story, not just the story itself.

That is worth keeping in mind as newsrooms reopen comment sections after years of shutting them down. If the payoff is a more loyal, more engaged readership, it is not spread evenly across a newsroom’s output. Start with the hardest stories, not the easiest ones that will always draw a crowd on their own. That is where the more reasoned conversation was hiding.

“While USA Today Co. CEO Mike Reed championed this partnership as a way to drive revenue, we believe it threatens to undermine trust in our news outlets and raises serious questions about the data security of our readers,” the statement reads, in part. “We have significant ethical issues with Palantir, whose artificial intelligence (A.I.) software has been used to advance widespread surveillance and power immigration crackdowns. Not only does partnering with Palantir, a major player in the news we cover, create an inherent conflict of interest, but it fails the journalists who have been assaulted and falsely arrested while covering immigration enforcement actions and protests.”

BREAKING: USA Today’s journalists and media workers demand that the media company end its relationship with AI firm Palantir.

— The NewsGuild-CWA (@newsguild.org) August 10, 2026 at 12:01 PM

USA Today Co. — formerly known as Gannett — announced the partnership during its second quarter earnings call on August 6 as a way to “strengthen how we collect, connect and activate audience data to drive more effective and faster monetization across our platform.” Palantir’s software would be used to analyze and monetize customer behavior amid an industry-wide decline in search traffic.

In the statement, unionized employees (most of whom are journalists) write that USA Today Co.’s “commitment to upholding First Amendment principles to serve the democratic process is in direct contrast with a company like Palantir, whose cofounder, Peter Thiel, once said freedom and democracy were not compatible, and whose technologies have been tied to human rights abuses and surveillance activities.”

The company provided “minimal details” about its work with Palantir, the statement says. Employees found out about the partnership in an email about 20 minutes before the earnings call, said Mike Davis, a reporter for the Asbury Park Press and a vice-chair of the NewsGuild of New York.

“Part of the problem is that we know very, very little about how Palantir’s software is going to be used,” Davis told me. “It’s pretty surprising that our company would unilaterally decide to bring in such a controversial partner like Palantir without providing a complete explanation to its employees, most of whom are naturally skeptical journalists.”

Reed told investors last week that USA Today Co. would retain ownership of its audience data, and Palantir’s AI software would help the company make the most of it. USA Today Co. owns more than 200 local newspapers in the United States, making it the largest newspaper chain in the country. It also owns Newsquest, one of the United Kingdom’s most prominent local news publishers with more than 150 daily and weekly newspapers.

“Our editorial decisions remain independent and are guided by our longstanding journalistic standards and ethics policies,” the company told Nieman Lab in a statement when the Palantir partnership was first announced.

“We respect that employees have a range of perspectives,” Amy Garrard, vice president of labor relations at USA Today Co., said in an email statement on Monday. “Our agreement with Palantir is a business decision focused on strengthening how we connect and better serve audiences, advertisers, and subscribers. Our data remains protected under our existing privacy commitments, contractual safeguards, and applicable laws. As always, our journalism remains fully independent.”

But the partnership still raises potential concerns about the safety of journalists’ data in the future, Davis said.

“An investigative reporter who keeps a ton of sensitive notes, source information, and confidential documents on their devices is going to be naturally concerned about protecting that information, and who could blame them?” Davis said. “It’s pretty hard to trust management’s intentions when the goalposts shift so often that there’s always a fear that what’s true today won’t be true in six months. We’ve all seen the years of disinvestment in local news, especially in the name of going all-in on AI.”

Read the full statement here.

This story has been updated to include comment from USA Today Co.

In its Q2 earnings call Thursday morning, USA Today Co. chairman and CEO Mike Reed told investors he expects the partnership to “strengthen how we collect, connect and activate audience data to drive more effective and faster monetization across our platform.”

“Every visit, every session, and every moment of attention creates a signal,” Reed said. “When we connect those signals, they become actionable intelligence that allows us to engage users more effectively and monetize those relationships faster and at much greater value. The work our team is doing with Palantir is a direct extension of this strategy. We are applying Palantir’s AI-powered platform to one of the largest opportunities in front of us, converting the sheer scale of our audience into known, orchestrated first-party relationships, because that is what turns our reach into sustainable, higher value revenue.”

USA Today Co. owns more than 200 local newspapers in the United States, along with its national daily USA Today. The partnership with Palantir comes as search traffic continues to fall — a challenge faced by publications around the world, and one The New York Times mentioned in its own earnings report Wednesday. USA Today Co. reported 158 million unique visitors in the second quarter of the year, down from 180 million in the first quarter.

This decline “does not reflect lower demand for the content,” said Kristin Roberts, president of USA Today Media. “What it reflects is lower referrals from traditional search because of those consumer discovery changes that we’re seeing and witnessing.”

Palantir has been the subject of controversy in recent months for providing its technology to Immigration and Customs Enforcement (ICE), the U.S. Department of Defense, and the Israeli military. The company’s co-founder and chairman is Peter Thiel, the billionaire entrepreneur who funded Hulk Hogan’s defamation lawsuit against Gawker, leading to the publication’s ultimate bankruptcy and closure.

USA Today Co. isn’t the only media company to enlist Palantir’s services. Axel Springer — which owns Business Insider, Politico, Bild, and The Telegraph — and Fox News have also partnered with Palantir for its data intelligence tools. Thomson Reuters has also been linked to Palantir due to ICE’s use of Palantir software and Thomson Reuters data, 404 Media reported in March.

“It’s important to note all of our data remains our data,” Reed told investors. “It’s really just leveraging this incredible AI and software that Palantir has to allow us to move so much faster and to be so much smarter with the data we have today. And…take a lot of the anonymous interactions we have today and turn those into known relationships…The more known relationships we have, the more data we create, the more signals we create, the more actionable intelligence we have.”

Roberts told USA Today Co. employees in an email Thursday that the Palantir partnership could result in “better recommendations and offers based on what people actually care about” and “smarter subscription and advertising experiences” for consumers and subscribers.

“USA TODAY Co. complies with applicable data protection and privacy laws and maintains strong standards for data security and governance,” the company said in a statement to Nieman Lab. “We require this of our vendors and partners as well, reflecting our belief that safeguarding protected personal information is essential to our business and the audiences we serve. Our editorial decisions remain independent and are guided by our longstanding journalistic standards and ethics policies.”

Roberts noted that World Cup coverage across USA Today’s network “generated 97 million pageviews, with search driving nearly 65% of that traffic. That reinforces an important point — when content meets a real and urgent need — search still delivers. But we are not building our future on search, even in our strongest categories. Great content still finds an audience, and our opportunity moving forward is to ensure that our distribution tactics keep pace with the way readers and viewers want to consume content in digital spaces.”

“I can see a day where we turn off scraping or making our content available for the [search engine links],” Reed said. “We’re actually more hopeful that we can be proactive with Google in negotiating a fair licensing deal. That would be our preferred path — to have our content appear both in traditional search as well as in AI summaries. But if we have to cut them off and block them in order to get to a deal, then we’ll do that for sure.”

Read the full earnings report here.

This story has been updated to clarify the relationship between Thomson Reuters and Palantir. 

Grokipedia allows humans to suggest edits, but Columbia’s Tow Center for Digital Journalism found in February that the tool was increasingly editing itself. “Grok-supplied edits spiked in December and have overtaken human submitters, making up more than three-quarters of the suggestions,” Tow Center fellow C.J. Robinson wrote.

Now, though, it appears that Grokipedia isn’t being edited at all. “The encyclopedia stopped editing itself and processing suggested edits from humans,” DiResta and the Stanford Cyber Policy Center Ronald Robertson wrote for Lawfare this week. “As far as we can tell, no entry has changed in more than three months.”

That matters, they say, because people actually use Grokipedia, and AI systems, especially ChatGPT, rely on its content:

The significance here extends beyond one internet encyclopedia. Grokipedia is not just a small xAI side project. The web traffic analytics firm Similarweb estimates that Grokipedia received 6.7 million visits in June 2026 and ranked 11,022nd among websites globally — substantial reach for a site launched less than a year earlier. Individual high-profile entries command significant readership: Grokipedia’s page-statistics API — which serves per-article view counts that are not displayed on the front-end website — records over 83 million views for President Obama and 16 million lifetime views for its Musk entry since October 2025.

Grokipedia’s contents also do not remain confined to Grokipedia. An Ahrefs analysis published in March 2026 found the site surfacing in roughly 356,000 citations across AI systems — most often in ChatGPT and Google’s AI Mode, and at lower volumes in Gemini, Copilot, and AI Overviews. That remains a small fraction of Wikipedia’s roughly 25 million such citations, but it is not trivial.

A reference work whose contents are silently frozen and whose editorial record is inaccurate could propagate incorrect or outdated information into answers that people encounter elsewhere. There are also potential reputational consequences for individuals with unreliable Grokipedia biographies who can no longer appeal to the AI for an edit…

As agentic AI systems begin to generate and maintain reference knowledge independently, the question is not only whether their output is accurate at a given moment. It’s whether their behavior over time is observable, and their record stable enough to be accountable for their contributions to the public information ecosystem. At the moment, Grokipedia is neither.

Boing Boing notes that Wikipedia is updated about a quarter of a million times per day.

Our book, published by Polity, argues it doesn’t have to be that way. AI can also offer journalism a rare chance to break that cycle and build something better. We draw on questions of creativity, democratic accountability, platform power, and human agency to offer readers a toolkit for reimagining journalism in the age of AI — not just as an industry, but as a profession dedicated to serving the public good.

