Get the choice wrong and the cost isn’t just a clunky dashboard. It’s an audience built on infrastructure that quietly works against the thing the creator actually wants, which is readers who eventually pay.

Migrating a list of any real size later, once a creator has actually noticed the platform isn’t converting the way they hoped, is its own slow, imperfect process, so the decision carries more weight up front than most people give it credit for.

What beehiiv actually publishes

In June 2026, beehiiv released its own State of Paid Newsletters report, built from first-party data across the thousands of paid publications running on its platform.

The headline number is blunt: the median newsletter on beehiiv converts 0.62 percent of free subscribers into paying ones. The spread above that median is enormous. Top-decile publications in categories like investing and economy pull conversion rates as high as 18 to 30 percent, while most niches sit closer to that sub-one-percent median.

beehiiv CEO Tyler Denk framed the broader growth trend behind those numbers this way: “Richer tools plus a more mature user base drove the explosion in subscription revenue growth.” beehiiv’s chief customer officer, Preeya Goenka, put a finer point on what actually moves a publication off that low median: “Paid isn’t just a paywall, it’s a product launch with value prop, pricing, offers, automations, nurture sequences.”

What Substack doesn’t

Substack, by contrast, has crossed five million paid subscriptions across its platform as of 2026, a figure the company has shared publicly.

What it hasn’t shared, in any report or disclosure available in 2026, is a comparable conversion rate: what share of the free readers on a typical Substack actually become paying ones. The handful of specific percentages floating around the internet attributed to Substack trace back to aggregator sites and unsourced “statistics” roundups, not to anything Substack itself has published or that a credible third party has independently verified.

That’s worth sitting with rather than glossing over. A creator comparing these two platforms in 2026 isn’t actually looking at two comparable numbers. They’re looking at one platform that has chosen, for whatever mix of confidence and marketing reasons, to publish granular conversion data down to the industry vertical, and one that hasn’t published anything close to it.

Why the asymmetry itself is the real story

It would be more satisfying to hand over two clean percentages and let a reader pick the higher one. But treating beehiiv’s self-reported 0.62 percent median as directly comparable to a Substack number that doesn’t actually exist would be its own kind of misleading, dressing up a one-sided data point as a head-to-head contest.

beehiiv has an obvious interest in telling a strong paid-newsletter growth story, since it competes directly for the same creators Substack already hosts, so even its own number deserves the same scrutiny a creator would apply to any vendor’s benchmark report.

What the two platforms do differ on, in ways that are independently documented rather than self-reported, is how they structure the free-to-paid conversion mechanism itself.

Substack tends to lean on mid-article paywalls that interrupt a piece partway through, which can convert aggressively but also frustrates free readers faster. beehiiv leans more on inline upgrade prompts inside otherwise-free issues, a gentler nudge that trades some immediate conversion pressure for a longer runway with the free list. Neither approach is simply better. They’re different bets on the same tradeoff, and which one fits depends entirely on what a specific newsletter is actually trying to build.

A broader lesson sits underneath these two specific platforms. Any vendor benchmark report, from any company selling a tool, is worth reading the same way: as a genuinely useful data point, and also as a document written by a party with a stake in the story it tells. beehiiv’s numbers are more specific and more useful than anything Substack has put out, which is worth crediting. That doesn’t make them neutral.

Where I’ve landed thinking about this for my own newsletter decisions

Content strategy is most of what I actually do for a living, across several very different properties, and platform choice is exactly the kind of decision that looks like a spreadsheet problem until you notice half the spreadsheet is missing real numbers for one of the two columns.

My honest takeaway isn’t “pick beehiiv because it publishes better data” or “avoid Substack because it doesn’t.” It’s that a platform’s willingness to publish its own weak points, alongside its strong ones, tells a creator something about how that company thinks about the relationship, separate from whichever platform’s mechanics happen to fit a given audience better.

Across the different properties I work on, the honest answer to “which platform converts best” has never been a single number anyway. It’s been a much slower process of watching what a specific audience actually does, then adjusting the offer instead of assuming the platform will do the work alone.

Even beehiiv’s own report closes on advice that has nothing to do with which logo sits on the dashboard, urging creators to test their own approach to pricing and value rather than copying a template. That holds regardless of which platform’s paywall a creator ends up using, and it’s a more honest starting point than chasing a clean percentage that, for at least one of these two platforms, simply isn’t public yet.

The post beehiiv’s 2026 benchmark data shows a median newsletter converts well under one percent of free subscribers into paying ones, while Substack still discloses no comparable conversion figures of its own, a transparency gap that creators researching where to launch a new paid newsletter are only now starting to take seriously appeared first on The Blog Herald.

It’s a tidy story. It’s also wrong in a specific, useful way, because it treats every blog as if it’s competing for the same kind of search, when the data shows something much more uneven happening underneath the panic.

The overall picture, before the split

Start with the baseline, because the shift is real even before you get to which categories are hit hardest. Researchers Athena Chapekis and Anna Lieb, publishing for the Pew Research Center, tracked nearly 69,000 real Google searches from a representative panel of US adults and found that “Google users who encountered an AI summary clicked on a traditional search result link in 8% of all visits.”

Users who never saw a summary clicked through roughly twice as often, in 15% of visits. That finding comes from a broad, methodologically careful look at ordinary search behavior across a real cross-section of searches, and it confirms the basic mechanism: when Google answers the question itself, fewer people keep going.

What that Pew data doesn’t do, deliberately, is break the result down by topic. It’s an average across every kind of search, which is exactly where the “every blog is doomed” story starts to fall apart, because an average hides enormous variation underneath it.

What actually decides who gets hit

A 2026 keyword study by marketing researchers Trevin Shirey and Macy Storm, who analyzed 2.37 million US search terms, gets at that variation directly. Some categories saw AI Overviews on roughly a third or more of their searches, among them pets and animal content, travel guides, mental health topics, and personal finance. Other categories saw them far less often.

The study’s own explanation for the gap is blunt: “If you rely heavily on informational queries to drive revenue, AI Overviews will hit you harder than an industry that relies more on transactional queries to drive revenue.”

That’s the one thing the split comes down to. The topic itself doesn’t decide it, and neither does how well-written the content is or how much domain authority a site carries in the old SEO sense. What decides it is whether the underlying question is one Google can fully answer on the results page, or one that still requires the reader to go somewhere and do something.

The same study found AI Overviews showing up on “39% of them producing AI Overviews” for informational searches, against 15.8% for transactional ones. A search for what a word means, how a condition is typically treated, or what year something happened is a question with a finished answer. A search meant to end in a booking, a purchase, or a specific account login generally isn’t, no matter how confidently an AI system tries to summarize it.

It’s worth sitting with why that split exists instead of just accepting the numbers. An AI system summarizing a page is essentially doing extraction. It pulls the checkable fact out of the writing and discards the rest, because the rest wasn’t the point of the query.

A reader asking what year a law took effect, or what the typical dosage range for a supplement looks like, wanted the fact, not the writer. A reader asking which specific hiking trail near a specific town is actually worth the drive, or whether a specific product held up after six months of real use, wanted a person who’d actually been there or used the thing. That second kind of question resists extraction because no single correct fact exists to pull out, only someone’s account of it.

The properties I watch don’t all move the same way

I manage content across several publishing sites at once, aimed at different audiences and built around different kinds of questions, and this pattern tracks with what I actually see in the dashboards, even without pretending I have a controlled study of my own.

The pages built around a pure factual question, the kind with one clean, extractable answer, behave differently from the pages built around a specific product, a specific place, or a specific person’s actual experience with something. The first kind is exactly what an AI summary is built to replace. The second kind is harder to compress into three bullet points and a citation, because the value was never really the fact itself. It was the specific voice delivering it.

A fair amount of my actual job is title generation and figuring out how to angle a piece so it earns attention instead of assuming it deserves it, and that work has gotten more interesting rather than less useful since this shift accelerated. A generic, fact-first headline is now competing directly with a machine that can answer the underlying question before a reader ever reaches the page. A headline built around a specific, opinionated angle, one an AI Overview has no equivalent version of because there’s no single fact to extract, is still doing the job a headline is supposed to do.

Where that leaves the rest of us

I don’t think this lets anyone off the hook, and I’m wary of the version of this argument that turns into “just make better content” as if that alone solves a structural shift in how search engines work. But panicking about a single average number, without asking whether your specific blog lives on the informational side of that split or the side that still requires a reader to click through and do something, wastes energy that could go toward an actual decision.

The honest question is whether the specific questions your blog answers are the kind a summary box can finish in two sentences, or the kind that still needs you, rather than whether AI answer boxes are a threat in some general sense.

I keep coming back to something I believe about most hard problems, which is that the mistake is usually treating a big, scary trend as one giant thing instead of breaking it into the specific pieces that actually apply to you. “AI is killing blogs” is the giant, scary version. “My mental-health explainer competes directly with an instant summary, and my review of a specific product doesn’t” is the version you can actually act on.

The post AI answer boxes are reportedly wiping out a majority of organic search traffic for some blog categories in 2026 — while others have barely lost a visitor, and the split comes down to one thing appeared first on The Blog Herald.

The instinct that keeps getting proven wrong is the one that wants to build up to a point slowly, set the scene, earn the payoff. Readers, it turns out, are not reading it that way at all.

The pattern researchers actually found

The original research behind this comes from Nielsen Norman Group, which tracked the eye movements of 232 users across thousands of web pages back in 2006 and kept studying the same behavior for years afterward.

The shape that showed up again and again looks roughly like the letter F: a horizontal sweep across the top of the content, a second, shorter horizontal sweep somewhat further down, and then a vertical drift down the left edge of the page as the reader’s eyes give up on reading in full and start scanning instead.

It isn’t a rigid rule so much as a description of what tired, busy attention does by default, and the group has always been careful to frame it that way rather than as a fixed law.

The exact shape shifts with layout, with how a page uses images or bullet points, and with whether someone’s reading on a phone or a desktop screen. What doesn’t shift much is the underlying habit driving it: most visitors decide fast whether a page deserves closer reading, and they make that call from the top of the content down, not after finishing the piece.

A follow-up piece from the same research group, updated as recently as this year, confirms the pattern held up well past its original discovery: “Our recent eyetracking research shows that the F-shaped scanning pattern is alive and well in today’s world, both on desktop and on mobile.” Nearly two decades and an entirely different generation of screens later, the basic shape of how people scan hasn’t changed as much as most of us assume.

Why this actually matters for anyone writing to be read

The practical consequence is blunt. If most readers only genuinely absorb the first couple of horizontal sweeps before dropping into scan mode, then anything essential that shows up after that point is competing with a reader who has already mentally decided how much attention this page deserves.

Jakob Nielsen’s own guidance on this, based directly on the eye-tracking data, doesn’t leave much room for a slow build: “The first two paragraphs must state the most important information.”

Put the actual point in the first two paragraphs

Not a teaser for the point. Not the color and mood that leads up to the point three sections later. The actual finding, claim, or answer, stated plainly, before the reader has decided whether this piece is worth their full attention.

A slow build might read beautifully to the one person editing it closely, which is exactly the problem, since that person already knows where it’s going.

Make the words that start each line do real work

Since readers scan down the left edge once they’ve stopped reading closely, the words sitting at the start of subheads, paragraphs, and bullet points matter more than the words buried in the middle of a sentence.

Kara Pernice’s more recent research at the same group put it directly: “Include the most important points in the first two paragraphs on the page,” and the same logic extends downward, into every subhead a scanning eye actually lands on.

Stop treating structure as decoration

A subhead that says something vague and clever instead of something specific is invisible to a reader who’s already scanning rather than reading. This is the one that took me longest to actually internalize, because a clever, understated subhead often feels like better writing in isolation. It isn’t better writing if nobody in scan mode understands what the section beneath it is actually about.

Read your own draft the way a scanner would, not the way its writer does

The hardest habit to build is reading a finished draft as if seeing it for the first time, tired, mid-scroll, half paying attention. Cover the piece with one hand, or read only the subheads and first lines of each paragraph in order, and ask whether that skeleton alone actually tells the story. If it doesn’t, the full version probably isn’t landing for the large share of readers who never make it past that skeleton anyway.

Where I’ve had to unlearn my own habits

Titles get almost all the attention in my job, and for good reason, but the body of the piece is where I used to let old habits slide. I’d write a clean, punchy headline and then let the actual article wander for two or three paragraphs before getting to the thing the headline promised, on the theory that a reader who clicked through has already bought in and will wait for the reveal. The eye-tracking data says otherwise. A reader who clicked through is still deciding, in the first few seconds, whether this particular page is going to reward the time they’ve already spent.

None of this means writing gets to be flat or joyless, front-loaded information and a piece with personality aren’t in conflict. It just means the personality has to sit around the point, not stand in front of it for two paragraphs while the reader decides whether to keep going. Nearly twenty years of the same eye-tracking finding is a long time for a piece of research to hold up unchanged, and it’s a strong enough signal that I’d rather build my writing habits around it than around whatever feels more satisfying to draft.

The post Eye tracking research on how people actually read web pages found that most visitors scan in a rough F shaped pattern down the left side of the screen, which is why professional editors front load the real information instead of slowly building up to it appeared first on The Blog Herald.

That distinction matters, because the print era and the digital era genuinely aren’t the same product wearing different clothes. A printed paper arrives on a schedule you don’t control, all at once, in an order an editor chose for you. A digital edition updates through the day, gets read on a phone during a lunch break instead of at a kitchen table, and can be searched instead of flipped through. Both can carry the same reporting. The experience of receiving it is a different thing entirely, and for a long time that difference was exactly why print held on even as its economics stopped making sense.

