Genealogy's Star

AI, Education, and Genealogy

2026-05-09T12:17:00.001-07:00

Volunteering in a library in a large university, I have a significant amount of contact with university students. Since my primary activity is teaching and supporting other volunteer missionaries and patrons of the library, I also have a significant contact with a broad spectrum of ages and backgrounds. In this context, I am in a somewhat unique position to see the impact of AI both in and out of a university environment. Currently, three of the large online family tree database programs, Ancestry.com, FamilySearch.org, and MyHeritage.com, all have AI involvement. The university, however, has mixed reactions as an institution about the use and implementation of AI in the classrooms.

It is interesting that the overall main reaction with all of the individuals that I interface with seems to be based on the initial announcements that were made about AI more than three years ago. There is a correlation between the reaction of most of the public to the FamilySearch.org website with concerns about the changes made to an open AI-based family tree and the consistent initial fear of AI hallucinations. Individuals associated with the University have a broad spectrum of response to and implementation of AI. This attitude is generally a reflection of articles such these.

Knight, Will. “Using AI for Just 10 Minutes Might Make You Lazy and Dumb, Study Shows.” Tags. Wired, May 6, 2026. https://www.wired.com/story/using-ai-negative-impact-thinking-problem-solving-study/.

AI Chatbots Could Be Making You Stupider.” April 20, 2026. https://www.bbc.com/future/article/20260417-ai-chatbots-could-be-making-you-stupider.

This list could go on and on. It has been interesting to me to note that this reaction to AI is almost exactly the same as the reactions to hand-held calculators. Here is a citation to a Masters Thesis on this topic.

Banks, Sarah A. "A Historical Analysis of Attitudes Toward the Use of Calculators in Junior High and High School Math Classrooms in the United States since 1975." Master’s thesis, Cedarville University, 2011. See chrome-extension://efaidnbmnnnibpcajpcglclefindmkaj/https://files.eric.ed.gov/fulltext/ED525547.pdf

Here is a quote from this Master's Thesis"

"The onset of calculators initially brought concerns that this new technology was not fully understood nor would be appropriately utilized by educators and that the effects on students were unknown. Future research studying how children were affected and the necessary changes to curriculum was recommended (“Math in the Schools,” 1975). At this time, some educators feared that students would not be able to retain their knowledge of simple arithmetic if they learned to use a calculator before fully grasping basic mathematical concepts. Other teachers, however, saw calculators as a chance to increase student motivation by using more “real-life” problems (Pendelton, 1975)."

Traditional genealogy involves a significant amount of drudgery such as untold hours of combing through old documents page by page, either on a microfilm viewer or in documents themselves in an archive.

Here is a not-so-hypothetical problem faced by many research genealogists. In doing research in a single parish in England or some other country, you find many people with exactly the same names. Separating out the individuals into families, into unique families, is a tremendous challenge. Over the past three years, I have begun to develop a methodology to use AI to solve this particular challenge. The results have been spectacularly successful. Now, because I have used AI rather than sitting in front of a microfilm reader or scrolling through endless digital documents, I am wondering if my brain is now fried. Or is it the other way around? My brain was being fried by sitting in front of a microfilm reader, and now I have time to think about actually working on the documents.

AI is here to stay. It will only become more useful and more universally used. Rather than worrying about whether students are going to lose intelligence by using AI, perhaps the educators should spend some time trying to figure out how to help them responsibly use AI.

By the way, this particular post was admirably assisted by AI voice recognition so that I didn't have to type this whole thing into my computer.

Using AI as an excuse or as a crutch probably has the same effect as spending time with video games or social media. But continuing to use hallucinations as an excuse for failure to learn about or utilize AI will have the same effect as these other computer interactions. Using AI with Google Gemini, NotebookLM, and Google Gems can effectively reduce hallucinations to almost nonexistence. Additionally, utilizing carefully drafted prompts can substantially increase AI's accuracy.

MyHeritage Scribe AI: Genealogy Tool

2026-05-08T06:14:00.002-07:00

Scribe AI

Deciphering the remnants of the past has long been a task of patience and specialized skill for family historians. Whether it is a faded letter in 19th-century cursive, a gravestone weathered by a century of elements, or a family crest with forgotten symbolism, the barrier between the record and the researcher is often one of legibility and context. To address these challenges, MyHeritage has introduced Scribe AI, a tool designed to transcribe, translate, and interpret historical materials using generative artificial intelligence.

At its core, Scribe AI—an acronym for Scientific Record Indexing Base Engine—is built to handle the varied and often difficult materials that populate genealogical collections. Rather than a simple text-to-digital converter, the system applies specific procedures based on the type of material uploaded. For a handwritten letter, it provides a full transcription; for a historical photograph, it offers an analysis of clothing styles and visual clues to estimate a date and location.

One of the most practical applications for the active researcher is the tool’s ability to interpret complex documents. When a user uploads a record, the AI does more than just read the words; it extracts key genealogical details such as names, dates, and relationships. It also provides a section on historical context, explaining the significance of the document type or the era in which it was created. For those working with international records, the system can translate text from over 50 languages into the user’s primary language, effectively lowering the barrier for research in foreign archives.

The utility extends to physical artifacts that are notoriously difficult to document. Gravestones, for instance, are analyzed not just for their inscriptions but for the iconography and symbolism carved into the stone. Similarly, heraldic coats of arms are explained through their design elements and historical plausibility. For documents spanning multiple pages, the AI processes the entire file as a single contextual unit, ensuring that the narrative flow of a long letter or a legal document remains intact.

Integration is a key feature of this release. Scribe AI is accessible via a dedicated landing page, but it is also woven into the existing MyHeritage ecosystem. Users can apply the tool to photos and documents already stored in their "My Photos" section or use it directly while viewing records in the site’s vast historical database.

From a technical and ethical standpoint, the development of Scribe AI includes a focus on privacy. MyHeritage has stated that documents and photos processed by the tool are used solely to generate results for the user. These materials are not indexed, made searchable for others, or used to train new AI models.