The book engages with AI’s impacts on multiple levels, covering everything from the hands-on realities of AI-assisted news production and the shifting architecture of media power to AI’s implications for democratic life, journalism education, and the profession’s future. Below is a portion of Chapter 2, which examines how AI is already reshaping every stage of news production and asks whether these tools will reinvigorate journalism or merely accelerate its hamster wheel. This excerpt focuses on emerging applications of AI across five stages of the newsmaking process — story ideation, sourcing, verification, storytelling, and distribution — highlighting both their transformative potential and their limitations.

You can download a digital copy of Journalism in the Age of AI from our website for free starting August 6, and purchase paperback copies starting November 6 in the U.K. or January 19, 2027 in the U.S.

Smaller organizations have also stumbled. In November 2025, reporters at Suncoast Searchlight, a Florida-based nonprofit news organization, sent a letter to their board of directors after discovering their editor-in-chief’s undisclosed use of AI editing tools — which included instances where the AI inserted hallucinated quotes into reporters’ drafts. Some staff members were unaware that AI was being deployed on their work at all, raising important questions about transparency, editorial control, and ethical boundaries.

These uses of AI illuminate the emerging complications around authorship, accountability, and transparency — questions that grow more vexing as AI becomes so integrated into workflows that some people no longer consider it worth mentioning. While the need for disclosure is often underscored in debates about AI-generated news, the thresholds become murkier as AI permeates routine support activities and sometimes even directly modifies journalists’ work without their knowledge or consent. The line between tool and author therefore continues to blur, challenging foundational assumptions about who is responsible when storytelling goes wrong.

AI and distribution

Media distribution fundamentally shapes who can access journalism and under what conditions. For decades, news organizations had considerable direct control over their distribution pathways. However, the rise of digital intermediaries — search engines, social media platforms, news aggregators — has dramatically altered this landscape. Today, algorithmic systems increasingly determine which stories reach which audiences and provide the marketplace for monetizing them, forcing news organizations to adapt their editorial practices to accommodate the demands of algorithmic distribution. At the same time, AI empowers news organizations to rapidly remix their own content for multiple platforms and to expand accessibility.

Optimized distribution: Media publishers have traditionally owned large segments of their distribution channels. They often owned the printing presses and the delivery trucks. They owned the cameras and microphones to record the news, as well as the antennas to beam their stories across the country. With the advent of the internet, newsrooms also had significant control over their own websites. As social media platforms started to gain popularity in the 2000s, though, media publishers gradually became more dependent on platform distribution by tech companies like Google and Facebook, including their social media feeds, search engine results, and news aggregators.

Thus, news organizations have had to manage their own proprietary distribution channels while optimizing their content for algorithmic visibility. Some newsrooms even required reporters to build personal social media followings, effectively transforming individual journalists into distribution channels. Meanwhile, audiences increasingly encountered news through algorithmic feeds, where platform logic, not editorial judgment, dictated the final presentation.

Several news organizations have adapted by integrating AI into their distribution through two main strategies. Algorithmic optimization uses AI to decode platforms’ shifting preferences, advising journalists on ideal formats, optimal publishing times, and strategic keywords to maximize visibility and reach across platforms. Targeted delivery harnesses AI for precision audience segmentation, enabling organizations to serve personalized content to specific demographic or interest groups.

BBC News exemplifies this approach. In March 2025, the broadcaster announced a new department dedicated to AI-powered personalization. As then-CEO Deborah Turness wrote to staff: “We must become ruthlessly focused on understanding our audience needs, on delivering the kind of journalism and content they want, in the places they want it, designed and produced in the shape that they enjoy it.” The path forward, she argued, requires deploying “AI to support, enable and accelerate our innovation and growth.”

It isn’t just major organizations doing this, either. The Rural News Network, a collective of more than 500 local news nonprofits across the United States, created Text RURAL to solve a distribution challenge. With member organizations producing nearly 5,000 stories monthly, the AI-powered tool automatically selects, aggregates, and summarizes the most relevant content for individual users based on their location and interests. It then delivers personalized weekly news roundups via text message. It’s a practical solution that makes local journalism more accessible to rural communities underserved by traditional distribution channels. However, it is an automated solution that bears risk, as even major publishers have discovered in their experiments with AI summaries.

Automating content adaptation: News organizations are also using generative AI for automated adaptation, or the rapid generation of derivative products tailored to different platforms and accessibility needs. AI systems can automatically produce social posts, video summaries, audio versions, and translations from content originally created by human journalists. While these derivatives are typically reviewed before publication, they nevertheless reduce the time, cost, and labor required to meet audiences across multiple platforms while expanding accessibility.

Dow Jones Newswires launched a custom AI language service in 2024 targeting this exact opportunity. The system produces what it describes as “fluent” translations of hundreds of English-language stories daily into Japanese, Korean, and French. “It gives us an ability to very quickly, almost instantaneously, translate our very rich existing English-language content into a language that will attract other audiences,” observes Chip Cummins, The Wall Street Journal’s chief Newswires editor. The service reaches professionals “who prefer to read their business news in their native language or who are suddenly tapping into the US market as a place where they want to invest, or their clients are investing.”

Similarly, the U.S. television station WCPO 9 uses AI to automatically reformat content for different platforms. “AI helps us take horizontal video storytelling you might see on our broadcast or YouTube and convert it to vertical on Instagram or TikTok,” notes senior manager PJ O’Keefe, “so that, again, we can make as big of an impact as possible, on as many platforms as possible.”

Dow Jones’ approach includes transparency measures: each translated article carries a warning about AI translation and links to the original story. However, as WBEZ reporter Araceli Gómez-Aldana observes, current models still struggle with the “accuracy of the translation for news stories, cultural understanding, translating quotes for tone and understanding — not only a literal word-by-word translation — and adapting to various dialects and literacy levels.” In other words, there are still risks associated with this kind of functionality.

Despite these powerful capabilities, what AI cannot do is liberate news organizations from platform dependencies. No matter how sophisticated their AI tools become, newsrooms cannot alter the ethical, economic, or political decisions made by platform owners regarding content moderation, algorithmic demotion of news, or sudden rule changes. Platforms can unilaterally reshape their algorithms — as Meta has done repeatedly in prioritizing and then de-prioritizing news content on Facebook — leaving news organizations scrambling to adapt their carefully crafted AI-optimized strategies to new, often opaque, rules. This structural power imbalance constrains AI’s potential to transform news distribution in ways that truly benefit journalism.

To some people, I’m sure, the answer will seem obvious: Journalism’s just gotten worse. The news used to be produced by upstanding proud Americans who only cared about the truth, and now it’s the product of a cabal of latte-sipping leftist elites willing to lie about anything in order to promote their dark, anti-Christian agenda and destroy all that is good in the world. Let’s call this the Fox News hypothesis.

That argument, as common as it is, leaves something to be desired. People who lose trust in “the media” aren’t doing so solely in response to their own observations and experiences. They live in a social environment filled with pre-baked messaging on who’s got your back and who’s out to get you, and “the media” makes for a pretty convenient target. How much of the decline in media trust is actually about the media’s actions, versus the narratives that get spun about them?

That eternal question was one reason I was excited to come across a brand new dissertation written by a journalist-turned-academic named David Beavers. Beavers spent five years working as an editor and producer at Politico before moving to a PhD program at Harvard. Six years later, he’s a freshly minted scholar headed to Wake Forest, where he’s been hired as an assistant professor in the Department of Politics & International Affairs.

Beavers’ dissertation is titled “Undermining the Fourth Estate: Elite Criticism and Journalism’s Fragile Foundation,” and it aims to untangle the question of media distrust causality through a creative mix of journalism history, natural language processing, and field experiments. (It just won Harvard’s prestigious Charles Sumner Prize.) Beavers advances two major data-driven claims. First, the American system of journalism has historically relied more on rhetorical appeals than structural forces to proclaim and defend its legitimacy. The First Amendment lets anyone be a journalist, which is great — but its lack of professional gatekeeping has also left the field uniquely vulnerable to attack from those who seek to weaken it. And second, the decline in public trust in the media was both preceded by and driven by attacks on the press by political elites, especially at two key pivot points — white southern Democrats defending Jim Crow in the 1960s and 1970s and the Lee Atwater/Newt Gingrich generation of Republicans in the early 1990s. As a result, criticism of the press — which used to be relatively equal on both ends of the political spectrum — has become a decidedly conservative behavior.

Here’s the full abstract:

American democracy is predicated on having a public that is sufficiently informed to hold elected officials accountable and to make sensible decisions at the voting booth. But a decades-long decline in Americans’ trust in the press threatens to undermine the fundamental role the media plays in cultivating an informed populace. This dissertation asks: What explains Americans’ declining trust in the press? I address this question in three parts. In Part I, I examine the historical development of journalism as a profession to offer a theory for why American journalism’s institutional forms make it uniquely vulnerable to political efforts to undermine its legitimacy. In Part II, I quantify congressional rhetoric about the press between 1939 and 2023 and relate it to long-standing survey evidence to show that elite press criticism has driven down Americans’ trust in the media over the last half-century. Finally, in Part III, I turn from causes of Americans’ declining trust in the media to consequences and possible solutions. I examine the effects of media trust and elite press criticism on news consumption using web-browsing and newspaper circulation data, and I conduct a trio of original survey experiments to assess interventions aimed at restoring trust in reputable journalism. I conclude with reflections on the state of American media and American democracy — and how those two are fundamentally intertwined.

I gave Beavers a call to talk through his findings; here’s a lightly edited version of our conversation.