MLive president John Hiner didn’t frame the decision as a mystery. “What has happened is, logistically, as the population has gotten older, subscribers have died,” he said, describing a subscriber base that was shrinking for the plainest possible reason, not a sudden loss of interest in local news. Delivering a physical paper to fewer and fewer addresses spread across a wide area gets more expensive per subscriber every year, until the math stops working regardless of how good the reporting inside it is.

It’s worth sitting with how long that math has actually been breaking down, rather than treating September’s announcement as sudden. The papers involved here trace back well over a century in some cases, through decades where a subscription was close to a default household expense in the towns they covered, not an optional add-on competing with a dozen streaming services and a news feed that refreshes itself. A gradual shift from “everyone in town gets the paper” to “most people in town read the news on a phone they already had in their pocket” doesn’t announce itself with a single dramatic headline. It shows up as a slowly thinning subscriber list for twenty straight years, until a company finally says out loud what the numbers have been saying quietly the whole time.

The digital side of that same ledger tells the other half of the story. “More than 99% of the company’s readership comes from digital formats,” Hiner said, and digital subscribers now outnumber print subscribers outright. Michigan isn’t unusual here. More than 3,500 newspapers nationwide have stopped printing over the past two decades, according to tracking by the Medill Local News Initiative at Northwestern University, and MLive’s eight papers join a list that’s been growing for a generation, not a sudden 2026 trend. Seen against that backdrop, the September announcement reads less like a company giving up on print and more like one finally acknowledging a shift its own numbers had already made.

What’s genuinely different this time is the tone of the announcement. Hiner didn’t pretend the print edition doesn’t carry real weight for the communities it served. “It’s nice to have a name on a building,” he said, acknowledging the pull of a masthead people grew up with, “but we’ve grown the brand in a lot of different ways that are just really in keeping with the times that we live in.” That’s a company choosing to say the sentimental part out loud instead of burying it under a press release full of “strategic realignment” language, which is worth noticing even if it doesn’t change the outcome for anyone who liked holding the paper.

I think about this kind of shift more than the average reader, because several of the properties I work on are built for an audience in their fifties, sixties, and seventies, exactly the readers most likely to have grown up with a physical paper as the default and to feel a real loss when it goes away, even while reading everything else on a screen. A masthead disappearing and a newsroom disappearing are genuinely different events, and it’s worth being precise about which one actually happened here. MLive’s reporters are still covering Ann Arbor, Flint, Grand Rapids, and the rest. What’s gone is one specific, aging way of getting that reporting into a reader’s hands, not the reporting itself.

The workshops MLive is planning speak to the actual thing readers stand to lose here, and it’s a smaller thing than “the news” but a real one anyway: a habit. Reading the paper over coffee, in a specific chair, in a specific order, is a ritual as much as an information source, built over years and not easily replaced by an app that technically contains the same stories in a different shape. Losing that ritual is a legitimate thing to grieve a little, separate from any argument about whether the reporting itself survives the transition. Both things can be true. The coverage of Ann Arbor and Flint and Grand Rapids keeps going, and something specific about how a longtime subscriber used to start their morning is still gone.

MLive says it will run workshops to help longtime print subscribers move to the digital editions before December, which is a smaller, more practical kind of help than a company owes its readers in a moment like this, but not a nothing one either. For a reader who’s spent forty years starting the morning with ink on their fingers, that’s a real adjustment, not a footnote, and it deserves to be treated as one instead of waved off as a formality on the way to a cleaner balance sheet.

The post MLive announced in September 2026 that it will end print production entirely for eight of its Michigan newspapers, shifting those communities to digital only coverage after years of steadily declining print subscriptions across the whole state appeared first on The Blog Herald.

Nota had pitched itself as a fix for the news desert problem, using data from Northwestern’s Medill School to pick counties with thin or vanished local coverage, from California to Georgia, and building AI-written sites there instead. The pitch was that a community losing its last reporter didn’t have to lose local news entirely.

A news desert is a real place, not an abstract policy term: a county where the last local paper folded, or shrank to a skeleton crew covering a wider area than three reporters can reasonably drive to in a week. The people affected still have school board decisions, zoning fights, and local elections happening whether or not anyone’s writing them down, and Northwestern’s own tracking has documented the spread of these gaps across the country for years.

Nota positioned itself as the response to exactly that documented, real problem, which is part of why the plagiarism findings landed as hard as they did. This was a company claiming to solve a genuine civic problem while quietly depending on the very reporters it said the problem had left behind, a much heavier failure than simply cutting corners on a low-stakes product.

That was the claim. Here’s what actually happened.

In Henrico County, Virginia, publisher Tom Lappas of the Henrico Citizen found that multiple stories on Nota’s local site were lightly reworded copies of his outlet’s reporting, with staff photos used without permission. His was one of at least 29 outlets whose work turned up, lightly rewritten but structurally intact, across more than 70 Nota stories, according to reporting from the journalism nonprofit Poynter.

At least 53 individual journalists had their reporting used without attribution. Lappas didn’t hear from the company directly once the story broke. “No one from Nota has reached out to apologize or explain how this happened,” he said.

The idea that AI could fill a reporting gap always rested on a specific assumption: that the shortage in a news desert is a shortage of writing. Get software to produce enough local stories, the thinking goes, and the coverage gap closes. What actually happened at Nota shows the assumption was backward from the start. Instead of generating original local coverage to fill an empty space, the company took coverage that already existed, produced by the exact reporters it claimed to be helping, and republished it as its own.

Even people inside the operation ended up saying as much. Jorge Rodríguez, one of Nota’s own reporters, told Poynter plainly: “I don’t believe Nota is a journalistic company at all.” CEO Josh Brandau didn’t dispute the core failure either, telling Poynter, “The fault absolutely lies with us in terms of keeping it open broadly to the web.”

What’s true, once the excuses fall away, is something local journalism has been saying for a while: reporting a place well requires someone who actually shows up to it. A city council meeting, a school board fight, a small business opening on a street that used to have three empty storefronts. Nobody has built software that attends those in person, asks a follow-up question when the official answer doesn’t add up, or knows which neighbor to call back. Nota’s own failure mode proves the point better than any defense of local journalism could. Faced with a genuine gap in original reporting, the fastest available option was to take someone else’s.

None of this means AI has no legitimate place in a newsroom. Plenty of outlets use it for transcription, data cleanup, or a first pass at a routine public-records summary, with a human editor checking the result before it runs. The difference is whether a human being who actually did the reporting is still in the loop, getting credit and getting paid. Nota’s model tried to remove that person from the equation entirely and still call the result local news.

What makes this specific failure sting more than a generic AI-content scandal is who it actually hurt. Nota republished the work of small outlets and individual local reporters, exactly the people already stretched thinnest by the coverage gaps Nota claimed to be filling, rather than anything belonging to a large media company that could absorb the loss. A freelance reporter in a shrinking newsroom, covering three towns because there’s no one left to cover the fourth, loses real, countable value when a better-funded AI product republishes their reporting without payment or credit. That harm already happened, to at least 53 people with names, beats, and editors who trusted them to file the story once, not to have it filed again somewhere else without their name on it.

I manage content for a living, some of it republished or reframed from existing material with permission, sourced, and credited by design, and the line Nota crossed is one I’d notice immediately in my own work, because it’s the same line that separates a legitimate rebrand from a lift.

Attribution is the part of the job that says, plainly, whose work this actually is, not a legal formality tacked onto the end of a piece. A company that skips that step is taking credit for someone else’s reporting and calling the result innovation, whatever softer word gets used in its own materials.

For readers, the honest takeaway is a smaller, more practical one than a lecture about AI in general. When a “local news” site with a name you don’t recognize shows up in a search result for your town, it’s worth a quick check of a byline and an about page before treating it the same as an outlet with a masthead and a phone number that rings through to an actual desk.

The reporters covering your specific city council meeting are doing something a script, however well trained, still can’t fake convincingly enough to survive scrutiny. Nota found that out the hard way, and so did the 53 journalists whose work it borrowed without asking.

Correction, October 4, 2026: The earlier headline incorrectly said Nota shut its entire operation. The reported closure was its eleven-site local news network, not the whole company.

The post Nota shut its eleven-site AI local news network in 2026 after journalists found their reporting copied without attribution, while the company’s separate publishing-tool business remained distinct from that failed experiment appeared first on The Blog Herald.

That’s an ordinary, slightly boring confession, and it’s exactly why a stat like this one stopped me mid-scroll instead of sliding past.

The number is 68.01 percent. That’s the share of Google searches in the United States that ended without a single click to any website in the first four months of 2026, according to an analysis the research firm SparkToro published using Similarweb’s clickstream data.

Flip it around and you get the headline: fewer than one in three searches sends a person anywhere at all anymore. Most of the time, Google answers the question directly on the results page and the visit simply never happens.

That 68.01 percent isn’t one single thing. Some of it is searches Google has always answered on the page itself, like a quick weather check or a math conversion, the kind of query nobody expected to produce a click even a decade ago. A growing slice of it is newer: AI Overviews summarizing an entire topic, pulling from several sources at once, and putting a finished answer above every blue link on the page. The searcher gets what they came for. The sites that supplied the material for that answer, often several of them stitched together, get nothing they can measure.

Digital marketing researcher Rand Fishkin, SparkToro’s co-founder, doesn’t soften the framing. “Thanks to AI features, instant answers, UI elements that keep searchers in the results, and shifting user preferences, Google is becoming a walled garden,” he wrote in the report.

That description matches what the clickstream numbers show, and it’s been building for years, from featured snippets to knowledge panels to, now, AI Overviews answering entire questions before a reader scrolls past the fold.

For a blog, the mechanism is simple even if the feeling isn’t. A search engine used to be a hallway that led somewhere. Increasingly, it doubles as the destination itself.

Here’s the part I’ll say plainly: for years, visitor count was treated at every property I’ve worked on as close to the entire scoreboard. That made sense when search reliably delivered people to a page. It makes a lot less sense when the search engine keeps most of them for itself and hands publishers the scraps.

A number that used to mean “the work is reaching people” now sometimes just means “the work is good enough that Google borrowed it to answer someone’s question, without sending them anywhere to read it.”

Fishkin’s advice to publishers watching this happen isn’t to fight harder for the old kind of click. “Replace traffic as a KPI for your digital marketing efforts,” he wrote, adding plainly: “Build a correlation dashboard instead.” The idea is to track whether your content’s visibility, even in zero-click form, correlates with outcomes that actually matter, whether that’s brand recall, direct visits from people who already know your name, or something else entirely, rather than treating raw click-throughs as the only metric worth watching.

Easier said than done, obviously. A correlation dashboard doesn’t produce the same satisfying jolt as watching a page count tick upward in real time, and plenty of the people this affects are paid, at least in part, on numbers that were designed for a version of the internet that’s already receding. I don’t have a tidy fix for that mismatch between what actually matters and what still gets measured. I just think it’s worth naming out loud instead of quietly pretending the old scoreboard still tells the whole story.

I won’t pretend that’s a comfortable shift for anyone whose job has revolved around one clean number for years. It isn’t comfortable for me either. But it lines up with something I already believe about work in general: the mistake is usually treating a big, vague goal as one giant unmovable thing instead of breaking it into smaller pieces you can actually check. “Grow traffic” was always a strange target to measure yourself against season after season, given how much of it depends on a platform nobody in this line of work actually controls. Watching that platform openly admit it’s becoming a walled garden is, oddly, permission to stop pretending otherwise.

What that looks like in practice, for me and for anyone else doing this work, is still being worked out in real time. Nobody I know has a finished playbook yet. What I’ve noticed, watching the dashboards rather than reading about them secondhand, is that properties built around a genuine, specific audience, the ones people seek out on purpose instead of stumbling into from a generic search, seem less shaken by a number like 68.01 percent.

Some of the sites I work on are written for women in their fifties, sixties, and seventies who came looking for that specific voice and keep coming back for it, not for whoever ranks first this week. That’s an observation from someone who checks these numbers for a living, not a settled fact, but it’s a pattern I keep noticing across different properties, in different niches, for different reasons.

None of that makes the 68.01 percent figure less real, and I’m not interested in pretending a loyal core audience cancels out a structural shift this large. It just suggests the sites that survive this era well probably won’t be the ones that chased the algorithm hardest. They’ll be the ones a reader actually remembers the name of.

If you run a blog and just felt your stomach drop reading this, you aren’t imagining the shift. The search traffic era isn’t fully over, but it plainly isn’t the era most of our habits were built around anymore, and treating that as the actual starting point beats mourning a number that was already changing underneath us.

The post Less than one in three Google searches actually sends a visitor to someone else’s website anymore — here’s what that 2026 number means if your blog still lives on search traffic appeared first on The Blog Herald.

She had been born in 1899, two years before Federation gave Australia the borders and the government it still has. That birth year put the First World War, fought from 1914 to 1918, in her late teens and early twenties. It put the Depression that followed the 1929 crash in her thirties. It put the Second World War, from 1939 to 1945, in her forties. She didn’t write about any of it as background history. It was just the middle stretch of an ordinary working life, the part most people her age had lived through too, except almost none of them were left to describe it firsthand anymore.

One of the people who helped her was a filmmaker named Michael Rubbo, to whom she dictated her posts. Together they called the blog The Life of Riley, and filled it with whatever Riley felt like telling people that week. There was no set schedule and no obvious theme beyond whatever had happened to her lately, which is a large part of why it reads less like a memoir project and more like an actual blog, written in real time by someone who happened to be a hundred and seven.