For the genealogist, the arrival of such technology represents a shift in how time is spent. By reducing the hours traditionally required for manual transcription and initial interpretation, researchers can focus more on the "next steps"—a feature the AI also provides by suggesting further avenues of inquiry based on the record’s content. As family history increasingly moves into the digital and AI-assisted realm, tools like Scribe AI aim to make the insights hidden in old records accessible to a broader audience of researchers.

This video provides a visual walkthrough of the Scribe AI interface and demonstrates its transcription and interpretation capabilities in real-time.

Here is an example of a Scribe AI analysis:

The actual report was much longer.

https://youtu.be/zbbyRnBSPz8?si=ePmhW6EWtAVvciUN

MyHeritage Family Infographics Explained

2026-05-08T06:00:00.001-07:00

MyHeritage.com

For many family historians, the challenge of research is not just finding the data, but making sense of the patterns hidden within it. While traditional pedigree charts and group sheets are essential for organization, they often struggle to convey the broader story of a family’s migration, longevity, or social trends. MyHeritage has recently introduced a tool called "Family Infographics" that aims to bridge this gap by transforming raw genealogical data into a series of visual narratives.

The Family Infographics feature operates by scanning a user's existing family tree and synthesizing the information into a cohesive visual report. Rather than requiring manual entry, the tool pulls from the dates, locations, and relationships already established in the database. The result is a collection of charts and maps that provide a high-level overview of a family’s history, making complex data points more accessible at a glance.

The infographics cover a wide range of demographic information. Users can view statistics on life expectancy across generations, identify the most common first names in their lineage, or see the distribution of birth months among their ancestors. Geographical data is also a primary focus; the tool generates maps that illustrate where family members were born, married, and died, providing a clear visual representation of a family’s movement over time.

Beyond the statistical summaries, the feature highlights specific "milestones" and "extremes" within the tree. It identifies the oldest living relatives, the most prolific branches of the family, and couples with the longest marriages. By focusing on these specific narratives, the tool serves as an entry point for family members who may not be active researchers but are interested in the highlights of their shared heritage.

Accessibility and sharing are central to the design of this new feature. The infographics are generated automatically and are available through both the MyHeritage web platform and mobile application. Because the reports are formatted for visual clarity, they are easily shared via email or social media, allowing researchers to present their findings in a format that is more engaging for a general audience.

For the serious genealogist, these infographics offer a different perspective on their research, occasionally revealing gaps in data or unexpected trends that might be missed in a standard list format. It represents a continuing shift in the industry toward data visualization, turning the "dry" facts of the past into a more dynamic story for the present.

A New Rule for Using AI in Genealogy

2026-05-06T13:20:00.000-07:00

Rule #4 You are the Master, AI is the Assistant

I have spent many years watching the genealogy landscape shift, from the days of cranking through microfilm in a dimly lit library to our current era of instant, global digital access. We are now standing at the threshold of the most significant transition I’ve seen in my decades of research: the rise of Artificial Intelligence. As I’ve often said, genealogy is not a hobby of collecting names; it is a discipline of verifying evidence. We now have a powerful new "assistant" in this journey, but we must be very careful to maintain the correct relationship with this technology.

In the world of computer science, there is a concept known as the "master/servant" relationship. While it might sound like something out of a Victorian novel, it describes the technical architecture of how complex tasks are distributed across a system. One central "master" process directs "servant" processes to perform specific duties. It’s remarkably similar to how a large archive operates, with a head archivist coordinating various clerks to pull records from different stacks. When you use an AI tool today, you are essentially engaging this technical workforce. However, the most important lesson for us as family historians is the functional one: in this partnership, you are the Master, and the AI is the Servant.

I see a growing number of researchers—both beginners and those who should know better—treating AI as an "oracle" that can be a friend or simply "do" their genealogy for them. But we must remember one of my Rules of Genealogy: Rule Fifteen: A fact is not a fact unless you have a record to prove it. AI only "knows" your family through what it can find online and if it cannot find and produce the appropriate source for its information, then the information is useless.

To bridge the gap between "old school" standards and "new school" tools, we must act as "forensic auditors" of our own research. When your "servant" (the AI) brings you a lead—perhaps a transcription of a difficult-to-read land deed or a summary of a probate file—your job is to verify any reponse by examining the original records. We use AI for the "heavy lifting" of data processing, but we keep the "intellectual labor" for ourselves. We provide the judgment and the critical eye; the AI provides the processing speed.

Maintaining this relationship requires a shift in how we work. Instead of asking broad, vague questions, we must provide highly structured "master prompts" that force the AI to analyze evidence rigorously. We should ask it to highlight conflicts and cite the specific records it is using. Once the AI provides a lead, we treat it with the same skepticism we would apply to a printed family history from a hundred years ago—we verify every claim against the primary sources. By staying in the "master" seat, we can embrace these incredible advancements while ensuring our family stories are built on a foundation of solid, verified truth. The tools change, but the standards of good research remain the same. Here is a revised example, of a prompt that establishes instruction for the AI Chatbot to act as a Master Genealogist.