The Times added 280,000 digital-only subscribers between April and June, a drop from the 310,000 subscribers added in the first three months of the year. (The Times now has 13.3 million subscribers in total, “roughly on pace” to hit its goal of 15 million subscribers by the end of next year.) Total subscription revenue has increased to $538 million, with most of that coming from digital subscriptions. In an investor call, Times CEO Meredith Kopit Levien attributed the numbers, in part, to the decline of search traffic.

“We delivered our Q2 results against the backdrop of a rapidly changing information ecosystem shaped by a small number of big tech companies whose moves continue to result in less traffic to publishers,” Kopit Levien said. “The Times isn’t immune to that impact.”

Kopit Levien also emphasized that “long-term bets” on video are essential to the company’s continued success. Adjusted operating costs increased 10% year-over-year, in part due to “investments in our video journalism.” The Times hired eight video journalists in January and is currently hiring for 12 video-focused roles. In 2024, The Times experimented with pushing vertical video on its homepage and by 2025, it created a Watch tab in its main app.

“We’re now producing thousands of new videos each quarter to reach the enormous audience for video in all the places people watch, including our own destinations,” Kopit Levien said. “Just this week we launched a Shows tab in our flagship app, creating a new way to experience our long-form franchises in news, opinion, culture and lifestyle…This is all part of our strategy to engage the people we already have more, and engage more people. As we do that, we intend to make the Times as preferred a brand for watching the news as it is for reading and listening.”

Many of the original videos published this past quarter were of reporters explaining their reporting to the camera, which Kopit Levien said is a format that is “inherently humanizing and trust building.” She did not provide specific numbers on engagement, but when asked whether the heavy video investment has resulted in advertising revenue, Kopit Levien said it’s played a “minor role.” In the coming months, she said, the Times will “really focus on scaling production, scaling engagement, and then scaling monetization.”

The Times also said it spent $4.6 million on generative AI–related lawsuits in the second quarter of the year. In total, it’s spent $32.9 million on those lawsuits since it began breaking out their costs in earning reports in the first quarter of 2024.

Read Kopit Levien’s full comments from the earnings call here.

A lot’s happened to media in the ensuing decade — perhaps most notably, the pandemic-fueled rise in work-from-home arrangements that, again, could have reduced the number of journalists overpaying for a Brooklyn walkup. But Bloomberg’s Justin Fox has an interesting datapoint today that shows the concentration of media keeps on growing.

The nation’s media capital is in many ways far less central to the national conversation than it used to be, with TikTok, YouTube, Substack, artificial-intelligence chatbots and other new modes of communication sucking up so much of Americans’ time and attention. Yet in the production of certain kinds of media — journalism in particular — New York is more dominant than ever.

My measure of dominance is employment, which is incomplete and flawed but better than anything else available. Having looked into the centralization of media employment a decade ago, I was inspired to revisit the data by an online complaint (from Minneapolis) about “the New York-ification of Democratic politics.” What I found is that New York-ification of media has for the most part continued to grow even as media industries have continued to struggle.

Fox pulls labor data for a number of media professions, including magazines, book publishing, and broadcasting. But of most interest for our purposes is newspapers (a category which now also include digital-only publishers). In 2016, when I was writing about coastal concentration, 1 out of every 30 U.S. newspaper jobs (3.3%) were based in Manhattan. That was already up markedly from 1-in-48 jobs (2.1%) in 2000 — which was, roughly speaking, the final peak of the print newspaper business.

But that trend hasn’t retrenched in the years since — it’s accelerated. By 2025, 1 out of every 12 U.S. newspaper jobs (8.3%) was based in Manhattan. What was once one of America’s least centralized industries is increasingly squeezed into a few square miles of Midtown.

Check out the full post to see how the distribution of “news analysts, reporters and journalists” has become increasingly Acela-adjacent, along with other slices of contemporary media employment.

On May 4, the Pulitzer Prizes recognized these stories among the winners and finalists across 15 journalism categories. The reporters behind each of these stories also disclosed using AI to the judging committee. Ultimately, five award winners and three finalists this year disclosed AI adoption in their submissions — the most since the disclosure requirement was added in 2024.

For the past two years, I’ve spoken to Pulitzer-recognized reporters about how they used AI in their reporting. Both years, generative AI took a back seat to more conventional machine learning technologies, like using embedding models to produce complex data visualizations and pattern recognition models to analyze satellite imagery in conflict zones. This year, though, generative AI tools and commercial large language models (LLMs) were more commonly used, largely to speed up the process of combing through document dumps.

“To state the obvious, perhaps, AI is here to stay,” Marjorie Miller, the administrator of the Pulitzer Prizes, told me. “The industry [used to be] far more apprehensive about AI tools than it is today, with a clearer understanding now of what uses might be appropriate — data collection and analysis, for example — and when it might not, such as in writing and editing stories in any format that might be considered for a Pulitzer Prize.”

Miller cautioned that as AI evolves, reporters will need to “ensure and reassure” the Pulitzers that submissions are ultimately produced by human beings, even when AI is used as an assistive tool. Given recent controversies over AI-generated text allegedly appearing in prize-winning literary works, Miller also said next year the Pulitzers will include an AI disclosure question in their book entry forms.

Finding the needle in a stack of public records

In the days after deadly floods hit Kerr County, Texas in the summer of 2025, reporters at The Wall Street Journal had a clear reporting question: Had this area ever dealt with dangerous flooding before?

To find an answer, the reporters turned to public records. The team built a custom scraper that pulled every public meeting minute, agenda, and transcript from the Kerr County website.

Journalists are trained to find the needle in a haystack, but doing so on a breaking news timeline can be challenging. To speed up the document review, the reporters leaned on a pre-built internal tool called WSJPT (a play on ChatGPT). The tool standardizes basic LLM requests across reporting projects, including prompts for summarization, classification, and image description. In this case, the reporters used the tool to summarize every page of every document scraped from the county portal.

The team combed through these summaries using a combination of LLMs and more old-school natural language processing (NLP) techniques (e.g. stemming, lemmatization) to find sections that referenced past flooding events.

“We aren’t obviating the need for human investigation of a pile of documents — and I don’t think we would if we could,” said John West, a computational journalist at the Journal. “Instead, we’re trying to sort the pile so the most relevant stuff is right at the top.” West clarified that every section flagged as possibly relevant by these tools was read by a reporter, and then every document deemed relevant was read in full.

Based on this analysis, the Journal identified former Sheriff Rusty Hierholzer, who had pushed county commissioners to install a stronger flood-warning system a decade ago. In 2016, Hierholzer called for the installation of outdoor sirens, recounting an experience flying in helicopters and “pulling kids out of trees here (in) our summer camps” when floods hit nearby Kendall County in 1987, killing 10 campers. Hierholzer’s recommendations were not implemented at the time, the Journal found.

The findings were foundational to one news story and several follow-ups on the floods. The Pulitzers named the Journal’s overall coverage a finalist in the Breaking News category. West says this playbook — using a “mix of off-the-shelf and custom software” to summarize and parse documents — was also central to the Journal’s reporting on the Epstein Files this year. That coverage was a finalist in the Public Service category.

Like many of the investigations this year that disclosed AI adoption to the Pulitzer judges, no AI disclosure (such as a label, footnote, or accompanying methodology) appeared in the Journal’s own stories on the Central Texas floods.

“We did not disclose the use of AI. It functioned as a sophisticated way of searching through the documents, but we read the docs, and ran the findings down,” said West.

He contrasted that choice with a recent Journal investigation about toxic fume incidents on U.S. commercial aircraft. That story used LLMs to read more than one million FAA documents and to generate incident rates per airline and aircraft. For that story, which did disclose AI usage, West said he “got to write the longest methodology statement I’ve ever written.”

Translating on a breaking news deadline

On August 27, 2025, a 23-year-old woman killed two children and wounded 27 others during a mass shooting at the Annunciation Catholic Church in Minneapolis, Minnesota. The shooting rocked the local community, but in the hours that followed, there were few answers about the shooter’s motivations.

When news of the shooting first broke, reporters at The Minnesota Star Tribune gathered in a Slack channel to coordinate their coverage. They identified the shooter’s YouTube account and videos showing her turning the pages of a journal written in a language the team didn’t recognize.

Dana Chiueh, an engineer in the Star Tribune’s AI Lab (now a fellow at ProPublica), took screenshots of the videos and entered them into an enterprise ChatGPT account. The chatbot recognized the text as Faux Cyrillic, a variant of Russian typography that can be used to spell out English words.

“[Faux Cyrillic] is not a real language. It can be thought of more as a type of code that one might use if they were trying to conceal what they were writing,” said Chiueh, explaining that ChatGPT allowed them to quickly see if there was relevant background information buried in the code.

After hours of tedious manual screenshotting, Chiueh wrote a custom script to pull the screenshots from the YouTube videos automatically. Ultimately, ChatGPT was able to produce an initial translation pass on hundreds of journal pages, over 600,000 words.

“At first we were doing something really scrappy. That spirit of being able to quickly prototype and iterate is something that is really useful for a breaking news situation,” she said.

A team of journalists then put the translations into Google’s NotebookLM, a Gemini-powered notetaking tool they used to search for keywords and pull out themes. They found mentions of past jobs, relationships, and locations the shooter had visited, like pawn shops and shooting ranges — all information that informed the outlet’s shoe-leather reporting.

“We were very conscious that AI hallucinates, so we made sure that any quotes, context, and anecdotes were reviewed by a human translator,” said Tom Scheck, investigations editor at the Star Tribune.

Rather than sending the entire AI-generated translation to a professional, reporters flagged important passages for review by two Russian language academics at the nearby St. Olaf College. For the most part, the AI-generated translations were correct, but the academics found a few errors, including a passage that misrepresented the shooter’s potential motivation.