Her posts amounted to a kind of memoir almost nobody else still alive could have written: growing up in the outback mining town of Broken Hill, working as a farm cook and a bartender, and raising three children. Over the next year and a half she published 74 of them.

Blogging by 2007 was already close to a decade old as a mainstream habit, and it was mostly a habit of people young enough to have grown up around computers, typing in real time about their own lives. A dictated blog from a woman old enough to remember a world before radio broadcasting didn’t fit that pattern at all. It skipped the usual barrier to entry entirely: she never had to learn the software, never had to sit through the part where a new user gets frustrated and gives up. Somebody else handled the mechanics. All Riley had to do was talk, which by then she’d had a hundred and seven years of practice doing.

The blog found readers fast, and not because a 107-year-old blogger was a novelty act. She stayed in contact with correspondents worldwide, and in one entry she thanked them for “a whole swag of e-mails and comments from my Internet friends,” which reads less like someone managing an audience and more like someone genuinely surprised to have one. Coverage of her settled on the same shorthand, the one used in the headline announcing her death: the world’s oldest blogger.

Her family noticed the shift in her too. Her great-grandson, Darren Stone, later described how startled she was by the scale of it all, a woman well past a hundred who was suddenly hearing from strangers on other continents who genuinely wanted to know what she thought.

That went on until June 2008, when Riley fell ill and moved into a nursing home in Woy Woy, on the coast north of Sydney. She kept posting from there too, mentioning at one point that she was feeling weak with a cough that wouldn’t go away. The entries stayed just as warm as the illness allowed. She said of the care she was getting, “I’ve never been treated so well in all my life. The nurses can’t do enough for me.”

Her last post went up on June 26, 2008. In it, she described an ordinary afternoon with the woman in the next bed, whose daughter happened to be a professional singer: “She and I sang a happy song, as I do every day, and before long we were joined by several nurses, who sang along too. It was quite a concert!”

Two weeks later, on July 12, 2008, Olive Riley died. She was 108. By one count she left behind hundreds of descendants and thousands of readers online who had never met her, all of whom had spent the past year and a half checking the same blog for the same reason: to find out what she’d noticed that week.

A hundred and eight years is already a long stretch of personal history to have lived through, most of it before anyone thought to write any of it down. Families usually end up reconstructing that kind of life secondhand, piecing it together from other people’s memories once the person who actually lived it is gone. What made Riley’s final year and a half different was that nobody had to do that work for her. She was there for the telling, dictating one sentence at a time to somebody willing to sit down and type it, describing the wars and the hard years and the small good afternoons in her own words while she still could, rather than leaving it to whoever came after her to guess at the details.

What’s easy to miss in a headline like “the world’s oldest blogger” is how ordinary the actual writing was. Riley wasn’t performing wisdom or delivering some grand summary of a century. She was mostly just describing her week, the way anyone with a blog does, except her week happened to include a childhood in outback New South Wales, a working life that spanned two world wars, and a nursing home room she shared with a woman who liked to sing. That gap, between the size of the life and the smallness of the details she chose to write about, is most of what made people keep reading.

The post In 2007, a 107-year-old Australian woman who could no longer type began dictating a blog about surviving two world wars and the Great Depression, and kept posting until two weeks before she died at 108 appeared first on The Blog Herald.

Two pools, then a second cut

YouTube’s August 2026 Partner Program update, whose terms take effect February 1, 2027, lays out two separate revenue pools tied to its subscription tiers: creators share in 30% of net subscription revenue from standard Premium, and 60% from Premium Lite, the cheaper, ad-reduced tier YouTube has been rolling out market by market. Premium Lite still runs ads on Shorts, music and search. YouTube doesn’t explicitly tie that carve-out to the pool sizes, but by its own account the split “factors in the costs of operating and promoting the service, including amounts we pay to music partners” — leaving more of each subscription fee inside the pool creators draw from. Once that pool is allocated to individual creators based on member watch time and views, YouTube applies a second cut on top: 55% of a Premium Lite payout goes to long-form video, 45% to Shorts. Working through ppc.land’s breakdown of the announcement, that puts long-form creators at roughly a third of the Lite pool’s net revenue and Shorts creators at just over a quarter.

A ratio YouTube didn’t invent for this

What makes 55/45 worth pausing on is that YouTube didn’t design it specifically for Premium Lite. It’s the same split the platform has used for years to divide ordinary ad revenue between the two formats: per YouTube’s own Partner Program documentation, creators keep 55% of net ad revenue from a long-form video’s watch page, and 45% of the revenue allocated to them from the Shorts Creator Pool. The same announcement confirms the 55/45 formula applies to the standard Premium pool as well as Premium Lite. Two structurally different revenue streams, ads and subscriptions, spanning two different price tiers, and YouTube keeps arriving at the identical ten-point gap in favor of the longer format.

Where the raised bar fits in

The same announcement doubles what a new creator needs to enter the Partner Program at all: 8,000 watch hours or 20 million Shorts views in the trailing year or 90 days respectively, up from 4,000 hours or 10 million views, starting February 1, 2027, with the 1,000-subscriber requirement staying in place. Most coverage of that change has focused on who it locks out of the program entirely. Less discussed is what happens to the creators who clear the new bar either way. YouTube’s own announcement puts the program at over 3 million creators, and every one of them is entering, or already sitting inside, a payout structure where the format decision was made for them well before this update and simply carried forward, at the same ratio, into every new revenue stream YouTube has added since. A Shorts creator already earning has to keep clearing 10 million Shorts views every 90 days just to stay inside the Shorts Creator Pool at all — a maintenance bar that exists entirely apart from the 55/45 split waiting on the other side of it.

The dollars behind the percentage points

The pools involved aren’t small. YouTube and Google One together reported 325 million paying subscribers as of the fourth quarter of 2025, a jump of 25 million in three months, while total YouTube revenue for the year reached $60 billion, up 17% year over year; that outpaces the 9% year-over-year rise in Q4 ad revenue specifically, though the two figures span different windows — full-year versus a single quarter — and aren’t a precise apples-to-apples comparison. Subscriptions are the part of YouTube’s business expanding fastest, and Premium Lite, recently raised from $7.99 to $8.99 a month in the US, is the tier built specifically to convert viewers who’d otherwise watch free with ads. YouTube has told creators plainly that “partners, on average, earn more than when the user was watching ads” once someone subscribes, which means every point of the 55/45 split is moving money through the fastest-growing part of the business rather than a shrinking one.

The volume doesn’t match the split

That ratio runs against how people actually watch. YouTube’s own figures put Shorts at over 200 billion views a day worldwide, a volume long-form video isn’t close to matching by view count. But the revenue math doesn’t run through view count directly — it runs through watch time first, then through a fixed percentage that was set independent of how any individual video performs. A Shorts creator can out-view a long-form creator by orders of magnitude and still end up dividing a smaller collective share of both the ad pool and the subscription pool. Popularity and payout rate are separate variables here, and YouTube’s formulas have never made them move together.

YouTube hasn’t published a rationale for why long-form gets the larger cut in either pool — its own materials explain the 60/30 gap between the two subscription tiers but say nothing about why the underlying 55/45 format split exists, or why it shows up twice in unrelated systems. It may simply be inherited: once a ratio is built into the ad-revenue infrastructure, reusing it for a new subscription pool is the easier engineering decision, not necessarily a fresh statement of priorities. Whatever the reason — deliberate prioritization or simple inheritance from the ad-revenue infrastructure — a company that wanted Shorts and long-form treated as equally valuable had two separate chances, in two unrelated revenue systems, to set that number at 50/50. It didn’t take either one, and YouTube hasn’t said why.

The post Premium Lite’s revenue split favors long-form 55/45 over Shorts — a ratio YouTube has reused, not invented, across its revenue systems appeared first on The Blog Herald.

Take that arrival away and there often isn’t a second engine to fall back on. Large outlets have apps, subscriber lists, and brand recognition built up over years. A two-person regional outlet or a specialist blog usually has none of that cushion. So when the mechanics of a Google search change, the effect doesn’t land evenly across the internet. It lands hardest on the publishers with the least room to absorb it, and for some of them, the traffic loss is close to existential rather than a line item to manage around.

Two data sets released months apart in 2026 put real numbers on that imbalance. Rand Fishkin, co-founder of the audience-research firm SparkToro, found that 68 percent of Google searches now end without anyone clicking through to a website. Separately, Sara Fischer, reporting an analysis of Chartbeat data for Axios, found that the smallest publishers lost 60 percent of their search referral traffic in just two years, a decline roughly three times steeper than what large publishers experienced over the same stretch.

What a search used to owe a website

Searching Google has never guaranteed a click. Some share of searchers have always gotten their answer from the snippet, the weather box, or the calculator built into the results page. But that share used to be closer to a coin flip than a foregone conclusion. Fishkin’s tracking put the zero-click rate at 60.45 percent in 2024, meaning a little under four in ten searches still sent someone to an outside page. By early 2026, that had slid further. “In the first four months of 2026, a whopping 68.01% of Google searches ended without a click,” Fishkin wrote, describing a jump he attributes largely to AI Overviews, the summaries Google began placing at the top of results pages that answer a question directly rather than pointing to a source for it.

Google introduced that feature as a way to have search do more of the work for the person typing the query, pulling information from multiple pages into one synthesized answer instead of asking the reader to visit each source in turn. For a reader, that can save a genuine step. For the sites that used to be that step, it’s a different story. Fishkin’s research also found that when an AI Overview appears above a search result, the click-through rate on the organic listings below it drops by close to 60 percent, which helps explain why the zero-click share keeps climbing even as the total number of searches keeps growing.

Neither study proves that one number caused the other. Fishkin’s figure covers the whole of Google search across every kind of query; the Chartbeat data covers a specific slice of publishers over a specific window. What lines up is the timing. Google’s rollout of AI-generated summaries at the top of search results began broadly in the United States in May 2024, sitting almost exactly inside the two-year period Chartbeat measured. Worth naming plainly: a shared timeline is a correlation, and a correlation is a long way from a proven cause, even when two trend lines move together this closely.

A dip for some, a collapse for others

Averages hide distribution, and this one hides a wide one. The 68 percent figure describes the whole of Google search, not any single publisher’s experience. Fischer’s numbers show the gap directly: “referral traffic from traditional search engines has declined by 60% for small publishers, compared with 47% for medium-sized publishers and 22% for large publishers, per Chartbeat.” In that breakdown, small publishers are outlets averaging between 1,000 and 10,000 daily page views, roughly the size of a well-read local news site or an active hobbyist blog. Large publishers, by contrast, are sites pulling more than 100,000 daily views, the scale of an established national outlet.

A 22 percent drop is a real loss for a large publisher; ad sales teams have built entire forecasting models around smaller swings than that. A national outlet losing a fifth of its search traffic and a small outlet losing three fifths of it are different events by any reasonable measure. The larger site typically still has direct traffic from a loyal readership, an app people open out of habit, or a newsletter list built over a decade. The smaller site, more often than not, was leaning on search to do most of that work in the first place.

Why size changes what the same headline means

Part of the gap traces back to what kind of content each tier tends to publish. Small, independent sites disproportionately run on informational queries: how something works, what a word means, how to fix a specific problem. Those are exactly the queries an AI Overview is built to answer directly, because the answer is short, factual, and doesn’t require the nuance or original reporting that a large newsroom might bring to a breaking story. A well-known outlet covering a fast-moving news event still tends to get the click, because an AI summary of a still-developing story is a riskier substitute than an AI summary of a settled fact.

The other part is capacity. A large publisher losing a fifth of its search traffic can shift budget toward newsletters, apps, or syndication deals and buy itself time to adjust. A small publisher rarely has the staff or the runway to build a second distribution channel from scratch while its main one is shrinking underneath it. Some of the outlets in Chartbeat’s smallest tier are one or two people running a site as a side project or a small business. For those publishers, a 60 percent drop in referral traffic over two years isn’t a metric to optimize around. It’s closer to a countdown.

Taken together, these two studies describe a widening gap more than a single event. Search itself is sending fewer people anywhere at all, and whatever traffic is still flowing out is increasingly flowing toward publishers who already had the resources to survive without it. The publishers who needed that traffic most are the ones now watching the most of it disappear.

The post A 2026 study of search behavior found that 68 percent of Google searches now end without a single click, and small publishers lost 60 percent of their referral traffic in just two years. appeared first on The Blog Herald.

What the survey actually measured

Eb Gargano’s Blogging Income Survey 2025 collected anonymous Typeform responses from 187 bloggers between January 24 and April 21, 2025, drawn mostly from her own email list, course students and Facebook group, skewed toward English-speaking countries and bloggers already focused on monetizing. Alongside total monthly earnings, Gargano’s survey tracks something most income reports skip: RPM, or revenue per thousand pageviews, broken out by how long each respondent had been blogging. RPM strips out the effect of traffic size and asks a narrower question: how much is each visitor worth, on average, once a blogger settles into their monetization mix.

The rate falls exactly where the money should be at its best

Gargano’s cohort breakdown moves in one direction for most of a blog’s life, then reverses. Bloggers under a year in averaged an RPM of $10.99. That climbed to $21.61 for the 1-to-3-year group, $24.97 for 3-to-5 years, and $52.59 for 5-to-10 years — a steady, unsurprising climb as bloggers learn what their audience will pay for. Then, at the 10-year mark, it drops to $43.41. Bloggers who have survived the longest, built the biggest archives and presumably learned the most about their own audience are earning less per thousand pageviews than bloggers who’ve been at it for as little as five years.