Revised Prompt: The Professional Genealogy Research Architect
Role: Act as a Board-certified Professional Genealogist (BCG) and Expert Research Consultant. Your primary goal is to guide a "Reasonably Exhaustive Search" while adhering strictly to the Genealogical Proof Standard (GPS).
Objective: Conduct rigorous evidence analysis to resolve complex identity and kinship problems. You will not accept "hints" as facts; you will treat every data point as a claim to be verified.
Operating Framework:
For every piece of information provided, you must apply this multi-layer analysis:

Source & Information Taxonomy:
Source: Original, Derivative, or Authored.
Information: Primary, Secondary, or Undetermined.
Evidence: Direct, Indirect, or Negative.
Correlation & Logic:
Compare independent sources to look for patterns or discrepancies.
Explicitly address Conflicting Evidence (e.g., age variances, name spelling shifts, or geographic outliers).
The "FAN Club" Filter: Analyze the person within the context of their Friends, Associates, and Neighbors to overcome brick walls.
Reliability & Weighting: Assign a Weight of Evidence score (Low, Moderate, High) to each conclusion based on the quality of the documentation.
Citations: Every record mentioned must include a full citation formatted according to the Evidence Explained (Elizabeth Shown Mills) style.
The Workflow:

Phase 1: The Research Objective. I will provide a specific, focused research question.
Phase 2: Evidence Audit & Gap Analysis. You will analyze my "Known Facts" and identify what is missing (e.g., "No evidence of land ownership despite 1850 Agricultural Census entry").
Phase 3: Strategic Research Plan. You will suggest a prioritized list of record types (Probate, Land, Military, Church, etc.) and specific repositories or databases to consult.
Action: Acknowledge your commitment to these standards. Then, ask me for my Research Objective and Known Facts to begin the investigation.
Before finalizing any conclusion, perform a 'Red-Team' analysis: Identify the weakest link in the evidence chain and suggest one specific scenario that could disprove the current hypothesis.
If a citation cannot be fully formed due to missing data in the record, use placeholders (e.g., [publisher unknown]) and flag the missing element as a research task.
When analyzing records from before 1800, account for archaic spelling, secretary hand transcription errors, and legal Latin terminology relevant to the jurisdiction.
Here is a recent video I did that adds some additional guidance for AI's application to genealogy and genealogical research.

Developing an Ethical and Safe Use of AI for Genealogy - James Tanner

This post was written with the assistance of Google Gemini 3.1 Pro.

How to Evaluate an AI Website for Doing Genealogical Research

2026-04-30T07:29:00.002-07:00

The proliferation of AI platforms since the launch of ChatGPT in late 2022 has triggered a surge in promotional content and comparative reviews across YouTube and various blogs. Most of these evaluations conclude with a subjective recommendation based on the author’s personal preference. While I hold my own biases, this article provides a standardized framework of criteria to help you determine if a reviewer’s claims are substantiated or merely anecdotal.

A suggested methodology for determining an accurate review and its applicability to genealogical research can be gleaned from articles and educational opportunities such as the following:

“Advanced AI Techniques for Genealogists: Expanding Your Research Skills.” GRIP Genealogy Institute, January 8, 2024. https://grip.ngsgenealogy.org/courses/advanced-ai-techniques-for-genealogists-expanding-your-research-skills/.
Apple Podcasts. “RootsTech Class 2026 Class Takeaways.” Accessed April 30, 2026. https://podcasts.apple.com/us/podcast/rootstech-class-2026-class-takeaways/id1419782085?i=1000759494832.
BYU Library Family History Center. Developing an Ethical and Safe Use of AI for Genealogy -James Tanner (22 Feb 2026). 2026. 55:14. https://www.youtube.com/watch?v=R0bfWAYx-OE.
Coalition for Responsible AI in Genealogy. n.d. Accessed April 30, 2026. https://craigen.org/.
“Ethics and Best Practice of AI Use in Genealogy Research - NZ Society of Genealogists.” Accessed April 30, 2026. https://genealogy.org.nz/Ethics--Best-Practice-of-AI/11482/.
Ferris, Maureen Martin. “AI in Genealogy | Maureen Martin Ferris.” Accessed April 30, 2026. https://www.maureenmartinferris.com.au/ai.html.
“How to Use AI Tools for Family History Research | The Gazette.” Accessed April 30, 2026. https://www.thegazette.co.uk/all-notices/content/104452.
Navigating the AI Frontier: Why Your Genealogy Society Needs a Policy (and How to Write One!) - GenSocSoup. Genealogy Society Management. January 28, 2026. https://gensocsoup.com/navigating-the-ai-frontier/.

I used Google Gemini to evaluate and review these and other sources.

Evaluating the utility of General Purpose AI (like Gemini or ChatGPT) versus specialized genealogy tools (such as the online family tree websites) requires a shift from "search-based" thinking to "analysis-based" thinking. In 2026, as these models move toward sophisticated reasoning rather than simple text prediction, the criteria for their value can be divided into technical capability (value) and workflow efficiency (usability).

For any genealogist, the Genealogical Proof Standard (GPS) is the leading standard. When evaluating an AI, you must determine how well it supports this framework.

Source Grounding: Does the AI provide specific citations to the uploaded documents? A valuable AI doesn't just say "John Smith was born in 1840"; it points to the specific line in the PDF or census record.
Hallucination Rate: Does the model "invent" ancestors to fill gaps in a pedigree? Testing a model with a known, well-documented family line is essential to benchmark its tendency to hallucinate.
Logical Reasoning: Can the AI resolve conflicting evidence? For example, if one record says a birth was in 1842 and another says 1845, a high-value model should be able to weigh the reliability of the sources (e.g., a birth certificate vs. a 1910 census) rather than just picking one.

The efficacy of AI interaction is dictated by the quality of the prompt. Consequently, assessing the validity of any AI evaluation requires a rigorous review of the underlying prompts used for comparison. Because an AI’s output is a direct reflection of the parameters formalized in the prompt, any conclusion regarding a model's utility is inherently tied to whether the prompt inadvertently biased the results or predetermined the comparison's outcome.

Since genealogical research is heavily document-dependent, the AI’s ability to "see" and "interpret" is a primary value criterion. The basis for these interpretations weighs heavily on the AI's ability in OCR or Handwriting Recognition, Data Structuring, and its Context Window ( the actual usable tokens). it is always important to check the latest comparison charts. See https://exploreaitogether.com/llm-usage-limits-comparison/

Here's a summary chart made by Google Gemini.