If it weren’t for the help of AI translation, Scheck says the Tribune would probably have hired a translator to go through the documents from the beginning, slowing down their turnaround time. Instead, the triaged translations informed an initial story on the manifesto the night of the attack and contributed significantly to a profile of the shooter published four days later. Both stories were a part of the coverage that won the Star Tribune a Pulitzer in the Breaking News category.

“AI allowed us to take a first pass on the content and then prioritize what we [might] use,” he said. “We know we have to run the marathon, but AI helped us start at mile marker five instead of at the traditional starting line.”

Making a trove of documents searchable

Over the past 25 years, the Chinese government has built up a sophisticated mass surveillance program. A series of investigations published by the Associated Press last year exposed just how many of the technologies fueling this surveillance apparatus were sold to China by American companies.

Reporters mapped the supply chains for state-of-the-art surveillance tools, tracking their development in Silicon Valley and deployment in China, implicating companies like Nvidia, Intel, IBM, Dell, HP, Cisco, Oracle and Microsoft. The reporting earned the AP a Pulitzer win in the International Reporting category.

Key to these stories were tens of thousands of leaked emails and databases from a Chinese surveillance company, as well as thousands of government records and procurement documents (like vendor bids, signed contracts, and invoices). AI was essential to sifting through these documents and making them searchable, according to Garance Burke, a global investigative journalist at the AP who worked on the project. The AP used LLMs to identify specific company contracts, summarize government records, flag specific people or technologies for further investigation, and organize all the information collected into more easily managed databases.

In other words, AI was an assistant in the early reporting and research stages of the investigation, helping to make sense of a massive pile of documents. The team unearthed evidence that IBM had worked with the Chinese defense contractor Huadi to design a national fingerprint database, and evidence that Intel and Nvidia helped enable AI capabilities on surveillance cameras used in Xinjiang and Tibet. The documents also showed that HP sold the Chinese police “digital fencing” products, which have been used to track when Uyghurs and other surveilled populations try to travel outside their home towns and provinces, among many other findings.

“The AI tools helped reporters search and review large volumes of public records more efficiently, but they did not replace the reporting or verification process,” said Burke. “Reporters manually reviewed documents surfaced through AI, independently assessed the accuracy of AI-generated summaries, and did not quote from those summaries.”

Using LLMs to double-check human work

Donald Trump is the self-declared “crypto president” — an industry booster who earned over $1.4 billion in personal crypto business dealings during his first year back in office. A team of reporters at The New York Times wondered if the administration’s pro-crypto stance had influenced the work of agencies that regulate the industry, namely the SEC.

To try to answer that question, the Times reviewed all of the SEC’s crypto-related enforcement actions dating back to 2017. The analysis surfaced a troubling trend. Since Trump took office again in 2025, the SEC had pulled back on more than 60% of its ongoing crypto cases, lessening penalties, freezing suits, and even dismissing cases entirely.

“It is unheard of for the agency to retreat from a swath of lawsuits against a single industry,” wrote the reporters in their investigation published last December. “Although the particulars of the crypto lawsuits differed, many of these firms had something in common: financial ties to Mr. Trump.”

The investigation is one of several stories that exposed Trump’s ongoing conflicts of interest with the crypto industry and earned the Times a Pulitzer win in the Investigative Reporting category.

Many of the Pulitzer awardees that used AI this year disclosed using LLMs to speed up and prioritize document review. The Times investigation stands apart. For their investigation, reporters downloaded more than 10,000 documents, including thousands of SEC news releases and over 700 federal court cases, according to Miller, the Pulitzers administrator. Reporters read and classified each of these documents manually over the course of several months. LLMs were only brought into the reporting process after this full human review was completed.

Reporters used OpenAI’s GPT-5 model to conduct a secondary review of the documents and check their work. They fed the model the documents from each of the lawsuits, as well as a detailed set of instructions on how to classify them. These classifications labeled “whether a case was crypto-related, whether it was inherited by the next administration and how liability was decided,” according to a methodology published by the reporting team.

The team compared the GPT-5 classifications with the ones assigned manually. When there were discrepancies, reporters went back and read the documents again to double-check their work. The Times declined to provide further details on this review process.

Hallucinations and other errors produced by LLMs are often cited as reasons not to use generative AI in investigative reporting. In this case, Times reporters turned the tables. They used LLMs as a tool to help keep human error in check.

In June, dozens of journalists, technologists, and academics gathered at NYU’s Arthur L. Carter Journalism Institute in Manhattan to discuss this growing threat to authentication and verification work. Hosted by Princeton’s Center for Information Technology Policy (CITP), the day-long workshop included expert panels, presentations, and breakout discussions. On Thursday, CITP published a report synthesizing the workshop’s findings and offering insight into the “fight for facts in the AI age.”

“Establishing one fact increasingly means establishing the facts beneath it,” write the report’s authors. “While AI tools can strengthen the verification process and introduce new opportunities to increase access to knowledge, there is an arms race with the volume of potentially fabricated facts.”

In response to the deluge of AI-generated media online, a host of verification initiatives, commercial products, and other tools have flooded the zone. The report suggests that many are not being developed with enough direct consultation from journalists.

One banner initiative in Silicon Valley to address authentication problems is C2PA, or the Coalition for Content Provenance and Authenticity. The initiative, launched in 2021 by Adobe, Microsoft, the BBC and others, embeds cryptographically signed “manifests” inside media files. It effectively creates a label the moment a photo is captured in a camera, recording its full history and origins as the image is edited, published, and circulated.

Despite gaining the support of major news organizations like the Associated Press and The New York Times, C2PA has hit major snags when it comes to newsroom implementation. One anonymous newsroom mentioned in the report said the software it used to ingest and edit photos stripped the manifest; it had to pressure the developer to patch the issue. Since the publisher’s CMS didn’t support C2PA, the manifest was viewable on images internally but deleted when published for readers. Most CMS vendors currently do not support C2PA.

“Will audiences believe the news organization has a chain of custody for its files if it cannot offer ‘full transparency’ of the manifest?” the report’s authors ask.

The report also notes that many social media companies strip C2PA and other metadata from images circulated on their platform, further eroding the initiative’s effectiveness.

There are also safety concerns. Cryptographic signatures could endanger journalists and sources by attaching identifying information to photographs or videos they take. This presents an immediate threat to those working in authoritarian environments, according to a recent study by the human rights organization Witness.

Similar challenges arise with other authentication tools, including AI image and video detectors. Many of these tools operate with confidence ratings — they might spit out a 80% likelihood that an image is AI-generated. The report explains that these types of ratings aren’t useful to most reporters, given that readers usually think about authentication as a binary: is it real, or is it fake?

“These confidence ratings are not (easily) translatable to audiences, since they want to know if an image is genuine or not and not a confidence rating that reporters often cannot even explain since the authentication tools are often ‘blackboxes,'” write the authors.

Given these challenges with bringing authentication tools into newsrooms, the report suggests that there needs to be more collaboration between technologists and the news industry to better meet the need for authentication tools.

“We urge technologists to work with reporters to develop the tools together and then let reporters evaluate their usefulness,” write the authors.

You can read the full report here, including more recommendations and an overview of how AI is shaping verification work at large.

Pageviews increased across features, sports, business, and news content — “exactly the behavior we had hoped to drive,” Hughes said. The Inquirer is a for-profit public benefit corporation owned by the nonprofit Lenfest Institute; Hughes said the initiative was supported and accelerated by philanthropic underwriting.

Other local newspapers saw traffic gains driven by round-the-clock coverage of major stories. In California, The Orange County Register got 1.3 million more web visits in May than in April, a 54% jump. (The Register is part of Southern California News Group, which is owned by Alden Global Capital.) Toni Sciacqua, Southern California News Group’s managing editor for digital, attributed the spike to a big story right before Memorial Day about a chemical tank at threat of explosion in Garden Grove, prompting an unprecedented evacuation. The team drew from its breaking news playbook for wildfires, pushing out constant, timestamped updates that included language like “as of 2 p.m.” so people would know what was new since the last time they read. The Register sent newsletters throughout the day, updated social media posts, and scoured social media and Google search trends for questions the newspaper could answer. “Lots of people were looking for whether this would impact Disneyland,” Sciacqua said. “The answer was no, but we did stories specifically answering that question and saw a lot of interest in that.”

The team published evacuation zone maps that included local landmarks, shelters, and freeways. Simple explainers defining technical terms used by officials became some of the Register’s most-read stories. Once the team had a handle on the breaking news, investigative reporters did a deep dive into the company that owned the tank and looked for similar facilities in residential areas throughout California “to answer more questions readers outside the impacted area might have had, too,” Sciacqua said.

In Augusta, Georgia, both The Augusta Chronicle and The Augusta Press saw traffic spike by more than 50% in April. The Chronicle is a Gannett publication, and it’s one of a few around the country to see traffic upticks of more than 50% in the last few months, along with Florida Today and The Courier-Journal.

“I’m quite happy to say that our April traffic bump was largely a response to two pieces of longer-form, accountability journalism — totally unrelated, with totally separate collaborations, that just happened to drop close to each other,” Gabe Rosenberg, the audience editor at KCUR, told me in an email. “If I have any takeaway, it’s really just the importance of headlines that answer the question, ‘Why does this matter? Why should I care?’ Audiences found and shared both of these stories because they could emotionally connect with them.”

The first piece that contributed to the bump was a story about the Trump administration shutting down the country’s premier bee lab. It came from Harvest Public Media, a multi-state collaboration of public media stations that covers agriculture, the environment and food systems and is based out of KCUR. The story quickly blew up on Google Search and Discover, performing “astonishingly well” according to Rosenberg.