Total monthly income follows a different, uninterrupted curve, which is part of why the RPM dip is easy to miss unless a survey specifically isolates it. Average earnings run $2.42 a month under a year in, $205.44 by year one to three, $1,044.91 by year three to five, $2,621.24 by year five to ten, and $5,624.91 past the decade mark. Every stage looks like unambiguous progress if income is the only number on the page. The rate is the number that turns over.

The traffic still wins the month

That drop in rate doesn’t show up as a drop in income, because the 10-year-plus group isn’t playing the same traffic game. Gargano’s data puts average monthly pageviews at 49,841 for the 5-to-10-year cohort and 129,562 for bloggers past year ten — roughly 2.6 times more. Multiply pageviews by RPM and the math resolves the apparent contradiction: $5,624.91 a month for the veterans against $2,621.24 for the 5-to-10-year group, more than double. A decade of accumulated posts, backlinks and search rankings buys enough extra traffic to outrun a worse rate per visitor. It doesn’t buy back the rate itself.

What’s actually dragging the rate down

Gargano’s own read points at monetization mix, not fatigue or algorithm penalties. “Most of the bloggers in this category get the majority of their income from ads,” she writes of the 10-plus-year group. “Many get almost 100% of their income from ads and, according to the data, ads deliver the lowest RPMs of all the different blog monetization methods.” She adds a second, compounding factor: bloggers who’ve been running a site for over a decade are “likely to have a lot of poor quality, out of date and/or unoptimized content” sitting alongside their newer posts, diluting the average value of every pageview the archive generates.

The size of the gap ads are leaving on the table

The survey’s own numbers make the ads-heavy strategy look worse the closer you look. Display ads across Gargano’s full sample averaged an RPM of $33.80. Digital products averaged $283.64 — not a marginal difference but roughly eight times the return per thousand pageviews. A blogger converting even a modest share of ad-supported traffic into email subscribers and a paid product isn’t chasing a nice-to-have upgrade. The survey doesn’t isolate how much of the gap between the 5-to-10-year and 10-plus-year cohorts traces to monetization mix specifically, but it’s Gargano’s own leading explanation for the drop.

What the sample can and can’t tell a reader

Gargano is upfront about who answered the survey: respondents skew toward her own email list, course students and Facebook group, which means the sample leans toward bloggers already treating their site as a business rather than a hobby. A 10-plus-year blogger who never tried to monetize seriously in the first place wouldn’t necessarily show up here at all, and the RPM figures likely understate how much worse the gap looks across the full population of aging, half-abandoned blogs that never had an income-focused audience reading along. If anything, that bias should make the 10-year RPM drop harder to dismiss, not easier — it’s showing up inside a sample stacked with people who are actively trying to monetize well.

This doesn’t mean blogging gets worse with time. The traffic compounds, the archive compounds, and by year ten a blog with any monetization at all is likely earning more in absolute terms than it did at year five. What compounds alongside it, for a meaningful share of long-running blogs, is a growing pile of old content still running the least efficient way to make money on it. The bloggers who avoid the RPM drop are the ones who keep changing what they sell as the decade goes on, rather than letting an aging monetization mix ride on accumulated traffic alone.

The post Bloggers past the 10-year mark earn more than double what 5-to-10-year bloggers make each month, but a 2025 survey of 187 bloggers found their revenue per pageview is actually lower appeared first on The Blog Herald.

What film school actually costs when it isn’t unofficial

The traditional route has a price tag attached to every stage of it. The American Film Institute’s conservatory program runs $72,954 for the 2025-2026 year, and Columbia’s MFA in film runs $77,840 a year in tuition for the first two full-time years, dropping to a reduced “Research Arts” rate for any additional years spent finishing a thesis — a path that can approach $200,000 in tuition alone before housing. Boston University’s film program runs over $90,000 annually; even an in-state public film degree can land between $7,000 and $23,000 a year before production costs. The debt follows students out the door: the Education Data Initiative puts median debt for a film and photographic-arts bachelor’s degree at $26,005, and for a film master’s at $99,757 — more than double the typical master’s-degree debt load, though it’s not the highest figure in that data set; majors like dance and design carry even more. A YouTube upload costs whatever the camera and the creator’s time are worth, and if it fails, nobody owes a lender anything.

The feedback loop film school can’t match

Tuition buys structured critique, but not much of it. A film student gets maybe a handful of graded screenings a semester, judged by a small faculty and a rotating cast of classmates. A YouTube upload gets judged by however many strangers click on it, within hours, in a comment section that doesn’t grade on a curve.

Curry Barker, who dropped out of the New York Film Academy’s Los Angeles campus with creative partner Cooper Tomlinson to start a channel called That’s a Bad Idea, has described the difference less as quality control and more as a cure for overthinking: “Any aspiring filmmaker’s worst nightmare is being paralyzed,” he told The Wrap, describing how the channel forced him to write things he and Tomlinson could actually shoot that week instead of waiting for a script to feel finished. A festival submission gets one verdict, months later. A YouTube upload gets a verdict by the weekend, and there’s a next one already in the queue.

Practice at a volume no curriculum allows

Barker’s public output before his 2025 breakout, “Obsession,” wasn’t one polished thesis film — it was years of visibly uneven attempts. A nine-episode web series called “Roommates” in 2018 and 2019. A 22-minute short called “The Chair” made for around $2,000 in 2023. Then his actual directorial debut: a found-footage horror feature called “Milk & Serial,” shot for $800 in 2024, that sat without a distributor for a year before Barker put it on YouTube for free. Each one is a public rep, watched and judged in real time, in a way a film-school student doesn’t get to log more than a few times before graduation. Volume is the entire mechanism: a curriculum built around four or five supervised projects over four years can’t replicate what a platform allows when the only cost of another attempt is making it.

When the body can’t do it, the software has to

Parsons’ path shows what that practice looks like when it’s forced rather than chosen. His arthritis diagnosis at 13 limited the kind of physical, hands-on production other teenage filmmakers could do — the running-around-with-a-camcorder version of learning the craft. He leaned into software, animation and VFX tools instead, teaching himself the pipeline a film-school cinematography track would have handed him piece by piece over several semesters. By the time “The Backrooms” went viral, he’d spent years uploading practice work nobody outside his own subscriber base was watching closely, which is precisely the condition under which a person is free to be bad at something without consequence.

Failing on camera, on purpose

Danny and Michael Philippou, who built the horror-comedy channel RackaRacka before either of them had a feature credit, describe the same mechanism more bluntly. “You want to make that initial stuff and fail like that,” Michael told No Film School. “That stuff… is a lesson that you take forward for the next thing.” His own follow-up makes the same point more plainly: “The best advice, really, is to just start making stuff.” Danny’s phrasing in the same conversation was blunter still: “Just get started. You have to.” Neither brother frames their early, rougher videos as embarrassing detours on the way to competence. They frame them as the mechanism competence came from — public, low-stakes, and repeatable in a way no film-school assignment schedule allows.

Film school and YouTube are now running in parallel

That doesn’t mean the traditional path is disappearing. Chapman University’s Dodge College of Film and Media Arts had a higher-than-expected yield this admissions cycle and ended up oversubscribed, according to The Ankler — the opposite of what a “YouTube killed film school” narrative would predict. USC film school dean Elizabeth Daley put the moment in context rather than alarm: “This is a cyclical business in terms of new generations coming up.” One path hasn’t replaced the other. A path that used to require an acceptance letter now also has an on-ramp that only requires an upload button, and both are currently producing directors Hollywood wants.

The two pipelines are optimizing for different things. Film school still teaches the parts of the job that a comment section can’t — sound design theory, production management, how a crew actually runs — and a faculty screening still catches craft problems an anonymous audience won’t articulate. What YouTube supplies is the one input structured education has never been able to offer at scale: enough cheap, public, immediate failure to get good before anyone with a checkbook is watching.

The post YouTube didn’t become an unofficial film school by accident. It became one because creators could fail publicly there, over and over, until they got good appeared first on The Blog Herald.

That version of events isn’t hard to believe. AI detection has gotten good enough to flag machine-written prose at scale, and sightings of obviously synthetic listings, ones with nonsensical author bios or covers that look assembled from the same three templates, have been easy to find in genres like romance and thriller. The assumption that follows feels almost like common sense: more AI books on the virtual shelf should mean more AI dollars landing in someone’s pocket.

A 2026 analysis by Tuhin Chakrabarty and three co-authors at Stony Brook University, Columbia Law School, and the University of Michigan tested that assumption directly instead of taking it on faith. The team ran full-text AI detection across 14,419 self-published genre-fiction ebooks sold on Amazon between January 2023 and March 2026, matching each title to its own daily sales record through the end of June 2026.

They used a detector called Pangram v3.3, which its developers report catches AI-generated text with a false-positive rate of just 0.04 percent, and sorted every book into one of three buckets based on how much of its text the tool flagged as machine-generated: none, light exposure at 25 percent or under, and substantial exposure above that threshold. That threshold matters. A book that used AI for a light copyedit or a few transitional paragraphs lands in a different category than one built mostly from generated text, and the study treats them separately rather than lumping every book that touched a chatbot into one pile.

The results complicate the flooding story rather than confirming it. According to the study, books with substantial AI text made up 20 percent of the catalog analyzed, but captured only 12.1 percent of unit sales and 11.3 percent of revenue. Books with no detected AI text, by contrast, made up 62.9 percent of the catalog and pulled in 72.5 percent of the revenue. As the researchers put it, these books “make up a large share of the catalog but a smaller share of sales.”

That gap is real. It doesn’t cancel out what comes next, though. The same data show AI-heavy books gaining ground over time, winning a growing share of sales and taking more of the scarce top-rank positions once held by books with no detected AI text. A fifth of the catalog earning an eighth of the money looks, at first glance, like proof the flooding strategy is failing. Followed across three years, it starts to look more like a foothold.

The bigger shift might not be about AI-generated books specifically at all. Across the period studied, the researchers found that the cumulative catalog of released titles grew 38.3-fold and the number of books selling in a given quarter grew 19.2-fold, while quarterly revenue grew only 8.9-fold. Thousands of new titles, AI-assisted and human-written alike, poured onto virtual shelves far faster than buyers showed up to pay for them. Revenue per selling book fell across most genres, and books with no AI text lost the most ground specifically in genres where AI-generated titles had spread the furthest, especially where Kindle Unlimited availability was high.

Kindle Unlimited matters here because it changes how a book earns money in the first place. Instead of a single purchase price, KU pays authors per page read from a shared pool, which rewards volume and frequent new releases more than it rewards any single title selling well. A catalog strategy built around publishing often, even with heavily AI-assisted books, can do reasonably well under those incentives even while any individual title underperforms on its own. That helps explain why the erosion for human-written books wasn’t evenly spread. It concentrated in exactly the genres and pricing models most exposed to volume publishing.

A second finding buried in the same dataset speaks less to sales and more to originality. Among top-selling titles, the researchers found that books with substantial AI text drew on more distinctive language already present in existing books than human-written top sellers did, and that this overlap climbed alongside revenue. Books with no detected AI text showed no such pattern. The authors note this bears on the market-effect question at the center of fair-use arguments over copyright, since one legal test for whether a use is fair asks whether it substitutes for the original work in the marketplace.

A browsing reader couldn’t have worked any of this out on their own. Amazon listings don’t disclose whether a book contains AI-generated text, and the researchers needed full-text detection software run against the entire manuscript, not just a skim of the sample pages Amazon shows before purchase, to sort the catalog this precisely. A cover, a blurb, and a few free pages simply don’t carry enough signal either way. That’s part of why the flooding narrative took hold in the first place: the visible evidence, a handful of obviously synthetic listings, was easy to spot, while the actual sales math required a proprietary sales panel and machine detection run across full manuscripts before anyone could see it clearly.

It’s worth sitting with what this data can and can’t tell you. This is one dataset, covering self-published genre fiction specifically rather than all of Amazon’s catalog, and it’s a preprint that hasn’t yet gone through peer review. The findings describe an association between detected AI text and sales performance across a large sample, not a controlled experiment into why any individual reader bought or skipped a given book. Detectors also aren’t flawless. Even at a reported false-positive rate of 0.04 percent, a sample of this size will misclassify some titles in both directions, and genre-fiction ebooks are not the same market as, say, self-published nonfiction or hardcover releases.

What the numbers do support is a narrower claim than the one usually repeated. AI-generated books aren’t quietly draining the same dollars away from human authors, title for title, the way the flooding story implies. They’re doing something slower and, in some ways, harder to counter.

By adding volume the market didn’t ask for, they’re making the whole shelf more crowded and slightly less profitable to sell into, whether or not the book sitting next to them was touched by a machine at all.

The post A study of nearly 15,000 self-published Amazon books found that titles heavy with AI-generated content now make up a fifth of the catalog, but earn barely one in eight sales dollars. appeared first on The Blog Herald.

What TikTok actually changed

The Creator Rewards Program, which replaced the Creativity Program Beta, requires creators to be 18 or older, have at least 10,000 followers, and have logged 100,000 views in the trailing 30 days. Only videos longer than one minute qualify, a threshold TikTok tied directly to the fact that its own users now spend half their time on the app watching videos over a minute long. Payout is calculated from four factors — originality, play duration, search value and engagement — plus a separate automatic reward tied to ad watch time. None of those four factors reward a creator simply for publishing more. They reward creators for making videos other people finish watching, which is a much narrower category than “videos that exist.”