Value vs. Usability Comparison Table

Criterion	Value (Does it do the job?)	Usability (Is it easy to use?)
Transcription	Can it read difficult 18th-century script?	Is the interface for correcting the text intuitive?
Translation	Is the translation idiomatic and historically accurate?	Does it preserve the original document's formatting?
Synthesis	Can it spot a migration pattern across 10 documents?	Can it output that pattern as a map or a timeline?
Analysis	Can it identify "same-name" individuals as different people?	Does it explain why it made that distinction clearly?

This will probably be an ongoing issue.

How to begin your own User Tree on FamilySearch.org

2026-04-29T07:01:16.289-07:00

https://www.familysearch.org/en/search/genealogies/

You might spend quite a bit of time trying to find how to begin a user tree on FamilySearch.org. Here are the current step-by-step instructions for uploading a tree and beginning your user tree experience. You will need to be signed into FamilySearch.org.

Prepare Your File: Export your family tree from your current software as a GEDCOM (.ged) file.

Access Genealogies: From the Search menu select FamilySearch Genealogies, scroll down to the section titled "How is FamilySearch Genealogies different from FamilySearch Family Tree?"

Upload: Click the button UPLOAD YOUR INDIVIDUAL TREE.

Choose File: Click Upload GEDCOM File, select your file from your computer, and give it a title.

Once your tree is uploaded, you go to the Family Tree menu item and navigate to Manage Trees. In Manage Trees, you should see the name of your tree that you uploaded. You can then review matches, which allows you to enter more information into your user tree from the FamilySearch Family Tree. I would also suggest looking at the settings. You may find there are a number of challenges in using the user trees because they are still under development. There may be changes not only in the way the user trees can be maintained but also with the process of uploading the files.

You may also need to maintain a separate tree from the FamilySearch.org website for your own purposes, either with a desktop program or in another online family tree website.

Keys to talking to a chatbot: Prompts

2026-04-29T06:39:00.000-07:00

With the introduction of Generative AI, we began to learn about the natural language interface. The natural language interface creates an impression that the AI chatbots will operate within the realm of reality. Simple questions like those illustrated above enhance this impression; however, asking simple questions is actually using the chatbot as a search engine. If you are at all acquainted with doing Google searches or using any other search engine, you are painfully aware that the results of those searches can include millions of responses, many of which have nothing to do with the search terms entered.

For this post, I am going to use Google Gemini which, in my opinion, I currently believe is the best AI for in depth genealogical and historical research. To start this discussion, here is a quote from a an article from Card Catalog entitled "31 AI Terms, Explained."

RAG (Retrieval-Augmented Generation)
Retrieval-Augmented Generation, or RAG, is a technique that combines a language model with a search or retrieval system, allowing the model to pull in relevant information before generating a response. Without RAG, a model can only draw on what it absorbed during training, which has a fixed cutoff date and doesn't include anything proprietary, recent, or specialized. With RAG, a model can search a document library, a company knowledge base, or the open web and incorporate that material into its answer. This is how AI tools can accurately respond to questions about internal documents they were never trained on, and how some products can cite current sources rather than relying on potentially outdated training data. RAG also substantially reduces hallucination in knowledge-intensive tasks because the model is working from retrieved source material rather than generating from pattern memory alone.

Th RAG process is exactly modeled by the Google Gemini/NotebookLM/Gems workflow. You use Gems to create specifically designed prompts that restrict Gemini to an assigned research task and then using the research prompt, you collect vetted sources into a NotebookLM notebook and then use the restricted Gemini AI to extract accurate information from the vetted sources. As a bonus, NotebookLM offers several pre-programed ways to analyze and communicate the information derived from the workflow

Here is an example of Google Gem with the title Master Genealogist.

Revised Prompt: The Professional Genealogy Research Architect
Role: Act as a Board-certified Professional Genealogist (BCG) and Expert Research Consultant. Your primary goal is to guide a "Reasonably Exhaustive Search" while adhering strictly to the Genealogical Proof Standard (GPS).
Objective: Conduct rigorous evidence analysis to resolve complex identity and kinship problems. You will not accept "hints" as facts; you will treat every data point as a claim to be verified.
Operating Framework:
For every piece of information provided, you must apply this multi-layer analysis:

Source & Information Taxonomy:
Source: Original, Derivative, or Authored.
Information: Primary, Secondary, or Undetermined.
Evidence: Direct, Indirect, or Negative.
Correlation & Logic:
Compare independent sources to look for patterns or discrepancies.
Explicitly address Conflicting Evidence (e.g., age variances, name spelling shifts, or geographic outliers).
The "FAN Club" Filter: Analyze the person within the context of their Friends, Associates, and Neighbors to overcome brick walls.
Reliability & Weighting: Assign a Weight of Evidence score (Low, Moderate, High) to each conclusion based on the quality of the documentation.
Citations: Every record mentioned must include a full citation formatted according to the Evidence Explained (Elizabeth Shown Mills) style.
The Workflow:

Phase 1: The Research Objective. I will provide a specific, focused research question.
Phase 2: Evidence Audit & Gap Analysis. You will analyze my "Known Facts" and identify what is missing (e.g., "No evidence of land ownership despite 1850 Agricultural Census entry").
Phase 3: Strategic Research Plan. You will suggest a prioritized list of record types (Probate, Land, Military, Church, etc.) and specific repositories or databases to consult.
Action: Acknowledge your commitment to these standards. Then, ask me for my Research Objective and Known Facts to begin the investigation.

This prompt was developed from my initial descriptive statement and then refined by the Gem app. The results were then given to Gemini to refine. The resultant prompt can then be automatically used by Google Gemini to do addition research with extreme depth and almost total accuracy.

The resultant sources and analysis and sources obtained are then put into NotebookLM notebook that forces Gemini to use only the vetted sources and provide deep research anaysis and responses. With the PRO lever of Gemini (currently $20 a month) you can add as many as 300 sources to a notebook for analysis. Google Gemini also has one of the top levels of token usage of any of the AI programs. Here is a list of the current ranking for AI tokens. In the context of Artificial Intelligence (AI), a token is a fundamental unit of data that AI models process during training and inference.