“[Harvest Public Media’s] coverage of the Trump administration’s actions on government agencies and their trickle-down effects have been particularly popular, as I’m sure they have been for many stations,” Rosenberg wrote. “I think in this particular case, the combination of the emotionality of bees as a poster image for pollinators and hurting the environment, the timing of these honeybee deaths that were news in and of themselves to most readers, and the narrative around the Trump administration led to a real home-run. Many of Harvest’s stories are national-focused, even while they are grounded in local reporting, so KCUR benefits a lot from their wide reach and appeal.”

KCUR’s traffic also got a boost through a story from another collaborative, the Midwest Newsroom. That story, about a police sniper in Missouri becoming a Missouri state trooper even though he had killed a two-year-old girl in a standoff, reached readers beyond the Kansas City metro area. The two stories together helped KCUR see nearly 170,000 more visits in April than in March.

KQED, based in San Francisco, also saw traffic bumps in April and May, thanks in large part to a tumultuous gubernatorial primary that saw Democratic frontrunner Eric Swalwell drop out of the race after being accused of sexual assault. The station’s primary elections coverage, along with a story about $700-a-month sleeping pods in San Francisco, led to major traffic bumps; KQED saw about 885,000 more visits in April than in March, and an additional 445,000 or so page visits from April to May.

“Underneath both content factors is a deliberate shift we’ve been making toward publishing the same reporting across formats, so a text story can find a second life as a vertical video months later,” wrote Peter Cavagnaro, director of communications and external affairs at KQED. “The sleeping pods piece is the clearest example: it ran in April and the vertical video version did about 900,000 views on Instagram Reels this month.”

LAist, which is home to 89.3 FM, also saw major bumps in May thanks to the California elections. According to Jon Cohn, LAist’s vice president for audience and community engagement, its Voter Game Plan and voter guides ahead of the primary led to a bump of more than 520,000 visitors in May, putting it at the top of our charts for biggest gainers by raw visits in May. (Nonprofit news publishers, too, saw traffic bumps in the quarter thanks to their coverage of primary elections.)

Here are the individual rankings for Q2 2026, broken up by month.

With the newsroom hobbled, the British publication’s management allegedly deployed generative AI tools such as ChatGPT and Claude to suggest headlines and write descriptions of photos for people using screen readers. The National Union of Journalists cried foul, calling the move a clear attempt to undermine collective bargaining.

As media and communication scholars, we saw the outrage over management’s alleged use of AI — the details of which the outlet’s management has disputed — as more than routine wrangling over adherence to labor law. The incident showed how AI is becoming integrated into daily news work through translation, interview transcription, research, writing, and audience analytics. At the same time, the incorporation of AI in newsrooms has raised new worries about accuracy and bias.

In our recent research, we examined how three different media outlets — Reuters, the BBC, and The Guardian — were deploying AI in their workflows.

We conducted a comparative case study, analyzing public documents published through 2025, including official AI policy statements, editorial guidelines, trial reports and industry presentations.

We found that there is no single story about AI’s integration into newsrooms. Instead, the way the technology is used appears to reflect the funding, values, and mission of each outlet.

Speed as market strategy

Reuters is a global wire service that syndicates news to other outlets.

Its coverage of financial markets and other fast-moving sectors, such as foreign exchange trading and corporate earnings, means that even a 30-second delay in reporting a central bank decision or a market-moving news event could cost traders significant sums of money.

For this reason, the company has heavily incorporated AI to speed up its workflow.

For example, the newsroom uses an AI tool the company developed called Lynx Insight to analyze financial data in real time, and another internal tool, Fact Genie, to process corporate press releases.

Thanks to these tools, Reuters has expanded its earnings coverage and can publish some breaking financial alerts in just six to eight seconds, turning what was once a race measured in minutes into one measured in seconds.

Reuters sees automation as a way to produce more content for different markets. Operating at speed, however, makes accuracy all the more challenging.

Reuters acknowledges that AI tools can introduce factual errors and requires its journalists to verify any AI-generated claims before publication. In other words, the high volume of AI-generated news alerts does not preclude editors from checking them for accuracy before they go live.

While speed always comes with risks, Reuters may also be well positioned to manage them because financial markets provide an immediate feedback loop, with traders quickly identifying and reacting to incorrect information. The same approach, however, could create greater challenges in political or investigative journalism, where errors may not become apparent for days, months or even years.

AI with guardrails

The BBC operates under a clear system of accountability. The publicly funded U.K. broadcaster depends on editorial impartiality for its credibility. Unchecked automation could put this core value at risk.

As a result, the BBC has taken a cautious approach to AI integration. Its main AI tools — At a Glance and BBC Style Assist — are used for tasks such as summarizing long articles and reformatting local news stories to match the BBC’s house style.

At a Glance creates bullet-point summaries of existing articles, while BBC Style Assist helps edit regional stories in the broadcaster’s signature style and voice.

Rewriting a story to match BBC house style can take about 30 minutes, and BBC Style Assist is designed to cut that time down substantially. The outlet also uses AI translation tools to reach audiences in Poland, Hungary, and Romania.

No AI-assisted story is published until a journalist reviews and approves it, a step that reflects the BBC’s emphasis on accountability. The BBC has acknowledged the limits of AI in its current form, publishing a study with the European Broadcasting Union in October 2025 showing that AI assistants misrepresented news content — through faulty sourcing, fabricated details, or outdated information — 45% of the time.

Rather than downplaying such findings, the BBC published them in an article on its media center and went on to create AI labels — a hexagon icon paired with the phrase “How we used AI” — that now appear at the top of content.

Principles before profit

When we conducted our research, The Guardian had an AI policy that barred the use of generative AI tools to create text or images intended for direct publication outside of exceptional and specific circumstances.

But in March 2026, The Guardian updated its guidance. The revised policy now permits limited use of generative AI for tasks such as suggesting alt text for images, analyzing parliamentary documents, or transcribing audio into text, provided the tool aligns with The Guardian’s editorial standards, is subject to human oversight, and receives explicit permission from a senior editor.

“Our approach to generative AI framework is designed to support our journalists’ expertise, never replace it,” the outlet told us. “Our journalists are always accountable for the journalism they create, and use of these tools requires absolute rigor and responsibility.”

In our view, The Guardian’s ownership model helps support a more cautious approach to AI. The outlet has been structured as a trust since 1936 to safeguard the paper’s editorial independence. There are no shareholders or sole owners. The outlet relies heavily on voluntary reader contributions in order to remain free to read.

This reader-funded model helps insulate The Guardian from commercial and political influence, allowing it to prioritize long-term editorial independence over short-term gains. Its principles on AI are clear: Any use of AI must help readers and respect both the public interest and creators’ rights.

When the paper’s management locked horns with workers during the 2024 strike, many staffers saw the use of AI to cut costs or counter labor actions as a betrayal of these principles. The incident highlighted the conflict that can occur when an organization’s values clash with pressures to keep things running.

Why this matters beyond the newsroom

Reuters, the BBC, and The Guardian each take a different approach to journalism, revealing unique ideas about the role of news in a democracy.

In our view, Reuters is more prone to view journalism as a business. The BBC is more likely to treat it as a public good. The Guardian’s approach to journalism is intertwined with its long-term values.

For these reasons, we argue that AI won’t uniformly change the news industry. No matter where it’s deployed, the time reporters and editors spend verifying AI’s accuracy can create what we call a “verification tax,” which can lessen the technology’s value. If newsroom staffers spend more time correcting chatbot errors than writing their own stories, the technology ends up being more trouble than it’s worth.

That said, the technology will likely accelerate differences already underway. Commercial news organizations will use more automation. Public broadcasters will include more human oversight. And independent outlets will focus more on stories written and reported wholly by people.

In the future, we think the most influential and successful newsrooms will not necessarily be the fastest or have the best AI tools. Instead, they will be those that use the technology wisely, balancing speed with accuracy, treating workers fairly, and protecting the trust and integrity that make journalism a valuable resource.

This article has been updated to include The Guardian’s dispute of the claims surrounding AI use during the 2024 strike.

In Nieman Lab’s quarterly ranking of nonprofit news outlets, four news outlets saw a bump in traffic thanks in part to their coverage of primary elections in May and June. We bring you four of these each quarter, looking at four different market segments: for-profit local news sites, local newspapers, nonprofit news outlets, and public media stations.

The 51st

Worker-led news cooperative The 51st attributed its uptick in site visits (from 120,000 in May to 179,000 in June) to its coverage of the Washington, D.C. primary election. The primary included high-profile races and the launch of ranked-choice voting, which The 51st cofounder Maddie Poore speculates led to increased readership of its voter guide, candidate profiles, and other stories.

Poore said the voter guide got 50,000 pageviews, and overall election coverage helped pull in 6,000 new newsletter subscribers. The 51st doesn’t have a paywall, but asks readers to register with their email addresses. Most of the traffic came from organic search.

“We found this heartening, in a time when many experts are writing eulogies for clickthrough from search engine results pages,” Poore said. “With our voter guides, profiles of candidates, and nuanced fact-checks, we’re providing information here that can’t be broken down at Google’s summary length. People still have to click through! Doing this successfully requires being super clear on your community’s information needs.”

CalMatters

In May, CalMatters topped our growth chart, receiving an estimated 3.8 million site visits. That’s up from 1.9 million in April. Editor-in-chief Kristen Go said the California gubernatorial primary drove readers to CalMatters’ voter guide, which often has record-breaking audience reach. But the impact of the voter guide extended beyond pageviews.