TikTok’s own numbers show why the company bothered rewriting the formula at all. When it announced the shift out of beta in March 2024, TechCrunch reported that TikTok claimed total creator revenue had risen more than 250% in the prior six months compared with the old, since-shuttered Creator Fund, which had become notorious for paying creators a few dollars on videos with millions of views. That 250% figure describes the total pool, not what any individual creator took home, and TechCrunch noted the same six months saw viewership of over-one-minute videos climb nearly 40% — meaning some of that revenue growth reflects a platform-wide shift toward longer content, and the formula itself may not have gotten proportionally more generous. The distinction matters: growing the pool and narrowing who qualifies to draw from it are two different moves, and TikTok did both at once.

YouTube’s version of the same fix

YouTube made a structurally similar move in August 2026, when it announced on its official blog that new applicants to the Partner Program will need 8,000 qualified watch hours in the trailing year, or 20 million qualified Shorts views in 90 days, starting February 1, 2027 — double the 4,000-hour bar that had stood since 2018. Creators already inside the program keep their existing status; the new bar only applies to whoever is trying to get in after it takes effect. YouTube framed the reason as scale rather than creator misbehavior, saying it needs to “keep pace with the growth of YouTube,” which now sees over 200 billion daily Shorts views and more than a billion hours of watch time on TV every day. And YouTube said the quiet part plainly: it expects “to pay even more to creators in 2027 than we did in 2026.”

The same update adds a second, ongoing gate that only applies to Shorts earnings: creators already in the program must maintain 10 million qualified Shorts views over a trailing 90 days to keep collecting Shorts revenue specifically. Fall below that and a channel doesn’t get removed from the Partner Program outright — its Shorts payments just pause until the number recovers, a mechanic that recurs on a rolling basis — a standing quality check on the video type YouTube considers easiest to flood with low-effort volume, not a one-time entry exam.

The same lever, pulled two different ways

TikTok’s fix operates on quality — it decides which videos, from an already-eligible creator, are worth paying for. YouTube’s fix operates on eligibility — it decides which channels get to enter the payout pool at all. Both levers do the same underlying job: they reduce the number of claims on a finite amount of ad money, so each remaining claim is worth more. YouTube’s own framing makes the mechanism explicit. Raising the bar to 8,000 hours doesn’t grow ad revenue by itself; it narrows who’s eligible to split it, which is exactly how a platform can credibly promise to “pay even more” without needing advertisers to spend more per creator. TikTok’s originality-and-completion formula does the equivalent work one video at a time instead of one channel at a time.

Why this should worry more than video creators

Every content platform monetized by advertising eventually runs into this same math: ad budgets grow linearly, if they grow at all, while the number of people trying to earn from those budgets grows however fast sign-ups allow. Blogs monetized through display ads and affiliate networks sit inside the identical pool logic — RPMs are a function of how many other publishers are chasing the same advertiser dollars in the same niche, not just how good any one blog is. The platforms with the leverage to police entry, the way TikTok and YouTube just did, will keep doing it whenever the math stops working. The platforms without that leverage, including most blogs, don’t get to raise a bar. They just watch their per-visitor revenue erode as more competitors show up to split it.

What happens to the creators on the wrong side of each line

Neither company frames its change as a loss for anyone, and in a narrow sense that’s accurate: nobody currently earning from either program loses access retroactively. But a channel that would have qualified for YouTube monetization under the old 4,000-hour bar and doesn’t clear 8,000 after February 2027 experiences the change as a wall regardless of how it’s described in a blog post, and a TikTok creator whose videos rack up views without holding attention past the first few seconds experiences the new formula as a pay cut regardless of how TikTok frames “originality.” The two platforms picked different filters — one runs on time served, the other on content judged in the moment — but creators on the losing side of either filter are the ones actually paying for the “pay even more” promise made to everyone else.

The post TikTok’s Creator Rewards Program and YouTube’s new watch-hour wall are solving the same problem for two different platforms: too many creators, too little payout to go around appeared first on The Blog Herald.

The person who said so out loud was Chris Best, Substack’s co-founder and CEO. Speaking onstage at The Information’s Future of Influence event in Los Angeles in June 2025, Best said “more than 50 people” were making over $1 million a year on the platform. The comment came out unscripted, in the middle of a conversation, the kind of figure a company usually keeps close until legal and marketing have signed off on the wording. It landed the way unplanned numbers tend to land: fast, and everywhere, within hours.

The moment it slipped out

The clearest sign that this was an unplanned disclosure came from inside Substack itself. Within hours of Best’s comment making the rounds online, Hamish McKenzie, Substack’s other co-founder, posted a note of his own: “Oh shit, didn’t realize we were disclosing this.” McKenzie’s note reads like genuine, in-the-moment surprise at watching a private number go public, posted the same evening Best said it.

That kind of reaction rarely happens by accident. Companies generally control what gets said about their own revenue and their own top earners, especially when the number is this flattering. A co-founder’s half-panicked note is a small, human confirmation that the figure was a real answer to a real question, given by someone who hadn’t quite worked out yet what happens once a number like that gets loose in the world.

What five years actually did

Whatever prompted the disclosure, the underlying shift is real, and it shows up in the platform’s own subscription numbers. Axios reported that Substack’s paid subscriptions grew from about 100,000 in July 2020 to more than 5 million by March 2025, roughly a fiftyfold increase in five years. That’s a jump big enough to turn a promising, unproven place to write into one where a few dozen people can apparently run a media business at seven-figure scale, without a network, a studio, or a masthead behind them.

Five years ago, the honest answer to whether someone could make a living writing a newsletter was maybe, for a handful of already-known writers, if the timing worked out. The honest answer now comes with a number attached to the top of that range, and the number is high enough to change what “maybe” means for everyone still climbing toward it.

Why now, and not five years earlier

Part of the answer is timing that had little to do with Substack itself. Newsrooms spent the years between 2020 and 2025 cutting staff on a scale that pushed a lot of experienced reporters and columnists to look for income that didn’t depend on one employer’s ad budget. A subscription newsletter, paid for directly by a few thousand readers instead of approved by an editor, was one of the few options that let an established writer keep a byline and set their own terms.

The other part is simpler. Five years ago, asking a reader to hand over a monthly card payment for a newsletter was still a hard sell, competing against a habit of expecting most writing online to be free. By 2025, that habit had loosened enough that a reader already paying for three or four other subscriptions barely noticed adding one more. Substack didn’t create that shift by itself, but it built the checkout button that a critical mass of readers were, by then, willing to click.

Fifty people, or fifty small operations

It’s worth being precise about what Best’s number does and doesn’t describe. “More than 50 people” earning over $1 million a year is a statement about income at the top of the platform. It says nothing about how that income gets produced. Some of those newsletters are probably one person handling everything: writing, editing, answering comments, running the business side alone. Others are more likely small operations, with an editor, a researcher, or a couple of contributors splitting the byline and the workload. A seven-figure newsletter run by four people and a seven-figure newsletter run by one person both count toward the same total, and they represent very different amounts of individual effort.

That distinction doesn’t make the figure less real. It changes what the figure is evidence of: a ceiling that’s newly reachable on the platform, not a job description that applies evenly to everyone chasing it.

The number also says nothing about how long any of those fifty-plus newsletters will stay above the line. A newsletter’s income depends on renewals, and renewals depend on readers deciding, year after year, that the writing is still worth the price. Crossing seven figures once, in 2025, is not the same as staying there in 2030.

A number that didn’t exist to report

What makes this moment different from a routine growth statistic is that five years ago, there was no comparable figure available to disclose, flattering or not. A platform can’t reveal that dozens of independent writers cleared $1 million a year when the entire platform hadn’t yet proven a handful of writers could clear six figures. The number Best said onstage only became sayable because enough individual newsletters had already gotten there first, one subscription at a time.

Whether that pace holds for the next five years is an open question, and Substack itself doesn’t seem to have a rehearsed answer, going by how quickly its own co-founder reacted to the number getting loose. What’s confirmed for now is narrower and more specific than any claim about the future of media. A threshold nobody could point to in 2020 has a number attached to it in 2025, and that number came from the person best positioned to know it.

For anyone weighing whether to start a newsletter of their own, that number is worth reading as a ceiling rather than a floor. Most writers on Substack do not clear six figures, let alone seven, and the platform has never claimed otherwise. What changed in five years is that the top of the range moved somewhere it hadn’t been before, far enough that even a skeptical audience had to notice.

The post More than 50 newsletters on Substack are now earning over a million dollars a year, according to the platform’s own co-founder — a milestone that didn’t exist five years ago. appeared first on The Blog Herald.

What five years of the same survey actually shows

Gargano’s Blogging Income Survey 2026 collected anonymous Typeform responses from 129 bloggers between January 2 and April 29, 2026, drawn largely from her own email list, courses and paid community, and skewing toward bloggers already focused on monetization. Most respondents were based in the UK, US, Canada, Australia and New Zealand. The headline numbers are blunt. It takes an average of 21 months for a blogger to earn their first dollar, though the median is 12 months and just over a quarter of respondents earned something within six months of starting. Reaching what a respondent personally considered a full-time income took an average of 4 years and 1 month, with a median of 4 years — up from a 3-year median last year — and 24% of this year’s bloggers hit that mark within two years, down from 28% last year. Gargano’s own read of the data, after five years of running the same survey, is direct: “it’s clearly getting harder to ‘break in’ to blogging, and it’s taking longer to get to a decent full-time income.”

The other half never earns a cent

That four-year runway assumes the blog eventually turns a profit at all, and a lot of them don’t. In February 2019, Kyle Byers at GrowthBadger surveyed 1,117 self-identified bloggers, split them into bloggers earning over $50,000 a year and everyone else, and found that a little less than half of the lower-income group made any money from their blog at all. Of the roughly 47% who did, the two most common methods were Google AdSense and affiliate marketing, the lowest-effort and lowest-control monetization options on the list. The gap between the two groups traced back to specific, repeatable choices: bloggers earning over $50,000 were 4.3 times as likely to do keyword research, used more than three times as many email list-building tactics, and were 5.6 times as likely to sell a product or service they controlled themselves.

“Full-time” and “financially sustainable” aren’t the same claim

Even a self-reported full-time income doesn’t settle whether the work actually supports someone long-term. A report from the Center for News, Technology & Innovation, published in April 2026, surveyed 43 independent writers and creators and conducted in-depth interviews with 26 of them. Only 5 of the 43 said their content income could fully fund their lifestyle. Just over half, 23 of 43, said they couldn’t fund their lifestyle from the work at all. The report covers independent writers and creators broadly rather than bloggers specifically, but the shape of its finding matches what both blogging surveys show separately: a small top tier making a living, and a much larger group for whom the work is a side project, a portfolio piece, or an experiment that hasn’t paid off yet.

What actually separates the two groups

Byers’ data gets more useful once “over $50,000” stops being treated as a single trait and starts getting read as a bundle of specific decisions. Higher-income bloggers in his survey rated Google organic search 83% more valuable than lower-income bloggers did, and rated email marketing 192% more valuable. Their most popular posts averaged 2,424 words, 83% longer than the top posts from lower-income bloggers, and they paid outside writers $347 per article against $96. The survey measures correlation.

Byers is upfront that adopting these habits out of order or too early can backfire. But it describes a business run with the rigor of any other small business — a different animal from a hobby that occasionally throws off ad revenue.

Gargano’s 2026 data points the same direction from a different angle: bloggers who did email marketing earned roughly twice as much as those who didn’t, and bloggers using four or more revenue streams out-earned single-stream bloggers by a wide margin.

Why the timeline keeps stretching

Neither survey pins down exactly why the under-five-year cohort lost ground this year while the ten-year-plus cohort gained it, but the direction fits what’s been happening to organic discovery generally: a newer site has less accumulated authority, fewer backlinks and a thinner archive to fall back on right as search traffic gets harder for everyone to earn. A blog compounding content, links and an email list since 2016 has a cushion a blog started in 2024 hasn’t had time to build. That lines up with Gargano’s own explanation for the split: bloggers past the five-year mark, she notes, “tend to have more revenue streams than bloggers who have been blogging for less than 5 years and they tend to be monetizing via more lucrative revenue streams.”

Blogging isn’t a bad bet. It’s a slower, less evenly distributed one than the “start a blog, quit your job” pitch suggests. The bloggers clearing $50,000 a year are running a small media business, complete with keyword research, an email list and diversified income streams, and they got there by staying past the point where most people quit, well beyond the four-year mark Gargano’s data says is typical now.

The post Is blogging still profitable? The bloggers who reach a full-time income wait an average of four years to get there, and the wait is getting longer, not shorter appeared first on The Blog Herald.

What the zero-click number actually measures

SparkToro’s analysis covers desktop and mobile web searches, weighted two-to-one mobile-to-desktop, but excludes Google’s mobile app, where zero-click behavior is reportedly even more aggressive. The rise itself isn’t new: zero-click search has been climbing steadily since at least 2016, driven first by featured snippets, knowledge panels, and direct-answer boxes, long before generative AI entered a results page. What’s new is the slope of the climb.

What AI Overviews specifically add

AI Overviews now appear on more than 20% of all Google searches. Ahrefs analyzed 300,000 keywords, split evenly between searches with an AI Overview present and similar informational searches without one, and found that as of December 2025, the presence of an AI Overview correlated with a 58% lower click-through rate for the top-ranking page than a comparable keyword without one — up from the 34.5% reduction Ahrefs measured when it first ran the analysis in 2025. Ahrefs’ Ryan Law was blunt about the gap between that finding and Google’s own public claims: “It seems that Google doesn’t want us to see the clickthrough rate for AI Overviews.” Pew Research found a similar pattern independently: searches that triggered an AI summary produced a click only 8% of the time, versus 15% for searches without one.