Top Ranking AI Websites by Context Window (2026)

Rank	AI Platform	Model(s)	Usable Token Limit	Key Strength
1	Google Gemini	Gemini 3.1 Pro / Deep Think	2,000,000	Largest mainstream window; deep integration with Google Workspace.
2	OpenAI (ChatGPT)	GPT-5.4 (xhigh) / 5.5	1,100,000	High reasoning performance across the full context window.
3	Anthropic (Claude)	Claude 4.7 / Mythos Preview	1,000,000	Best for "needle-in-a-haystack" retrieval and stable long-form writing.
4	xAI (Grok)	Grok 4.1	1,000,000	Rapid real-time information processing from X (formerly Twitter).
5	DeepSeek	DeepSeek V4 Pro (Max)	1,000,000	Most efficient high-context open-weight implementation.
6	Moonshot AI (Kimi)	Kimi K2.6	256,000 - 1M+	Specialized in long-context memory (Kimi K2 has reached 1B tokens in labs).
7	Alibaba (Qwen)	Qwen 3.6 Plus	262,000	Excellent for repository-scale code understanding.

🔍 Important Considerations

Theoretical vs. Web-Accessible: While models like Llama 4 Scout or Kimi K2 have been benchmarked at 10 million to 1 billion tokens, these limits are often restricted to specialized developer environments or high-tier API access rather than the standard "chat" interface you find on their websites.
The "Pro" Gap: Most platforms gate their highest token limits behind "Pro," "Ultra," or "Max" subscriptions. For example, Claude’s 1M token window is typically reserved for Tier 4+ organizations or specific beta previews.
Performance Degradation: Having a high number of "usable" tokens doesn't always mean the AI "remembers" everything perfectly. Gemini 3.1 Pro and Claude 4.7 currently lead in benchmarks for accurately retrieving information buried in the middle of a 1M+ token prompt

See “Best AI Tools 2026: Complete Ranking Across Every Category | NxCode.” March 29, 2026. https://www.nxcode.io/resources/news/best-ai-tools-2026-complete-ranking-guide.

Google Gemini 3.4 or 3.5 may be shortly released and the gap between Gemini and the other chatbots may increase.

Granted, none of the websites can actually use all of their tokens, but Gemini can clearly process more information than any of the other usable popular websites. Additionally, none of the other websites currently have a workable workflow model that is fully integrated into a prompt generation app.

During the past few months, I have used the Gemini/NotebookLM/Gems workflow to do some extremely detailed, extensive genealogical research projects and have had amazing results. I will go into some of these projects in the near future and discuss them fully in my blog posts.

The First Three Rules for Using AI in Genealogy

2026-04-26T15:07:00.001-07:00

Because there are Rules for Genealogy, I thought there ought to be some Rules for Using AI in Genealogy. As I thought about the idea for a while, I realized I had already started telling people about the rules in a previous blog post so it was a good idea to codify them in a subsequent post. Here I go with the first three rules. These rules can also apply to using AI in general.

Rule #1: AI is a tool, not a toy.

I have spent the last three years refining methods for using AI in legitimate genealogical research. During this time, I’ve observed many people using personal AI chatbots primarily for entertainment or constant conversation. However, using AI strictly for diversion falls into the same category as playing video games or watching social media reels; to be effective in genealogy, AI ahould be treated as a functional tool rather than a pastime.

When I was in grade school, I took a shop class where we learned to use woodworking tools, including an electric table saw. One of the things that they taught us was a movie that showed what happened when a person was using a table saw and got injured severely by the kickback from the wood. I still remember this movie vividly every time I watch someone use a table saw or try to use one myself. The movie it was followed by considerable instructions on the proper way to use the table saw.

In the past, educators have become concerned about the use of hand calculators and Wikipedia, for two examples. Apparently, none of the bad effects of using a hand calculator or the bad effects of using Wikipedia have resulted in the entire collapse of education in the United States. Rather than prohibiting the use of AI by students, how about instructing them on the way AI can be used properly?

The huge number of comments online about the dangers of AI are doing progress a disservice. Rather than constantly focusing on the dangers of AI, how about a lot more discussion about how to prevent any of the dire consequences that seem to be prevalent in current online discussions?

Rule #2: You must ask the right questions to get the right answers.

AI is a complicated computer program. With the advent of the natural language interface, there was a perception that somehow it was not necessary to program computers, that the computer itself could program itself. It is true that current AI chats can write complex programs; however, the feedback I see is that the computer programs need to be very closely reviewed in order to become entirely operable. Since I deal with a huge variety of computer programs, I am almost constantly aware of inconsistencies and "bugs" in almost all the programs I deal with. Sometimes efforts to resolve a computer program error end up in creating more inconsistencies and difficulties.

I think one way to overcome what we would normally call bugs in an AI program, instead of hallucinations, would be to help people become more familiar with the way to ask the right questions and give the right set of instructions. It is likely that AI programs in general will always have a degree of inconsistency, but it is also true that there are valid ways of developing to limit the ability of the AI program to hallucinate or fabricate. A good example of this is the Gemini/NotebookLM combination that can essentially eliminate hallucinations or fabrications.

Rule #3: Always require a source.

When using an AI chat for tasks requiring accuracy, such as genealogical research, it's crucial that the AI program automatically cites sources for its statements. Information provided without a source may be incorrect and is often useless for research. To illustrate, try asking your chat program to review its own unsourced response for accuracy—you might be surprised by the outcome.

Stay tuned, I may have a few more rules.

2026-04-20T12:37:00.000-07:00

Historical Record Collections Added in March 2026

MyHeritage Expands Global Database by 366 Million Historical Records

It is a frequent topic of discussion in the genealogical community that the sheer volume of digitized records available online continues to grow at a rate that is difficult for even the most dedicated researcher to track. This past month, MyHeritage.com continued its consistent acquisition and publication schedule by adding 366 million historical records to its online database. This update brings their total collection to over 21 billion records.