“In May we saw a surge in new members [and] a record-low monthly churn, and the coverage drove donations,” Go said.

CalMatters also created video guides for the primary. Journalists interviewed gubernatorial candidates for vertical videos, which led to organic shares on TikTok and Instagram.

The Colorado Sun

Coverage of primary elections in Colorado increased The Colorado Sun’s site visits to an estimated 1.9 million in June, up from 1.2 million in May.

Stephanie Rivera, the Sun’s director of audience and engagement, said the biggest traffic drivers were a voter guide published in early June, and election night coverage on June 30. The Sun published more than 30 election stories that got a combined 800,000 pageviews. Much of the traffic came from Google Search. For example, The Sun’s guide to secretary of state candidates got 51,000 pageviews.

“While this wasn’t considered a competitive race, nor did it have controversy surrounding it, I think it reminds us that voters want to make informed decisions beyond key races and will seek out balanced reporting to help them cast their ballots,” Rivera said.

NonDoc Media

NonDoc Media, a nonprofit news site covering Oklahoma, published 70 stories in the lead-up to the state’s primary elections in June. It got an estimated 81,000 site visits in May and 133,000 in June.

Editor-in-chief Tres Savage said NonDoc, an outlet with 10 full-time employees, increased its publishing pace leading up to the election to “ensure readers had timely, comprehensive information about candidates, races and ballot issues.” NonDoc also hosted five public political debates, “which contributed to generating audience interest and expanded our election coverage.”

Other gains

Mississippi Free Press

Mississippi Free Press saw its site visits increase by 302% between May and June — from 87,000 site visits in May to 353,000 in June.

June’s most-read story was about a police officer killing one-year-old Kohen Wiley on June 16. Mississippi Free Press (MFP) has since published 12 stories about Wiley’s death and compiled them on a separate landing page.

About 46% of traffic came from social, and much of that was from Reddit. Since then, MFP has posted its stories to Reddit more frequently. Audience manager Chris Ellis also attributed the growth and attention to MFP’s journalistic rigor.

“We did not hide the officer’s action behind passive or indirect language, nor did we make the vehicle the central focus of the story about a child’s death,” Ellis said. “We continued following the story, distinguished verified facts from official claims and gave the family the space to speak for themselves. Readers responded to that combination of clarity and humanity, sharing the coverage far beyond Mississippi.”

VTDigger

VTDigger, an outlet covering the state of Vermont, received 1.1 million visits in April, up from 863,000 in March. Audience and product director Taylor Haynes attributed the growth to a surge of traffic from Google Discover.

“Several of our top-performing stories received the majority of their traffic from Discover, which was a noticeable shift from what we had seen earlier in the year,” Haynes said.

VTDigger’s top story in April was a profile about a Vermont teenager who became the youngest runner from the state to compete in the Boston Marathon. About 87% of the story’s total traffic (380,000 pageviews) came from Google Discover, Haynes said. Three other major stories — about Medicaid, housing, and immigration — received 65% or more of their total traffic from Google Discover.

“A common thread was that they had relevance beyond Vermont and featured strong human-interest angles,” Haynes said. “While VTDigger is a state news organization, these stories appeared to resonate with broader audiences and aligned with the types of timely, widely appealing content that often performs well in Discover.”

Haynes said VTDigger started adding meta descriptions to its stories in April, and while it’s too soon to tell if that caused the Google Discover traffic bump, it’s something the team is monitoring.

Block Club Chicago

Much of what drove readers and traffic to Block Club Chicago in April was the outlet’s nuanced coverage of crime stories.

One story covered the release of a video related to the death of a Chicago police officer. Another reported on the shooting of two police officers in a neighborhood hospital.

“I would say this was a more crime-heavy month than we’ve seen in awhile, in terms of the types of stories that performed well,” cofounder and executive editor Jen Sabella said. “Block Club works to cover crime when we’re able to include context beyond a police narrative. All of the stories that performed well here fit that bill, which I think shows that people want news on crime impacting their communities, but with context and community voices included.”

As a result, Block Club Chicago saw those stories boom on Google Discover and enjoyed more engagement on Instagram and more podcast listeners.

The Intercept

The Intercept saw a 30% increase in traffic from March to April, CEO Annie Chabel said, totaling 2.1 million visits. The month’s top three stories were all scoops: FBI director Kash Patel’s previously undisclosed alcohol-related arrests, the Pentagon’s undercounting of the number of U.S. troops killed or wounded in Iran, and the story of an Alabama grandmother who was arrested for wearing a penis costume to a No Kings protest. (After The Intercept published the story about her case, she was acquitted.)

Chabel said The Intercept’s increase in visits came from a variety of sources, and “no single platform accounted for more than 9% of pageviews.”

“As Big Tech platforms increasingly serve users with AI slop instead of real news, readers are flocking to The Intercept’s independent investigative journalism by word of mouth,” Chabel said.

The Salt Lake Tribune

The Salt Lake Tribune attracted 3.6 million visits in April, up from 3.2 million in March. It saw another bump in May, to 4.3 million visits.

In May, the Tribune published 50 stories about Canadian businessman Kevin O’Leary’s proposal to build data centers in Box Elder County, Utah. Those stories got 400,000 pageviews combined, according to chief of content Danyelle White. Google Search and Discover accounted for more than half of referral traffic to those stories.

Some of that can be attributed to Google’s algorithm changes, White said, along with new editorial strategies.

“One of our tactics this year was to reduce the number of stories that failed to reach 1,000 readers,” White said. “So far, we’ve reduced those stories by 10%, freeing up more staff time to focus on stories that resonate with audiences.”

The depressing one-year anniversary is being marked by a slew of articles and updates on how public media is faring. The good news? The situation is not as bad as feared, thanks in part to a surge of public support — “white-hot rage-giving,” as a South Dakota Public Broadcasting executive put it to NPR’s David Folkenflik. But it’s unclear whether that level of giving is sustainable. “A boom in fundraising after a crisis often precedes a big decline when donors move on to the next emergency,” Current notes.

Here are some interesting bits from reporting on the anniversary.

An unprecedented level of donations gives stations “breathing room.” Tim Black, the CEO of WNIN in Evansville, Indiana, told NPR he’d “expected to pull about $770,000 from a reserve fund. But he hasn’t had to touch a penny of it. He says the station’s members have contributed a bit more than $1 million — about 40% more than usual.”

A recent Contributor Development Partnership report found that “all types of public media stations — radio, TV and joint licensees — saw double-digit percentage increases in total donors from 2024 to 2025,” Current notes. “For radio stations, the surge ended four years of year-over-year declines.”

“The number one thought on my mind in the last year has been, how do I retain these donors?” Rob Risko, VP of development and membership at LAist, told Current. “How do I keep this momentum going? How do I turn this movement into longevity?”

(It’s worth noting that the increase in donations and gifts has not prevented layoffs at stations large and small. Semipublic’s Public Media Layoffs Tracker counts 550 public media jobs lost between July 18, 2025 and May 27, 2026.)

Some states come to the rescue; others add their own cuts. Rural and tribal radio stations, which received a significant portion of their overall funding from the federal government, have been particularly hard-hit by the funding cuts. In response, Native Public Media (NPM) has established two new funds to sustain tribal broadcasting, the Alliance of Rural Public Media reported in April. And “New Mexico’s state legislature has included language in its FY27 budget to continue funding public media stations within the state, including four tribal stations in NPM’s network.”

Other states, though, have added their own funding cuts on top of federal ones. “In Alaska and Alabama, the next governor will influence whether state money for public broadcasting comes back (Alaska) or goes away (Alabama),” NPR public editor McBride notes. “The Wyoming legislature recently came close to cutting the $800,000 it contributes every year to public media, but restored the money to its budget after a public outcry.”

Some universities are walking away. More than half of NPR member stations are affiliated with a public university or college. Now, some of those institutions are reconsidering whether those relationships are worth the trouble, reports McBride:

Penn State University and the University of Missouri are both divesting their public media licenses.

After negotiations to transfer its license initially broke down, Penn State’s board of trustees voted last September to shut down WPSU as a way of saving $3.4 million a year in subsidies the university was supplying to the station. But in response to public pressure, they agreed to transfer the licenses to WHYY in Philadelphia. WHYY will take on the additional fundraising to keep WPSU, which serves central Pennsylvania, running with local programming. That deal is awaiting FCC approval.

The University of Missouri announced in 2025 a three-year window to convert the public media stations in St. Louis and Kansas City to community-run boards. The NPR and PBS stations in both communities are currently working to manage the transition. Mizzou is keeping KBIA in Columbia, which is run by the university’s legendary journalism school.

Other colleges that hold public media licenses are cutting the level of support they provide. WOUB in Athens, Ohio lost $1.2 million in federal funding, which resulted in staff cuts earlier this year. The station then learned that Ohio University was also trimming the station’s annual allocation, prompting the acting general manager to warn that more cuts may be coming this year, including possible program cuts.

Stronger together? WHYY and Penn State’s WPSU aren’t the only stations merging. The Public Media Bridge Fund, a philanthropic effort to sustain public media through the funding crisis, told McBride it is “currently reviewing around 20 applications for collaborations, mergers, and other cooperative agreements.” Boston’s GBH and New England Public Media merged in May. Rocky Mountain Public Radio announced earlier this month that it will acquire Community Radio for Northern Colorado. And this week, WAMC Northeast Public Radio announced its own collaboration with newsrooms around the region, including the Times Union, The Berkshire Eagle, and the Adirondack Explorer.

“The initiative reflects something we’ve long believed at WAMC,” the station said in its announcement. “Public media’s job isn’t just to produce good journalism, but to make sure good journalism finds its audience.”