The metric change bloggers are already seeing in Search Console

Anyone checking Google Search Console for their own blog has probably noticed a pattern that used to be unusual: impressions climbing while clicks fall, instead of moving together the way they always used to. Ahrefs calls this “the great decoupling” and traces it to two mechanical causes. AI Overviews often cite several source URLs at once for a single query, spreading impressions across multiple sites instead of concentrating them on whichever result ranks first — and when an AI Overview answers the question directly, a person has no reason to click through to any of the cited sources at all. Ahrefs found the correlation between impressions and clicks across its own tracked data flipped from a positive 0.425 in late 2024 to a negative 0.352 by early 2025, around the same time Google’s March 2025 core update roughly doubled how often an AI Overview appeared. A blog’s impressions going up is no longer, on its own, good news.

Who’s actually losing traffic, and who isn’t

The decline isn’t evenly distributed. Digital Content Next tracked Google search referrals for 19 major publishers and found a median year-over-year decline of 10% in May and June 2025 — 14% among non-news publishers, a smaller 7% among news publishers. Branded searches move the opposite direction: according to research from marketing agency Amsive, reported by Search Engine Journal, AI Overviews are associated with an 18% increase in click-through rate when someone searches for a specific site or brand by name rather than a general topic. That split matters more than the headline zero-click number: a blog that ranks for “how to grow an email list” is exposed to the full brunt of the AI Overview effect, while a blog whose readers search by name is largely insulated from it.

What publishers are actually doing instead of waiting it out

The publishers already living with this aren’t treating it as a search-optimization problem to be patched. Reach’s David Higgerson described the shift bluntly in comments to the BBC, reported by Search Engine Journal: “We’ve got millions of people who receive our alerts on WhatsApp. We’ve built newsletters.” Bauer Media’s Stuart Forrest made a related but distinct point about what still works inside search itself, framing it as a content-quality problem rather than a technical one: publishers need to be the site an AI Overview cites instead of a rival, and “it’s amazing the number of publishers that just give up on” doing that well. Both point toward the same underlying shift: an audience a publisher owns directly — a subscriber list, an app, a community — doesn’t lose value when Google changes how a results page looks, because Google was never what was holding that relationship together.

What this means for a blog that lives on organic search

A blogger whose traffic comes almost entirely from Google is exposed to a trend that started before generative AI and has now picked up speed because of it, and neither Google’s own explanations nor a single SEO fix are likely to reverse it. The workable response looks like what larger publishers are already doing at a bigger scale: building a channel Google doesn’t control, whether that’s a newsletter, a podcast feed, or a community, so a chunk of the audience arrives without going through a results page at all. Building that channel won’t recover lost search traffic. It puts a ceiling on how much of the business any single results-page redesign can threaten.

The post Zero-click Google searches passed two-thirds in early 2026, and the newest data shows AI Overviews didn’t start that trend — they just sped it up appeared first on The Blog Herald.

What Beehiiv actually shipped

The centerpiece is Community: a discussion space attached to a creator’s newsletter where subscribers can talk to each other, which creators can moderate and gate behind a paid tier. CEO Tyler Denk’s explanation for building it was specific: “People following your content have a shared interest in what you’re creating, but they can’t communicate with each other… being able to have a community where your audience can actually engage with one another is super valuable.” Alongside it, an AI Copilot analyzes a newsletter’s content and audience metrics to suggest growth moves and draft outreach campaigns, and the editor got a split-screen redesign showing edits and the rendered preview at once. Beehiiv said its programmatic ad network now pays publishers on the platform more than $1 million a month in aggregate.

A creator can export a Beehiiv subscriber list in an afternoon; the conversation history inside a paid community tier stays on the platform. That’s the mechanism behind July’s launch — Community’s real function is making more of a creator’s audience relationship non-portable.

Why a newsletter tool is building a chat room

A newsletter, structurally, is a list — the entire relationship lives in an email address a creator could export and take to a competitor tomorrow. That portability is the whole pitch newsletter platforms make to attract creators away from a rival, and it’s also the thing that keeps any single platform from ever fully owning a creator’s audience. A community feature works against that portability: conversations between subscribers, unlike the email list itself, don’t export. A creator can leave Beehiiv with their subscriber list intact; they can’t take the conversation history, the paid-tier chat access, or the relationships subscribers have built with each other. For a platform competing on retention, that’s close to the entire point of building Community in the first place.

The pitch is ownership, and at least one creator’s numbers back it up

Denk has framed Beehiiv’s difference from Substack and Patreon as being “unopinionated” about how a creator monetizes, rather than steering them toward one model. That pitch has a concrete data point behind it: journalist Frankie de la Cretaz reported a 78.4% increase in net income worth $21,325 a year after migrating a newsletter from Substack to Beehiiv, citing both financial and what they described as ethical reasons for the switch. That move isn’t an isolated case: Nieman Lab has tracked several other writers who left Substack for Ghost and Beehiiv and reported receiving a larger share of subscription revenue after the move. The company is betting that pattern continues — Beehiiv added 50,000 active users and posted its best quarter yet, $4.5 million in new annual recurring revenue, in the first quarter of 2026 alone.

What this means for anyone building an audience on borrowed infrastructure

Bloggers and newsletter writers building on any platform — Beehiiv, Substack, WordPress.com, YouTube — are making the same trade Beehiiv’s own creators are: portable reach in exchange for platform-specific engagement tools that don’t travel if they leave. The $1-million-a-month ad-network figure is early evidence that programmatic advertising can pay real money at scale without a creator selling a single sponsorship themselves, but that revenue is also the newest and least portable part of the arrangement — a sponsor relationship built through a platform’s ad network doesn’t move with an exported subscriber list the way the list itself does.

The useful question isn’t whether to use these tools; it’s tracking which parts of an audience are actually a creator’s to keep and which parts only exist because one specific platform is hosting them.

The retention math behind the community push

Beehiiv has spent the past year adding podcasts, webinars, and paywall options on top of newsletters — the podcasts feature hit its quarterly goal within 24 hours of launch and beat its Q2 target by more than 10x, according to Denk, and half of new podcast hosts on the platform migrated in from elsewhere. That kind of expansion only makes sense if the company expects creators to consolidate more of their business inside one platform rather than spreading it across specialized tools. Community is the piece of that stack least like the others: podcasts and webinars are content formats a creator could rebuild on a different platform, but a community’s value is the specific conversation history that formed inside it — the one part that doesn’t come along when a creator leaves.

The post Beehiiv just gave newsletter subscribers a way to talk to each other, and it’s a hedge against the one thing that makes newsletter platforms replaceable appeared first on The Blog Herald.

Min’s explanation was blunt: “Substack is an incredible launchpad, but we reached a point where it wasn’t built for what we were becoming.”

That is a specific, checkable complaint, and it is not the same complaint as “the rules keep changing under me.”

What actually left, and what didn’t

The Ankler’s departure wasn’t a clean break. The outlet said it would keep a weekly newsletter and live video presence on Substack even after moving its primary infrastructure to Passport — a hedge, not a divorce. That distinction matters, because the same caution applies to the other names attached to this story.

The Bulwark, Mehdi Hasan’s Zeteo, and Emily Sundberg’s Feed Me have been reported as privately frustrated with Substack and quietly exploring alternatives, but none has announced a move.

Treating all four as already gone overstates what’s actually happened; only one has.

Why the math changes as a newsletter becomes a media company

Substack takes a 10% cut of subscription revenue. On a solo newsletter earning $50,000 a year, that’s a rounding error. On an operation earning eight figures with a payroll to cover, it becomes real money fast: media reporter Oliver Darcy, writing in his Status newsletter and relayed by The Wrap, estimated the cut runs to “hundreds of thousands of dollars, if not north of a million each year” for top-tier outlets. Ten percent of The Ankler’s roughly $10 million in revenue lands right in that range. What’s driving the frustration is that a percentage that made sense when the business was one person and an email list stops making sense once the business has 18 salaries to fund.

What Substack is actually offering in return

Substack’s counter is that the fee buys distribution as well as infrastructure: the company says its discovery and recommendation tools drive roughly 30% of new paid subscriptions for creators on the platform. That’s a real number worth taking at face value. The publishers pushing back aren’t disputing it so much as arguing they’ve outgrown its value relative to its cost — they want customization, product control, and a design system that doesn’t make a seven-figure media brand feel like one of, as reporting has put it, the platform’s “sub-brands within its own ecosystem.”

The difference between this and a rules problem

It’s tempting to file this next to YouTube’s monetization-bar increase as another case of platform-dependent creators getting burned by rules changing underneath them. The reporting doesn’t support that comparison. Nothing here describes Substack changing its fee, its recommendation algorithm, or its terms unannounced. What’s described is a handful of publishers outgrowing a fee structure and a design ceiling that were built for a smaller kind of business than the ones they’ve become. That distinction — outgrowing a fee structure versus getting blindsided by one — is what determines whether your own platform risk looks like The Ankler’s or looks like YouTube’s.

Who this actually applies to

For the overwhelming majority of people writing on Substack, this fee fight changes nothing. A newsletter earning $50,000 or $200,000 a year is nowhere near the point where a 10% fee competes with payroll, and the platform’s discovery tools are still doing exactly the job they were built for at that scale. This is a top-of-the-market story about four publications that grew past being newsletters, not a warning sign for the thousands of writers who haven’t.

Ten percent of $10 million is roughly $1 million a year, the size of the line item that turned a platform fee from a rounding error into a decision. The Ankler is the only one of the four so far to act on it.

What this means for anyone building past a one-person operation

The lesson generalizes past media trades and Hollywood newsletters. Any blogger, YouTuber, or newsletter writer who starts as a one-person operation eventually faces the same fork The Ankler did: the platform fee, ad-network cut, or marketplace commission that was invisible at $50,000 a year becomes a line item worth renegotiating once there’s a staff to pay. The math here isn’t really about fairness — a 10% subscription cut, a 30% app-store fee, and a marketplace’s take rate are all defensible prices for the discovery and infrastructure they provide at small scale. What changes is leverage: a platform’s audience-finding tools matter less to someone who already has 150,000 people showing up directly, and that person can afford to pay more to own the relationship outright instead. The useful test isn’t “is this fee too high,” it’s “would I still need this platform’s discovery tools if my audience doubled” — and for most writers, the honest answer is yes, for a long time yet.

The post The Ankler didn’t leave Substack for a bigger audience – it left because the 10% cut that works at newsletter scale stops working at media-company scale appeared first on The Blog Herald.

The speed of that interim ruling matters for platforms trying to plan around a new law. The distinction between a law’s effective date and a temporary bar on enforcement matters just as much.

The judge concluded that NetChoice was likely to succeed on its First Amendment challenge. Her reasoning offers other states a detailed account of the constitutional concerns a similar proposal could face, subject to further litigation and appellate review.

That turns this from a Virginia story into a preview of the fight headed for every statehouse still working on a similar bill, and into a real complication for any publisher or platform whose audience skews young enough that these laws are the reason they were paying attention to Virginia in the first place.

1. A judge temporarily blocked enforcement

Virginia’s Senate Bill 854 required a one-hour daily default limit per service or app for users under 16, age determination, and a process for verifiable parental consent to increase or decrease that limit. It was not a cap that parents were unable to adjust.

The law took effect January 1, 2026. NetChoice sued before that deadline. On February 27, U.S. District Judge Patricia Tolliver Giles granted a preliminary injunction against enforcement. Her decision applied the standards for interim relief, including likelihood of success, rather than entering a final judgment on the merits.

2. The reasoning was about speech rights, not really about whether social media is harmful

Judge Giles recognized Virginia’s interest in protecting young people while finding that NetChoice was likely to establish a First Amendment violation. Protecting minors and restricting access to lawful speech raise separate questions, and the opinion addresses both.

The written opinion examines whether the law is sufficiently tailored, including the speech burdens of determining users’ ages and exemptions for other services. Those concerns supported preliminary relief. They are reasons to scrutinize a proposal’s design, not a guarantee that every similar law will fail or that a small rewrite could never survive.

3. Virginia’s law was blocked well after it already took effect

This part gets flattened in quick coverage. The statute’s January 1 effective date preceded the February 27 injunction. An injunction restrains enforcement; it does not erase the statute from the books.

For a platform that had prepared a Virginia compliance flow, that creates a planning problem rather than a reason to discard the work. Requirements can return if an injunction is changed or reversed, and a provider may have obligations under other laws even while this one is blocked.

4. The sponsor is not backing down, and neither is NetChoice

State Senator Schuyler VanValkenburg, who sponsored the bill, made clear this is not over. “Obviously, I disagree with the decision,” he said. “I think our legislation does thread the needle of respecting free speech while also empowering parents and protecting kids. I look forward to the case working its way through the legal system and I think, and hope, that ultimately the courts will agree with my analysis and that of the commonwealth.”

On the other side, NetChoice, the trade group whose members include most of the largest social platforms, framed the ruling through its litigation center co-director Paul Taske as validation of a broader pattern the group has pushed in court after court, in Virginia and well beyond it: “Today’s decision underscores a core truth: unconstitutional laws do not help anyone. Moreover, laws requiring age verification and other privacy-invasive measures actually make everyone less safe and more prone to data breaches.” Both sides are treating this as round one, not the final word.