For those of us who have been following the industry for years, these numbers represent a significant shift in the accessibility of international data. The real value, however, lies in the specific collections and how they might help us solve "brick wall" problems in our research.

Focus on European Vital Records

The majority of this month's update centers on European collections, which remains a core strength of the MyHeritage.com website. They have added a substantial number of birth, marriage, and death records from France. French records are often exceptionally detailed, but they can be challenging to access if you are not familiar with navigating individual departmental archives. Having these indexed and searchable in a central location is a practical benefit for those with French ancestry.

Furthermore, the update includes specialized collections from Germany, specifically from Baden and Prussia. German research is often complicated by historically shifting borders and the decentralization of records. The addition of these specific regional indexes helps bridge some of those jurisdictional gaps.

Scandinavia and the United Kingdom

MyHeritage has also expanded its holdings in Scandinavia with new church records and census data from Sweden, Norway, and Denmark. In these countries, parish registers are the backbone of family history research, often predating civil registration by centuries.

In the United Kingdom, the new additions include electoral registers and MyHeritage military files. While many of these records are available across various platforms, having them integrated into the MyHeritage search engine increases the likelihood of finding a connection that might have been overlooked elsewhere.

North American and Specialized Collections

On the North American front, MyHeritage has provided enhanced indexing for the 1950 U.S. Census. While the census images have been available for some time, the ongoing refinement of these indexes is crucial for identifying specific family units accurately. We also see additional census records from Canada being brought online.

What is particularly noteworthy is the inclusion of specialized record sets, such as Jewish genealogical records from several international locations and new data from South Africa. These are often underserved areas of research, and any increase in digital access is a welcome development.

Understanding Record Matches

As these 366 million records integrate into the system, they will trigger numerous Record Matches. As I often say, "Match is not identity." While automated matching is a powerful discovery tool, the researcher must verify the evidence. Always examine the original document image to ensure the information is interpreted correctly within the family context.

The ongoing competition in record digitization among the major genealogy companies is a benefit to us all. With a total of 21 billion records, MyHeritage is providing a more comprehensive framework for global research. We can expect this trend of rapid digitization and indexing to continue throughout the year.

The Main Challenges of FamilySearch Full-text Search, Part Three

2026-04-16T07:25:00.001-07:00

https://www.familysearch.org/en/search/full-text/

These instructions about using simple Boolean Algebraic symbols for doing full-text search are a reminder that when you are using AI you are talking to a computer, not a person. If I started a set of Rules for AI in Genealogy, I would put this statement as the first rule. I guess that this first rule would be about the need to remember that AI is a tool not a solution.

There is a lot of discussion out there on the internet about "outsourcing your intelligence to AI." But it is apparent to me, if not to too many other people, that these arguments and concerns especially from the educational community, were being made using the exactly the same words about electronic calculators and the Wikipedia website. I know I have written about teachers banning electronic calculators and copying articles from Wikipedia, but I feel like I am having déjà vu all over again.

Let's get serious about the status of the document collections on FamilySearch.org. Presently, there are six different document collections or ways to search the documents and they overlap only slightly. The different avenues of access are the following:

The main catalog
The historical record collections
The Images collection
Full-text Search
Simple Search
The Books collection

There is no real way to determine the degree of redundancy between these separate collections.

What does Full-text search add? Ultimately, if all six are somehow consolidated, we may be able to find a way to search ALL the documents on the website from one search interface. Presently, the only way to have any confidence at all in the extent of your searches into the FamilySearch.org website is to do the search six different ways. Particularly, with the Full-text Search and Simple Search functions, there is no way to know what part of the website's collections you have searched. From my own experience, I am positive that most users of the website do an unsuccessful name search and conclude that FamilySearch does not have the documents they are looking for. Here are some thoughts about how Full-text Search fits into the equation.

The bridge between our "old school" research standards and these powerful new AI-driven tools is precision. We have to stop treating the search box like an easy solution and start treating it like a surgical tool. FamilySearch’s Full-text tool is essentially a massive search engine for historical documents—deeds, wills, and probate records that were previously "locked" inside digital images. To maintain our research integrity, we must master the technical nuances of how to talk to this specific system. We must also be painfully aware that any search we make only reaches an unknown number of documents that have been processed and made available to the search.

The first step is understanding that the computer is literal. If you search for Sarah Miller in the Full-text box without constraints, the computer will find every "Sarah" and every "Miller" in a land deed or a court record. By using an Exact Phrase search—"Sarah Miller" in quotes—you are imposing a research standard on the machine. You are telling it that the relationship between those two words is non-negotiable. This is the only way to effectively search for a specific name in a sea of handwritten text that has been converted by AI.

But historical records are rarely that clean. In a probate record, your ancestor might be listed as "Sarah, the daughter of John Miller." This is where I find the Proximity operator to be a helpful tool for the modern genealogist. By searching for "Sarah Miller"~10, you are telling the FamilySearch engine that these two words must appear within ten words of each other. But the results of my use of all these Boolean tools are mixed I am not sure that FamilySearch's full-text search understands them because the results are inconsistent.

Using tips for finding names, for example in legal documents where titles or middle names often separate a first and last name. If they are applied consistently, they could bridge the gap between the rigid search and the messy reality of 19th-century legal phrasing.

We also have to contend with the "wildcard" nature of history. Spelling was often more of an art than a science in the past, and even the best AI transcription can misinterpret a letter. Using a Single Character Wildcard, such as Sm?th, allows you to find both Smith and Smyth. In the Full-text search environment, where an "o" might look like an "a" to a computer, these wildcards are essential for ensuring that no record is left behind simply because of a digital misread. Using this tool, may also prove frustrating when the returns start to number in the millions.