Today’s edition looks at local for-profit news sites. (Eligible sites for this ranking are for-profit members of the trade group LION Publishers — LION standing for Local, Independent, Online News — so these are mostly digital-born newsrooms, not newspapers or TV stations.) The rankings for Q2 feature many of the same sites we told you about last time, like TAPinto, Oil City News, Muddy River News, and Lost Coast Outpost. This time, we’ve got details from local news sites in California, Illinois, Wyoming, and British Columbia. Below that are the Top 25 rankings for April, May, and June.

L.A. Taco

At L.A. Taco — which promises “street-level news stories that often cover L.A.’s most marginalized communities” — boomed 51% in May, and part of the explanation is right in the site’s name. May was the peak of the site’s annual Taco Madness event, which offers both an online 64-taco tournament and an in-person event that draws more than 2,000 people. “We love leaning into our roots and promoting small businesses that make delicious food with passion,” said publisher Alex Bloomingdale. (The winners: Los Dorados LA and Tacos Los Cholos.)

But the two biggest traffic drivers of the month were more electoral broccoli than birria de chivo. Two voter guides for L.A.’s June 2 election — one a detailed overview of the candidates’ positions, the other an interactive guide to all the endorsements made by 14 progressive organizations, noting which candidates were unanimous or consensus picks and where different groups were divided. It’s a nice reflection of the editorial approach of a site comfortable covering carne asada and ICE raids, lucha libre and the LAPD.

Energeticcity.ca

Energeticcity.ca has covered Fort St. John (you’ll never guess the city motto!) and the surrounding area in northeastern British Columbia since 2008. Its traffic in June was up 52% over May. Adam Reaburn, the site’s owner and general manager, cited a few high-performing stories and said the site’s been working on improving its SEO.

One twist: Because it’s a Canadian site, Energeticcity.ca can’t post its work on Facebook as a normal user. Three years ago, Meta banned Canadian news stories from its platforms in response to the Online News Act. So the only way for the site to put its stories in front of Canadians is buying a Facebook ad.

Facebook was often responsible for close to 40% of site traffic before the ban. Efforts to push readers to its newsletters and mobile apps have helped the site recover the lost visits. But now instead of posting to Facebook, Reaburn buys ads to reach its readers still on Facebook, “mainly to reach people who otherwise aren’t seeing our content anymore.”

“We’re honestly conflicted about it, because every ad dollar ultimately supports Meta, the same company whose ban cut us off in the first place,” he told me. “But we’ve decided it’s worth it if those ads convert Facebook users into newsletter subscribers, app users, or paying supporters, audiences we actually own rather than rent through a platform.”

The TRiiBE

In June, traffic to The TRiiBE, which covers the Black community in Chicago, spiked 65%. Tiffany Walden, the site’s cofounder and editor-in-chief, gave some of the credit to a profile of actor Cortez Smith, a West Side native who plays Nuck on “The Chi.” The site emphasizes cultural coverage in the summer months and this profile generated more than 20,000 pageviews.

“The TRiiBE strikes a balance between news and culture,” Walden said, “and we do that intentionally, as a way to showcase the multifaceted essence of Black Chicago.”

Buckrail

Buckrail, a site in Jackson Hole, Wyoming, saw its traffic jump 96% in April, and editor Marianne Zumberge credited an unusual source: the site’s tradition of running satirical April Fool’s Day stories. And these are very local April Fool’s Day stories, to the degree that some are almost incomprehensible to outsiders — like this year’s “Elk shed hunt canceled after researchers find mineral cure for CWB.” Or 2023’s “Bridger-Teton to utilize JHMR tram as migratory tool for bighorn sheep.” (Others are less opaque to us city dwellers, like “Flock of owls training to pick up litter in Yellowstone.”)

“They’re typically shared widely, which is fun,” Zumberge told me. “Apparently [this year’s elk-disease satire] was so convincing that people in the community still believe it to be factual. It got a lot of play.”

The Hendersonville Lightning

The Hendersonville Lightning covers a North Carolina mountain town of about 15,000 people, just south of Asheville. A typical story gets around 500 to 700 pageviews, according to editor and publisher Bill Moss. But its May traffic zoomed up 88% over April. The month’s top-performing stories were a pretty standard mix of local news (courts, zoning, crime), with No. 1 being about a record $101 million civil judgment in a suit over a retaining wall that collapsed onto construction workers in 2021.

Moss said he’s been surprised this year by unexpected traffic levels to seemingly garden-variety stories — like a Senate candidate’s campaign stop scoring 9,110 pageviews. The site’s daily newsletter is strong and growing, he said, and “we’re working on a new strategy to convert newsletter subscribers to paid consumers of the print and digital product.”

Now, it’s seeking to expand.

Following a feasibility study, the nonprofit LOOKOUT — whose tagline is “Fierce. Independent. Queer” — is working to move into Colorado and throughout the region. The outlet takes inspiration from ’90s street zines and the digital vanguard like Vice.

Beyond the Southwest, the Chicago-based Joyce Foundation has also supported a landscape study to gauge the need for such a publication in the Great Lakes region.

Currently based in Phoenix, the newsroom prints a quarterly magazine, produces about three separate newsletters each week to more than 14,000 subscribers, and publishes news, investigations, and accountability journalism on its website each week. The style marries the voice of an alt-weekly — “Join Our Drunken Live Blog of the Republican Governor’s Primary Debate” — with sharp and rigorous reporting.

The newsroom counts a full-time staff of four and a stable of freelancers, interns, and fellows. At this month’s Association of Alternative Newsmedia convention in Palm Springs, it picked up four awards, including one for excellence in journalism. The site doesn’t have a paywall. The organization hosts more than three dozen in-person community events across Arizona each year.

Recent stories include how the state liquor board debated queer culture for eight hours, how an heir to the U-Haul business tried to stop a gay leather bar from opening, and how Arizona’s leader of the Stonewall Democrats has a “history of incendiary conduct.”

“I’ve always been a bit of a rabble-rouser,” said LOOKOUT’s founding editor, Joseph Darius Jaafari, a tattooed 38-year-old Iranian journalist who has written for Rolling Stone, The Marshall Project, and others. “I like to poke, I like to prod, I like to be a bit of a jackass to The Man. That’s kind of, I think, why I went into this career in the first place.”

A grant from the American Journalism Project seeded the publication and the MacArthur Foundation supported its recent feasibility studies. But the founders worry that anti-LGBTQ+ rhetoric from the White House has made it difficult to attract future or sustained funding.

“I do believe this is a pivotal moment, and philanthropy has the ability to step up and support organizations serving communities that are being targeted,” said Jake Hylton, LOOKOUT’s executive director.

The success and ambition of the local news operation is an example of a small, bootstrapped nonprofit newsroom building a community and relying on that community to help fuel its sustainability and reach new places.

“We cannot wait for philanthropy to fund us to do this,” Jaafari said about the site’s move into Colorado. “We’re going in there bare fucking bones.”

“People like us asking this everywhere”

The 2024 presidential election of Donald Trump galvanized Jaafari, a veteran journalist, and his partner, Hylton, who worked in real estate and housing advocacy. They thought about leaving Arizona, which they didn’t feel was a particularly safe place for queer and brown people.

“Very quickly, the conversation molded itself into ‘Oh, shit, there are people like us asking this everywhere. What can we do about that?'” Hylton said. “There are queer people everywhere, and the type of work we do doesn’t really exist this side of Ohio.”

Colorado is the first step as they seek to expand throughout the Southwest to include Utah, New Mexico, and Nevada. It’s also part of their own survival strategy as a news organization.

Colorado has a robust nonprofit news scene and a strong philanthropic journalism support network. The extent to which a new statewide outlet can break through and compete for funding will be a test for future growth elsewhere.

Last month, LOOKOUT’s founders hosted an invite-only event at a Denver bar for partners, funders, and local publishers. The gathering received support from the Knight Foundation and the digital marketing firm BlueLena.

Danika Worthington attended the party. As the membership manager for the nonprofit statewide Colorado Sun digital outlet, she has had an inside look at the state’s media scene while helping grow a startup through most of its eight years in business.

There is plenty to cover in Colorado’s queer space, both good and bad, she said, but she often feels that queer news is relegated to arts and culture. The Denver journalist, who has donated to LOOKOUT, said she’s excited for an outlet to take a serious approach to hard news in her state.

“I think it might even lead other publications to kind of step up their game,” she said, “or at least to partner with them.”

“Stories that wouldn’t have been told otherwise”

Jaafari had been working for the Gannett-owned Arizona Republic newspaper as an investigative reporter for about a year in 2022 when he said he approached his editors about covering the LGBTQ+ community more robustly.

After feeling rebuffed more than once, he started a monthly Substack newsletter.

“It was a protest, essentially,” he said. In more colorful terms, he described it as a “fuck you.”

The project grew but created friction at his day job. A $400,000 startup grant from the American Journalism Project allowed him to quit the Republic and work on his startup full time. He and Hylton built a website, produced consequential reporting, and launched a series of in-person events.

“I’m stealing everything I learned from my time at Vice and from little raggedy zines in the punk rock world,” Jaafari said. “I’m stealing every single brand and voice and guide, and everything from the old ’90s Christopher Street magazines and queer magazines … I’m stealing everything that is cool in our community and putting it into this.”

While the reporting is mission-driven, he doesn’t consider what he does as activist or movement journalism. “We don’t pull punches within our own community,” he said. “That, to me, is a defining factor of whether or not you’re advocacy.”