5. Other states drafting similar caps just got a preview of their own lawsuit

NetChoice is not new to First Amendment challenges involving age determination and limits on access. For publishers relying on younger readers, the useful question is which restrictions courts consider sufficiently tailored. Virginia already allowed parents to adjust the one-hour default; New York’s forthcoming rules address addictive feeds and overnight notifications instead. Parental choice alone therefore does not distinguish the two designs or settle their legal prospects.

6. Keep the compliance work and follow the case

Any publisher or marketer who prepared for the Virginia limit should distinguish the February 27 enforcement pause from permanent relief. The law was enjoined, not repealed. Whether a particular compliance task can be paused depends on the orders that apply and the provider’s other obligations.

Virginia appealed the preliminary injunction on March 3, 2026. Keeping the compliance notes available makes more sense than assuming the issue is closed. I have gone through more than one platform policy change that looked permanent and turned out to be a first draft, so my instinct here is to keep those notes filed away rather than delete them.

None of this settles whether an hour a day is the right default for a teenager, or which legal design will ultimately survive. It means the February ruling temporarily blocked enforcement while the constitutional challenge continued, rather than finally striking down Virginia’s law.

Correction, October 4, 2026: An earlier version described the preliminary injunction as a final decision striking down the law and incorrectly suggested there was no parental override. Enforcement was temporarily blocked, and the statute allowed parents to adjust the one-hour default.

The post A federal judge temporarily blocked enforcement of Virginia’s one-hour default limit on social media use by under-16s, a preliminary First Amendment ruling that could influence similar bills without finally settling the law’s constitutionality appeared first on The Blog Herald.

What YouTube actually said, and didn’t say

YouTube’s own explanation for doubling the bar was about scale: the company pointed to more than 200 billion daily Shorts views and over a billion daily hours of watch time on connected TVs as evidence the platform had outgrown its decade-old thresholds. It did not frame the change as a response to spam, low-effort content, or advertiser complaints about where their money was landing, and that connection is this article’s inference — YouTube hasn’t offered it as a reason.

What YouTube confirmed in its own support documentation is narrower than the headline suggests: the doubled bar, 1,000 subscribers plus either 8,000 watch hours in a year or 20 million Shorts views in 90 days, applies to creators applying to the Partner Program on or after February 1, 2027. Current members’ status is not affected by the change.

The separate, ongoing requirement to keep an already-monetized channel active stays at 1,000 watch hours a year, or 1 million Shorts views in 90 days, or a minimum upload cadence, with a 90-day cure window if a channel dips below it — and every creator has until January 31, 2027 to accept updated terms or lose monetization features the next day.

What podcast advertisers are actually paying for

The IAB’s $2.862 billion figure for 2025 is real growth, but it’s decelerating growth: podcast ad revenue grew 72% in 2021 and 26% in 2022, so 17.6% is a much calmer number than the headline “podcast advertising keeps climbing” implies on its own. What hasn’t decelerated is trust among the format’s most engaged listeners. Sounds Profitable found that 86% of podcasting’s most active users recalled hearing an ad in the past week, and that 46% of respondents trust ad-supported podcasts’ moderation and content quality, a figure that only Twitch and streaming music, at 51%, beat among the channels studied. Sounds Profitable partner Tom Webster has made a related point: podcasting’s growth hasn’t come by bloating ad loads past what listeners will tolerate, which is a large part of why the ads that do run get noticed instead of ignored.

Why a doubled bar and a trust premium are the same story

Two hundred billion daily Shorts views describes how much inventory exists — it says nothing about how much of it an advertiser should want their name next to. Podcasting has spent years solving that problem from the demand side: a handful of host-read ad slots per episode, sold by a comparatively small number of shows, produces scarcity by construction, and scarcity is a large part of why podcast ads earn a recall and trust premium that programmatic video hasn’t matched. Doubling YouTube’s monetization bar does the same job from the supply side: it shrinks the pool of channels whose inventory YouTube is willing to sell against, whatever the company’s stated reason for doing so. Both moves push the same direction: toward less inventory, sold at a higher price, to buyers who trust what they’re getting.

The part that doesn’t fit neatly

The picture is messier than a single cause with two effects. Podcast ad dollars are concentrating more heavily toward the largest shows this year: Magellan AI’s ad-spend tracking, reported by Inside Audio Marketing, found the top 500 podcasts captured 52% of Q2 2026 ad spend, up from 48% in Q1 — the opposite of a story where trust and ad dollars spread evenly across the format. And YouTube’s stated reason for its change is platform scale — the company hasn’t described it as curation — so treating the two announcements as coordinated responses to the same market pressure is this article’s read of the pattern; neither company has made that claim about the other.

What this means for anyone selling ad space on their own site or newsletter

$2.862 billion and a doubled monetization bar were announced four months apart, by two companies, for two different stated reasons.

Whatever each company says about its own motives, both numbers move the same lever: less inventory advertisers can verify, sold at a higher price, to buyers who trust what they’re getting.

A blogger or newsletter writer with a smaller, high-completion audience is sitting on the exact asset advertisers are chasing in both of the trends above: attention a sponsor can actually verify shows up and pays attention, rather than attention that merely exists in a subscriber count. A newsletter with a 45% open rate and 3,000 subscribers is a better pitch to most sponsors than one with 30,000 subscribers and an 8% open rate, for precisely the reason a 46% trust score beats a bigger, noisier audience with a worse one: the number that closes a sponsorship deal is verified attention, not raw reach.

The post Podcast advertising rates are climbing and YouTube just doubled its monetization bar – here’s why that might not be a coincidence appeared first on The Blog Herald.

That is not a comfortable thing to admit as someone who writes for a living. It turns out I am not alone in not knowing, and a lawsuit filed this summer is trying to force an answer for a much bigger group of publishers than just me.

What the lawsuit actually claims

On June 24, 2026, a coalition of 34 newspaper companies, representing nearly 400 local publications across 33 states, filed suit against OpenAI and Microsoft in federal court in New York. The Arkansas Democrat-Gazette and its parent company, WEHCO, are named plaintiffs. The complaint alleges that the two companies systematically scraped publisher websites, copied articles without permission, stripped out copyright management information, and fed the material into the datasets that trained ChatGPT and Copilot. The legal claims rest on the federal Copyright Act and the Digital Millennium Copyright Act, and the newspapers are seeking statutory damages along with an injunction that would force the removal of their work from the underlying models and training data, not just a promise to stop scraping going forward.

Part of the newspapers’ evidence is OpenAI’s own words. In written evidence the company submitted to the UK House of Lords in late 2023, OpenAI stated plainly that “it would be impossible to train today’s leading AI models without using copyrighted materials.” The complaint frames that admission as the whole case in miniature: a company that argues copyright is essential to its own product while treating other people’s copyrighted work as free raw material. As the filing itself puts it, the publishers are trying “to hold Defendants to the same standard they insist upon for themselves.”

Why this matters even if your blog is not one of the 400

Most independent bloggers and small digital publishers are never going to be plaintiffs in a case like this one. Joining a coalition lawsuit takes lawyers, resources, and a body of published work large enough to be worth the fight. But the legal theory being tested here does not only apply to newspapers with decades of archives. If a court agrees that scraping a publisher’s site to train a commercial model without permission or payment is copyright infringement, that finding does not stay contained to the plaintiffs named in this one complaint. It becomes a data point every smaller publisher can point to.

A quieter reason this case matters sits beyond whatever legal precedent it eventually sets. Most independent writers have never had the option WEHCO and its co-plaintiffs are exercising, the option to say no and be taken seriously. A newspaper chain with decades of archives and in-house counsel can credibly threaten litigation. A single blogger, or even a small network of them, generally cannot, which means the practical leverage over how AI companies treat smaller publishers’ work has always sat with whoever could afford the lawsuit, not with whoever actually wrote the content.

A quick check on where you actually stand

1. Look at your own robots.txt file

Most AI crawlers, including OpenAI’s, are supposed to respect a robots.txt file that blocks them, at least when a site owner has actually configured one to do so. Check whether your site’s file currently blocks known AI crawler user agents, or whether it was never updated past whatever your CMS shipped with by default. A lot of sites are wide open simply because nobody ever touched the setting.

2. Search for your own content inside publicly available training-data trackers

Several independent researchers and outlets have built lookup tools that let a site owner check whether their domain appears in commonly used AI training datasets, including the kind of web-scrape datasets named in this lawsuit’s complaint. Running your own domain through one of these is a five-minute task that at least tells you whether you are asking a hypothetical question or a concrete one.

3. Decide whether blocking or licensing is actually the goal

Blocking every AI crawler outright is not automatically the right call for every publisher. Some are actively negotiating licensing deals instead, trading access for payment rather than trying to keep every crawler out. Figure out which outcome you actually want before you spend time implementing either one. A blogger who depends on search traffic to survive has a very different calculation than one whose income comes mostly from a newsletter or direct sales, since blocking a crawler can carry search-visibility tradeoffs of its own that have nothing to do with AI training specifically.

4. Keep records now, even if you take no other action

Whatever you decide to do about crawling going forward, save dated screenshots or archived copies of your own published work and your site’s historical robots.txt settings. If this legal theory keeps winning, a record of what you published and when, and what you did or did not authorize, is the kind of thing that turns a vague sense of being wronged into something a lawyer could actually work with later.

Final thoughts

I am not a lawyer, and nothing above is a legal strategy, just the plain, practical version of what I would want to know about my own site before deciding whether any of this is worth acting on.

Whatever happens to this particular case, the underlying question, whether the internet’s smaller publishers get paid or just get scraped, is not going away because the newspapers with the resources to sue are the only ones asking it out loud. I would rather know the answer for my own site now than find out the hard way later that I never bothered to look.

The post Arkansas’s own Democrat-Gazette is among nearly 400 local newspapers now suing OpenAI and Microsoft over claims their articles were scraped without permission to train ChatGPT and Copilot, and the case could set the terms for whether smaller blogs and digital publishers ever get compensated when their content trains someone else’s AI. appeared first on The Blog Herald.

On paper, this reads like a studio discovering a promising young filmmaker. What it actually priced was something else: an audience that had already watched the thing once, said so out loud on three different platforms, and would plausibly show up again without a single dollar spent finding them.

What these deals are actually buying

Call it a spec deal, an option, or an overall deal — the label changes, the object being purchased mostly doesn’t. Linden Lane Films, a new venture pairing traditional studio talent with creator partnerships, signed twin creators Alex and Alan Stokes, who bring 137 million YouTube subscribers between their channels, plus fellow creator Ben Azelart’s 48.6 million. Alan Stokes described the appeal as access to people who could “elevate our game and produce content that can garner new audiences worldwide” — audience-first language from a creator, not a studio, which tells you where the leverage actually sits in these conversations. The screenplay, the pitch deck, the three-minute short: all of it is packaging around the one number that gets a deal signed, which is how many people are already subscribed to finding out what happens next.

Why marketing costs make the arithmetic work

Independent film researcher Stephen Follows has tracked what movies actually spend to find an audience once they’re made, and the numbers explain why a pre-built one is worth paying for. “Hundreds of Beavers” spent roughly $135,000 promoting a $150,000 film — a 90% release-to-production ratio. “You, Me & Her” spent $350,000 marketing a $500,000 production. “Columbus” spent $189,032 on P&A, more than a quarter of it on publicists alone. A24’s “Backrooms,” adapted from Kane Parsons’ viral analog-horror YouTube series, cost under $10 million to make and carried a domestic marketing spend in the teen millions — before the film went on to gross more than $390 million worldwide. A creator who arrives with 15 million people already watching has, in effect, pre-paid a bill that regularly runs as high as the production budget itself.

The ceiling nobody’s pricing in yet

This doesn’t scale the way the trend pieces imply. Digiday’s reporting on the creator-deal wave has found that only the top tier of creators keep any ownership stake in what gets made from their work; most, including creators with audiences in the tens of millions, sign into partnerships where the studio holds the rights. The genres that travel are narrow — horror, comedy, true crime, unscripted — and one industry observer put the limit bluntly: a creator known for chick-flick horror content isn’t the person a studio calls when it wants prestige drama. There’s also a saturation problem building underneath the enthusiasm: multiple deals are getting greenlit on audience size alone, with little vetting of whether the underlying material can sustain a ninety-minute runtime it was never built for.

The number worth trusting is a subscriber count that shows up across unrelated platforms for the same piece of content, because that means the same story pulled attention through three different discovery algorithms and three different audiences. Fifteen million views split across YouTube, TikTok, and Instagram for one three-minute short is a harder thing to fake or inflate than a single platform’s follower total. A studio evaluating a creator deal is, whether it says so or not, running the same due diligence a savvy marketer runs on influencer partnerships: not “how many followers,” but “did this specific thing actually travel.”

What this means for anyone building an audience before they have something to sell

Bloggers and newsletter writers chasing a book deal, a sponsorship, or a syndication pickup are being evaluated on the exact same math, even when nobody involved calls it that. The pitch a studio is paying six figures for is not “this is well-written” — Hollywood has an entire industry of well-written unproduced screenplays sitting in drawers to prove that craft alone rarely closes a deal. The pitch that closes is “these specific people already showed up, on their own, more than once, across more than one place I don’t control.” That is a number a blogger can build years before there’s a book to sell, a course to launch, or a brand willing to sponsor a newsletter — and unlike a screenplay, it doesn’t require anyone’s permission to start compounding.