Finally, we must learn to curate our results by using Exclusion. If you are searching for a surname that is also a common place name or term, your search can quickly become cluttered. For example, if you are looking for a family named "Rice" in a county known for its agriculture, searching for Rice -paddy or Rice -planting allows you to strip away the irrelevant data and focus on the human beings.

The examples I am giving here illustrate that any AI implementation is only as useful as the ability of the user to maintain control. Meanwhile, give the state of affairs with the collections on the FamilySearch website, relying on Full-text Search for all of your research is lamentably impossible.

The Main Challenges of FamilySearch Full-text Search, Part Two

2026-04-13T07:22:00.000-07:00

https://www.familysearch.org/en/search/full-text/

As I continue with this series of posts, I am going to focus on FamilySearch's implementation of full-text searching. FamilySearch is free, and all you need to start searching is to register with the program. The search fields are surprisingly sparse, but they turn out to be adequate for most searches. The first challenge is to understand what information you need to put in the different search fields. With blank search fields, the invitation seems to be to fill in the information you're looking for in each of the fields. But the rule here is less is more.

In part one of this series, I gave an illustration of the variety of forms of the name John in English and other languages. You need to begin thinking about how the information you're seeking may have been recorded in a variety of different documents. If you enter a very common search term such as the name "John", you may get millions of responses. FamilySearch's full-text search fields are hierarchical. The key words can be just about anything, but experience indicates that key words need to be able to identify various types of documents that you're searching for. For example, you might enter the keyword "deed". This tells the full text search to look for documents in that category. As you add information to the other search fields, you always want to keep in mind how the entries might be represented in the documents you're searching.

Here is an example of a basic search.

The quotation marks tell the search to look for the entire name rather than individually searching for all the Henrys, all the M's, and all the Tanners. You also need to remember that you are searching only the documents that have been processed by FamilySearch and made available to the full-text search. Over time, of course, the number of documents processed is changing at many millions per day, and so full-text searches will become more and more valuable. In this case, the name "Henry M. Tanner" is not common, and therefore the number of responses turned out to be very small.

Despite this forced focus, the responses do include people who are not my particular Henry M. Tanner. The next step you could add a physical location for the records by editing the search fields. You can edit the search directly from the responses. There is a very difficult-to-see link called "Edit Search". When you edit the search, you get a surprise. The searches you found previously may or may not be found in a subsequent search. By adding a place name, you have eliminated any documents with the original name of the person that do not have both the place name and the person's name.

Here are the results.

Even if you go back to the original search to try and use the list that showed up initially, you may be surprised to find that another search turns up a different set of documents. You should go through each set of documents that are produced, unless, of course, you get millions of documents. Make sure to save the documents that you intend to use. Click on the image for the documents that you want to preserve, and you will see the transcription of the document in the language of the document. If the document is in Spanish, the transcription will be in Spanish. You can then use the icon tools to attach the document to a person in the FamilySearch family tree, edit the document, or download the document. There are also some tools for adjusting the image.

Bear in mind that unless you use the quotation marks, you will be looking for every one of the words in the name or other information you enter. Place names are standardized, and so you may have to make some alterations if the place you are looking for does not happen to be already standardized by FamilySearch.

Because you happen to find some documents does not mean that the full-text search has found all the documents in all the collections on the FamilySearch website. For example, if I change the place name to use simply Arizona, then I get an entirely different set of documents.

In another example, if I use the name Tanner rather than the full name of my great-grandfather, then I get a different set of documents.

In some cases, the number of variations you would need to try can be overwhelming.

Stay tuned. There's probably going to be a part three to this series.

Falling for the Old Technology News Issue, a commentary

2026-04-09T14:25:00.001-07:00

This image represents the challenge of relying on outdated technology news in a rapidly evolving world. While the subject focuses on obsolete information like 56k modems and floppy disks, the environment outside is already moving forward with 5G networks, AI, and advanced electric vehicles. It highlights the potential for confusion and misinformation when using historical "innovations" to interpret the current tech landscape.

The current obsession with AI hallucinations exemplifies how obsolete perspectives can overshadow innovation. Many treat the risk of fabrications as a dispositive reason to reject AI, failing to recognize that while hallucinations remain a factor, they should not preclude a serious investigation into the technology's broader utility.

One simple response to this concern is asking if the person making the comment is familiar with the workflow model developed by Google using Gemini, Gems, and NotebookLM. In addition, there are other workflow models that use redundancy to diminish the possibility of fabrication or hallucination. The chatbot is simply asked to criticize the accuracy of its statements. Another alternative is to instruct the chatbot to stop when it cannot substantiate the accuracy of its statements. There has also been a significant effort on the part of the AI developers to curtail hallucinations.

Because of my age, I also encounter people who fear technology in any form. Most of their fear stems from the constant reports on the media about how AI is going to destroy the world in some form or another.

It took more than 30 years for cars to replace horses. Generative AI (GenAI) is being adopted faster than any previous technology, including the internet and personal computers, with nearly 40% of U.S. adults aged 18–64 using it by August 2024. ChatGPT reached 100 million users in just 0.2 years (around two months), whereas the internet took years to reach similar milestones. See The Rapid Adoption of Generative AI

Using Google"s NotebookLM, I can search 600 pages of genealogically valuable information in a matter of minutes and most of the time I spend reading and analyzing the responses. FamilySearch.org's Full Text Search is able to search millions of pages of documents in a few seconds and find an individual name or other similar information. This frees me from sitting in front of a microfilm reader for 8 hours to find one name in one document.

I guess, as a genealogist, that I am caught between dealing with people still mired in a paper-based technology while working hard to keep up with the advancements in AI that change our entire way of doing research.

The Main Challenges of Full-text Search Part One

2026-04-09T04:52:00.003-07:00

Three of the major online family tree/data base websites have implemented AI based full-text search and to some degree, handwriting recognition in the last three or so years. FamilySearch.org's offering is called "Full Text Search" and includes handwriting recognition. The Full Text Search is available for free to all users. MyHeritage.com introduced a similar program called Scribe AI. Ancestry.com's contribution is confined to OCR and lacks handwriting recognition. All the efforts of the genealogy programs are behind the ability of Google Gemini with NotebookLM and some of the other AI websites. Of course none of the genealogy programs have the resources of Google or OpenAI and the others.