Earlier this year, reporter Tori Gantz exposed how two of Arizona’s most prominent LGBTQ+ advocacy groups were working on legislation that would allow pastors to discriminate against same-sex couples by not performing weddings. Jaafari reported a story that was critical of a popular Phoenix drag bar and how its actions were causing a rift in the community.

Jim Small, editor of the nonprofit Arizona Mirror digital news site, describes LOOKOUT as filling a needed gap in the state’s media ecosystem. Even when it reports on major stories that other outlets are also covering, he said, it takes a different approach and angles that provide valuable context and insight.

The Mirror re-publishes LOOKOUT’s work, but its newsroom has also partnered with it on joint reporting projects. That’s something the founders say they would like to do with outlets in Colorado and elsewhere as they move into other markets.

“The result has always been solid journalism, really unique stories, and doing the thing that I think we all want to do, which is telling stories that wouldn’t have been told otherwise,” Small said.

Partnerships, collaboration, and coordination

Last year, the consultant Fiona Morgan conducted a survey on LOOKOUT’s behalf and held focus groups that included LGBTQ+ people and allies living in western states. The goal was to better identify their needs and also to map what is currently available to them in terms of support.

She put a special emphasis on rural communities, she said.

Hylton said what the team learned from the feasibility studies is that growing throughout the region, and potentially beyond, will take partnerships and collaboration.

“We built this organization in Arizona on partnerships, we did it by listening, we did it by engaging with community first,” he said.

Next month, LOOKOUT plans to travel to a pride event in Pueblo, Colorado, a former major steel hub in the West about two hours south of Denver. Some who have been involved in local media there have complained about feeling neglected by the state’s broader local media support networks. Hylton said LOOKOUT is already in talks with the 5-year-old nonprofit Pueblo Star Journal about potential partnerships.

Expansions can be tricky. What looks good on a whiteboard and sells in a boardroom may crash against the reality on the ground. A potential partner or collaborator may view an emerging news organization as a competitor that could draw away scant resources.

Morgan, who has worked in the nonprofit and philanthropic news-and-information space for years, said she believes the folks behind LOOKOUT know what they’re doing and have a clear vision for where they fit in the media ecosystem.

“They are very sincere about wanting to support the mission by collaborating with the place-based newsrooms and LGBTQ organizations that currently exist, not replacing them,” she said. “I think what’s going to work is that they’re very strategic in how they planned this expansion, and I think they have been extremely thoughtful about it.”

The renewal email included one sentence in all caps: “THIS PRICE WAS SET BY AN ALGORITHM USING YOUR PERSONAL DATA.” (Subscription prices on The Wall Street Journal’s website are a bit lower: $16.25 per week for a print/digital subscription, after a year at 40% off, or $845 a year.) The reader canceled her subscription.

You may have encountered dynamic pricing — sometimes referred to as algorithmic, personalized, or surveillance pricing — when booking a flight or buying groceries on Instacart. It’s when a company’s algorithm decides what to charge you based on your purchase history and other behaviors. And news subscribers are increasingly noticing it when their subscriptions come up for renewal.

“Like many businesses, Dow Jones uses data to inform the marketing, positioning and pricing of certain consumer subscription products,” a Dow Jones spokesperson told me when I asked about our reader’s experience. “We do this responsibly, guided by robust governance and compliance protocols. Adherence to these protocols is actively monitored and enforced by stakeholders across the business including legal, privacy and compliance.”

Earlier this month, NJ.com subscriber Adam Lisberg got a renewal notice stating that his annual subscription price of $130 was determined by an algorithm using his personal data. NJ.com is owned by Advance Local; its website states that annual subscriptions start at $100 for basic access, while a premium subscription with various perks costs $200.

Algorithmic pricing comes to the news business — just got this renewal notice from nj.com that my $130 annual renewal cost was set personally for me. What are other people being charged?

[image or embed]

— adamlisberg.bsky.social (@adamlisberg.bsky.social) July 13, 2026 at 8:41 AM

Lisberg’s Bluesky post caught the attention of other NJ.com subscribers. One was being charged $145. New York Times’ New Jersey reporter Tracey Tully was told her subscription would renew at $175.

Dynamic pricing isn’t illegal in the United States unless it’s deployed deceptively, discriminates against protected classes, or violates antitrust laws. Last month, consumers filed a class action lawsuit against The Washington Post for failing to disclose its use of dynamic pricing. Tim Giordano is a partner at Clarkson Law Firm, the firm behind the Post lawsuit. He said that news subscribers have the right to understand how their personal data is being used.

“Journalism has real value, and news organizations have every right to pursue a profitable business model,” Giordano said. “But individualized pricing determined by opaque AI systems using personal data, and even legally protected characteristics, is outside the bounds of fair competition and consumer protection law. This kind of AI-driven price gouging gates essential information behind a paywall that moves simply depending on who you are. It undermines public trust in one of the core pillars of democracy.”

“It signals that really vulnerable for-profit news companies are racing to the bottom trying to survive on price and subscriber retention,” said Consumer Reports investigative journalist Derek Kravitz, who helped break the story about Instacart’s AI pricing last year. (After the investigation was published, Instacart suspended the experiment.)

Several states have introduced legislation to regulate or ban dynamic pricing. New York’s Algorithmic Pricing Disclosure Act, which went into effect in January, requires companies to tell New York residents about their use of dynamic pricing. California, meanwhile, banned competitors from sharing “common pricing algorithms” with each other to price similar products and services.

My colleague Andrew got a renewal notice for his Wired subscription on June 29 that set his annual subscription price at $40; because Andrew lives in Brooklyn, the renewal notice included the line, “NEW YORK RESIDENTS: PLEASE NOTE THAT WE ARE REQUIRED TO INFORM YOU THAT THIS PRICE WAS SET BY AN ALGORITHM USING YOUR PERSONAL DATA.” (New Yorker and Albany Times Union subscribers have received similar notices.) The email also mentioned that his $40 offer is below the annual rate of $80, “based on your past purchases and engagement with our products.”

News companies may disclose their potential uses of customer data for pricing in their privacy policies in a few different ways. I looked at policies from Dow Jones, Condé Nast, and Hearst. In these policies, the language generally refers to “personalized offers” and states that artificial intelligence and machine learning may be used to accomplish business objectives related to personal data collection.

Hearst, which owns the Albany Times Union, said it deploys dynamic pricing across all of its newspaper markets at renewal time. It did not disclose the retention or churn data, but that the approach “results in fair pricing for subscribers while supporting our investment in local journalism.”

“Like many newspaper publishers, for many years we’ve used a structured pricing approach based primarily on subscriber tenure and, more recently, digital engagement,” a spokesperson for Hearst said in an email. “What’s new is that New York’s law requires us to use specific language to describe this pricing practice.”

Have you been told that your news subscription price was determined by an algorithm using your personal data? Nieman Lab wants to hear from you. Please fill out a short survey here.

The problem, The Wall Street Journal reports, comes down to Google’s web crawlers, which the company uses to both index websites for search and train its AI models. Publishers can turn off the bots that crawl their sites, but doing so also means their site won’t turn up in traditional search. A few years ago, killing search discoverability to save your traffic would have been unthinkable. But, the Journal shows in a chart worth looking at in that story, traffic declined by more than 40% for some publications between June 2025 and June 2026, according to Semrush data. (The Guardian and the BBC saw gains.)

According to Cloudflare, bots now make up more than half of all web traffic, and beginning on September 15, the tech company will block multi-purpose crawlers — those that scrape for both search indexing and AI training — unless site owners opt out.

The Journal piece is the latest in a host of news stories about publishers threatening to leave Google. “We’ve been clear about what we want,” Cloudflare chief strategy officer Stephanie Cohen told Adweek earlier this month. “We want a technical solution that allows you to be discoverable without having to give your content away for free.”

“It’s time to take a stand and say enough is enough,” USA Today Co. CEO Mike Reed told the Journal. The company, which is currently suing Google over an alleged ad-tech monopoly, says it may turn off Google’s access completely if search traffic continues to decline.

“I wouldn’t call it a big decision because we’re blocking other crawlers,” Reed told Adweek. “For those with licensing agreements, they get our content. For those without, we block them.”

USA Today isn’t the only publisher considering blocking Google. According to the Journal, Reddit is reevaluating its $60 million per year deal with Google after seeing traffic declines, and Politico and Reuters are both considering limiting Google bot access and putting up registration walls. Last year, Penske Media sued Google over its AI summaries. People Inc. CEO Neil Vogel told the Journal that “Turning [Google] off and blocking them entirely is 100% on the table.”

Last month, a British regulatory court ruled that Google must allow publishers to opt out of participating in AI features without affecting traditional search results. Google is testing that ability in the U.K. (and has nine months to make the change) but hasn’t rolled it out globally. The Journal reports that the U.K.-based Economist has been “actively discussing” opting out of AI search, but its VP of generative AI (whom you might imagine has a bit of a stake in the whole AI thing) said that opting out of Google entirely isn’t an option because “traffic can be choppy and declining, but traffic is still traffic.”

Time, meanwhile, is trying to monetize the bots by using what the Journal calls “a sort of Trojan horse approach — creating text-based ads that regular readers never see but that AI crawlers might find.” The idea is that AI chatbots might then reference those ads when responding to user queries without knowing they are ads, like some sort of insidious sponsored content.

The fact that publishers are considering opting out of Google entirely marks a significant turning point. For years, anyone who wanted to be found on the internet had to operate by Google’s rules, making constant SEO tweaks to maximize reach. But Google also needs those companies to play ball; without them, its headline product would be gutted.

“No publisher wants to make do without Google, but many are increasingly prepared to do so,” Mark Stenberg wrote in Adweek. “Google, on the other hand, has no such backup plan.”