Fifteen million cross-platform views bought a three-minute short a six-figure development deal with its original writer still attached. A marketing budget that regularly matches or exceeds a film’s production cost is the reason that math works: the audience is the expensive part, and a creator who already has one is selling a studio the one line item it can’t easily discount.

The post Why studios chasing spec deals with digital creators are really just buying an audience-acquisition cost they can no longer justify building from scratch appeared first on The Blog Herald.

Most of the coverage since then has framed this as a story about platforms, TikTok and Instagram and whichever app comes next, being forced to rewrite how their algorithms treat minors. That framing captures part of the real story, though it leaves out the people who publish into those feeds, not just the companies that run them.

The assumption most publishers are making

The quiet assumption going around blogging and small-media circles is something like this: this is a platform-level compliance problem, so it belongs to Meta and TikTok’s legal teams, not to a publisher who just posts content and hopes it gets seen.

Under that assumption, a blog with a young readership can treat the SAFE for Kids Act as background noise, a story to read about, not a policy to plan around.

Why that assumption does not hold up

The problem with treating this as someone else’s compliance issue is that it ignores how teen readers actually find content in the first place.

A recent Pew Research Center report on young adults and news found that 70 percent say they get political news because they happen to come across it while doing something else, against just 30 percent who go looking for it on purpose. That report covers young adults broadly rather than minors specifically, but it points at the same mechanism the SAFE for Kids Act is built to interrupt: a huge share of younger readers do not search for what they read, a feed hands it to them.

Once a platform defaults an under-18 account into a plain chronological feed instead of a personalized, algorithmically ranked one, the recommendation engine that used to occasionally surface a publisher’s post to a teenager who had never heard of that blog before stops doing that job by default. A publisher’s own website looks exactly the same as it did yesterday. The part that quietly stops pulling its weight is a channel a lot of publishers never had to think about, because it used to work in the background without anyone touching it.

What the rules actually require

From January 25, 2027, the final rules will restrict addictive algorithmic feeds for covered under-18 users unless a parent consents. They will also prohibit notifications between midnight and 6 a.m. without parental consent. Those restrictions are forthcoming, not already in operation under these rules.

The Attorney General’s announcement describes age-assurance requirements beyond a simple birthdate field. Platforms must offer at least one method other than government ID, test accuracy annually, and retain the testing results for at least ten years. Data collected for age assurance must be deleted once it has served its purpose. The testing-results retention requirement should not be confused with keeping individual users’ underlying verification data for ten years.

The comparison with Virginia needs care. A federal judge temporarily blocked enforcement of its usage-limit law in February. But Virginia’s one-hour default also allowed parents to increase or decrease it. The distinction is between time limits and restrictions on particular feed and notification features, not between having parental choice and having none. The Virginia injunction does not establish whether New York’s different design will survive a challenge.

“Our kids’ mental health isn’t for sale,” Governor Hochul said when the final rules were announced. “We promised action and the SAFE for Kids Act is delivering by ensuring that any platform that targets our children with addictive feeds will face legal consequences.” Attorney General James, whose office built the age-verification framework, put the practical side of it more plainly: “These rules set standards for age verification and parental control to ensure tech companies are doing their part to keep kids safe on their platforms.”

What publishers can actually do about it

A traffic source that was never guaranteed to begin with is not much of a reason to panic. It is, though, a good prompt to stop treating algorithmic discovery as the whole strategy.

A newsletter a teen reader already subscribed to still lands in their inbox at 7 a.m. regardless of what their feed looks like. Content built to actually answer a specific search question still gets found by someone typing that question in, whether they are 15 or 50. And a follow button, the kind that surfaces a publisher’s posts in chronological order rather than an algorithm’s, becomes more valuable than it has been in years, precisely because chronological is about to become the default for an entire age group.

I am not going to pretend to have strong feelings about whether this law is good policy for kids specifically. That is a hard question and a personal one for a lot of families, including mine, and it is not one a single state’s rulebook is going to settle for anyone. For a publisher relying on algorithmic discovery, January 25, 2027 is the date to plan around, rather than treating the restrictions as a change that has already happened.

Correction, October 4, 2026: The earlier headline described future restrictions as already effective and incorrectly contrasted New York with a Virginia limit having no parental override. The effective date is January 25, 2027, and Virginia also allowed parental adjustment.

The post New York’s SAFE for Kids rules will restrict addictive algorithmic feeds and overnight notifications for under-18s without parental consent from January 25, 2027, giving publishers a deadline to prepare for changes in teen discovery appeared first on The Blog Herald.

Since June 9, 2026, that question has a legal answer in New York, and it is not a gray area. If you made the ad and the person in it is not a real human, you tell people. If you did not make it, and you are just running it on your site, you generally do not have to.

What the law actually requires

New York’s new statute grew out of Senate Bill S8420A, sponsored by State Senator Michael Gianaris and signed into law on December 11, 2025. It took effect 180 days later, on June 9, 2026. The operative line requires that an advertiser “shall conspicuously disclose in such advertisement that a synthetic performer is in such advertisement, where such person has actual knowledge,” according to the bill text itself.

“Conspicuously” is doing real work in that sentence. Per ArentFox Schiff’s reading of the law, a disclosure has to be “reasonably visible and comprehensible, not buried in fine print or flashed briefly on the screen.” A label sitting in a wall of terms and conditions, or on screen for a quarter of a second in a video ad, would not clear that bar.

What counts as a “synthetic performer,” exactly

The bill text defines a synthetic performer as a digitally created asset, built with generative artificial intelligence or a software algorithm, made to look like it is a human performer in an audiovisual or visual ad, while not being recognizable as any actual identifiable person. A fully AI-generated brand mascot or spokesmodel fits that definition cleanly. A real actor whose voice was digitally cleaned up, or whose face is simply filtered, is a murkier case, and one that a lot of small marketing teams have not thought through yet.

Per ArentFox Schiff’s breakdown of the penalties, they are civil rather than criminal, and they scale with repetition: $1,000 for a first violation, $5,000 for every one after that. That is not a number that bankrupts a large agency, but for a blogger running their own sponsored campaigns, or a small brand producing its own ads, it adds up fast if the same undisclosed AI spokesperson runs across a dozen posts before anyone notices.

Notice the phrase “actual knowledge” sitting inside that statutory language quoted above. That is a meaningfully lower bar for liability than it first sounds. An advertiser cannot claim ignorance of a synthetic performer it commissioned and then handed off to a production shop without asking questions. If the ad was built to look human and you never checked whether it actually was, a regulator can reasonably ask what you should have known, not just what you happened to know.

Why publishers are treated differently from advertisers

This is the part of the law that actually matters most for anyone running a blog. The statute exempts media outlets and publishers that simply host or publish someone else’s ad, without being involved in creating it, unless they are given written notice of a violation and five days to fix it. In other words, if a brand hands you a banner ad or a sponsored video that already breaks the disclosure rule, you are not automatically liable for running it. You become liable if you know about the problem and do nothing.

The line gets blurrier the moment a blogger stops just hosting an ad and starts producing one. If you are the one generating an AI spokesperson to sell your own affiliate product, writing the script for a synthetic host to read on your own sponsored video, or building a fully AI-generated testimonial for a brand partnership, you are the advertiser under this law, not a neutral publisher. That distinction turns on who actually made the ad, no matter how big or small your audience is.

What about ads that are exempt from all of this

A few categories sit outside the disclosure requirement entirely, per the bill’s own exemptions. Advertisements for expressive works like films, television, video games, and similar media are exempt, along with audio-only ads and cases where AI is used solely to translate a real human performer’s speech into another language rather than to create a fake one. Those carve-outs exist because the law is aimed specifically at ads that could mislead someone into thinking they are watching a real person, not at every piece of AI-assisted content. Using AI to write ad copy, clean up a product photo, or generate background music does not trigger any of this. The trigger is a fake person, presented as a real one, trying to sell something.

So how do you actually know if this applies to you

Three questions cover most of it. Did anyone on your team, or anyone you hired, generate a performer using AI rather than filming a real person? Would a viewer reasonably believe that performer is human if nobody told them otherwise? And is the disclosure you currently use, if you use one at all, sitting somewhere a reader would actually see it, rather than buried in a caption’s last line or a terms page nobody opens? A yes to the first two and a no to the third is exactly the gap this law was written to close.

I am not a lawyer, and this is not legal advice specific to your situation. But before your next AI-generated spokesperson goes live in a New York-facing ad, it is worth sitting with one plain question: did I make this ad, or am I just running it? If the answer is the first one, the disclosure needs to be visible enough that nobody has to squint to find it.

The post New York’s new AI advertising law took effect June 9 and requires a “conspicuous” disclosure any time a synthetic AI performer appears in an ad, with fines of $1,000 for a first violation and $5,000 for each one after that, and while the law spares publishers who merely host the ad, any blog or brand actually producing AI-generated spokespeople for a New York audience is squarely on the hook. appeared first on The Blog Herald.

That is the civil penalty New Jersey attached to selling or licensing sensitive personal data without permission, and it applies whether the seller is a total stranger or a company you have quietly renewed a contract with for years. The law says a violator “shall be liable to a civil penalty of $50,000 for each record sold, offered for sale, or licensed,” with no floor on how small the transaction has to be before the fine applies.

I have spent a fair amount of time throughout my career on the buying end of audience data. Email lists, retargeting pixels, lookalike-audience packages built from somebody else’s list. Most of that happens through a vendor’s dashboard, a few clicks, a monthly invoice, no real visibility into where the underlying data came from. New Jersey’s law is aimed at that exact gap.

What actually counts as a “data broker” here

The statute defines a data broker as a person or entity that “knowingly collects or purchases the personal data of a consumer with whom the person or legal entity does not have a direct relationship and sells or licenses that data to a third party.” That covers a lot more of the ad-tech and email-list ecosystem than the term “data broker” usually conjures up. A company you think of as an audience-targeting platform, or a list-rental service, can meet that definition without ever calling itself a broker. According to Wiley Rein’s reading of the law, it carves out no exemption based on how much data a company handles or how many consumers it touches, and skips a consent-based exception too.

“Sensitive data” is defined broadly too. Per the same alert, it covers information revealing racial or ethnic origin, religious beliefs, mental or physical health conditions, certain financial information, sexual orientation, immigration status, transgender or nonbinary status, genetic or biometric data, precise geolocation, and any data collected from a known child. If a list you bought was built from health-adjacent browsing behavior, or segmented by anything touching those categories, it is worth a second look before you use it to target New Jersey readers specifically.

Picture a fairly ordinary setup: a parenting or wellness blogger pays a third-party platform for a “New Jersey moms, ages 28 to 45” audience segment to run a sponsored newsletter placement. The platform built that segment by combining pregnancy-app data, pharmacy loyalty programs, and a few retailers’ purchase histories, none of which the blogger ever saw or asked about. Under the old assumption, the platform’s terms of service quietly absorbed that risk. Under this law, a segment built that way touches health data, reproductive information, and possibly a known child’s data all at once, and the buyer has no real way to know that from the dashboard alone.

The compliance clock is already running

Most of the law’s obligations took effect immediately once it was signed on June 30, 2026, which is part of what makes it unusual. The sale-and-licensing ban itself came with no grace period at all. A separate piece, the requirement that data brokers publicly register with the state, activates 270 days later, landing on March 27, 2027, with an initial registration window running from April 1 through June 30, 2027, according to Wiley’s analysis. So there are effectively two deadlines to track: the ban that already applies, and the registry that will let you actually check who is compliant starting next spring.

Buying a subscriber list once in a while does not put every blogger in legal danger under this law. What has genuinely changed is how much responsibility now sits with the buyer to know where the data actually came from, instead of trusting the seller by default. If how you do anything is how you do everything, this is one of those quiet backend habits worth tightening even when nobody is watching.

A quick check before you target New Jersey readers

None of what follows is legal advice, just the practical questions I would want answered before I trusted a vendor with my own newsletter targeting.

1. Ask your vendor directly whether they are registered

Once the registry opens next spring, this becomes a simple yes-or-no question you can ask any vendor selling audience data that touches New Jersey residents. Until then, ask whether they are tracking the requirement at all. A vendor who has never heard of the law is not automatically doing something wrong, but it is a signal to look closer.

2. Read what kind of data is actually in the list you bought

Segment names like “health-conscious moms” or “recently divorced” sound like marketing copy, but they can map directly onto the sensitive categories the law names. If a list was built around health, financial situation, immigration status, or a child’s data, treat it differently than a generic newsletter signup list.

3. Check whether the seller relies on consent language you have never seen

The law does not include a consent-based exception, so a vendor’s privacy policy promising that “users consented to data sharing” is not, on its own, a shield. Ask to actually see the consent mechanism rather than taking the bullet point at face value.

4. Put the 2027 registry window on your calendar anyway

Even if none of this feels urgent today, the registration window closing June 30, 2027 is the moment this stops being a hypothetical and starts being a public, checkable list. Vendors who are not on it by then will be a lot easier to spot, and a lot harder to justify still working with.

Final thoughts

I am not a lawyer, and nothing here is legal advice. If any of this genuinely applies to how you run your list or your ad targeting, a privacy attorney can look at your actual contracts in twenty minutes in a way that no blog post, including this one, ever really can.

But knowing the four questions above is enough to figure out whether you need that call in the first place.

The post New Jersey’s new data broker law bans the sale of sensitive personal data outright and fines violators up to $50,000 per record, and any blogger or marketer who buys third-party audience lists to target New Jersey readers now needs to check whether their vendor is registered before the state’s compliance deadlines hit. appeared first on The Blog Herald.