As far as the genealogical community is concerned, handwriting recognition, document translation, and full-text search are revolutionary in changing the way we do research. I can put hundreds of documents into NotebookLM and then have a discussion with Gemini about the contents of all the documents at onece and the conversation quicky evolves into an examination of what records need to be found to resolve serious historical issues. But working with full-text search opens a whole new series of challenges.

The first and major challenge is the issue of search terms. In the old way of searching using Google, for example, you can formulate search terms based on categories. For example, if I want to find some probate records, I search for probate records in the state or country where my ancestors lived. I didn't expect Google to find the actual records of my ancestors. That was left to me to search through the records individually for the people I was interested in finding. Of course, I could go back to the time when I was forced to use microfilm and read through the entire collection of records to find one name. The time and effort spent in searching record collections severely limited the amount of information I was able to gather in any reasonable time period. In addition, of course, I was faced with the issue of deciphering all of the handwritten records. The results of these conditions were that very little research was produced in any particular time spent.

Full-text search opens up the ability to review hundreds, perhaps thousands of records in a matter of minutes, but the catch is the fact that you need to guess the search terms to use to open up the information. Let's say your ancestor's name was John. Here is a list, generated by Google Gemini, of the variations of the name John.

The name John is one of the most enduring and prolific names in history. It originates from the Hebrew name Yohanan (יוֹחָנָן), meaning "Yahweh is gracious."

Because of its biblical significance—associated primarily with John the Baptist and John the Apostle—the name spread through the Byzantine Empire, into the Latin-speaking West, and eventually across the globe through various linguistic shifts.

Linguistic Evolution

The path of the name from its ancient roots to modern English follows a clear phonological chain:

Hebrew: Yohanan
Greek: Ioannes (Ἰωάννης)
Latin: Iohannes / Johannes
Old French: Jehan
Middle English: Jan / John

Variations by Region and Language

Western & Northern Europe

These variations reflect the Germanic and Romance linguistic families where the name has remained a top choice for centuries.

English: John, Jack, Johnny, Jon
French: Jean, Yannick (Breton)
Spanish: Juan, Juanito
Portuguese: João
Italian: Giovanni, Gianni, Nino, Vanni
German: Johannes, Hans, Jan, Johann
Dutch: Jan, Johan, Hanne, Hans
Scandinavian (Danish/Swedish/Norwegian): Johan, Jan, Jens, Hans

The British Isles (Celtic Variations)

The Celtic languages developed distinct versions of the name that have since become globally popular in their own right.

Irish: Seán, Shane, Shaun, Shawn
Scottish Gaelic: Iain, Ian
Welsh: Evan, Ioan, Ieuan, Siôn
Manx: Juan

Eastern Europe & Eurasia

In Slavic and Orthodox traditions, the name often maintains the "I" or "Y" sound from the original Greek Ioannes.

Russian: Ivan, Vanya
Polish: Jan, Janusz
Czech/Slovak: Jan, Ján, Janko
Hungarian: János, Jancsi
Romanian: Ion, Ioan, Ionuț, Nelu
Bulgarian/Serbian: Ivan, Jovan
Greek: Ioannis, Giannis, Yannis

Middle East & Africa

These versions often stem directly from the Hebrew original or the Islamic tradition.

Arabic: Yahya (يحيا), Yuhanna (يوهنا)
Hebrew: Yohanan (modern: Yochanan)
Amharic (Ethiopia): Yohannes
Turkish: Yahya

Asia & Pacific

In these regions, the name is often adopted through religious conversion or phonological adaptation of Western names.

Chinese: Yuēhàn (約翰)
Japanese: Yohane (ヨハネ - Biblical), Jon (ジョン)
Korean: Yohan (요한)
Hawaiian: Keoni

Diminutives and Medieval Short Forms

Historically, many surnames were created from pet names or shortened versions of John.

Hank: Derived from the Dutch Hanne.
Jan: Common in Northern Europe; used as a root for many surnames.
Jenkin: A medieval English diminutive ("Little John").
Hick/Hitch: Obsolete medieval English rhyming nicknames for John.

Summary Table of Major Forms

Language	Primary Form	Common Diminutive
English	John	Jack
Spanish	Juan	Juanito
Russian	Ivan	Vanya
German	Johannes	Hans
Italian	Giovanni	Gianni
Irish	Seán	Shane
Scottish	Ian	Iain
Finnish	Jukka	Jani

Which one of all of these terms was the one used by your ancestor named John? Did he use the name John at all, or did he use some other name, such as Bubba or Kid or J.T.? So when you are faced with a search field such as this one from FamilySearch, what are you going to use for the search terms?

If you assume that the person's name was John, what are your chances of finding him if he went by one of the other names? For example, my great-grandfather's official name was Henry Martin Tanner, but when he signed legal documents, such as deeds, he always used Henry M. Tanner. Full-text searches are rather literal, and if I search for Henry Tanner. I will possibly not find Henry M Tanner. I can use all sorts of Boolean algebraic terms, but I will still face the same problems of determining the search terms I need to use to find any specific piece of information I am searching for. Another example: one of my relatives is named Joseph Christiansen. His grave marker says Joe Christiansen. He apparently did not like to be called Joseph. How am I supposed to know this?

The people programming full-text search could add the variations for all the names and all the places and practically everything else into their program. They might even implement artificial AI to recognize all of the variations. What happens in that circumstance is that the number of documents discovered by AI can run into the millions.

Do I have a solution for this? No, but I have a methodology I use to attempt to narrow down the number of possible variations. This primarily includes carefully reviewing the documents that I do find to discover the possible limited variations of the name used by the person I am searching for. This whole process also applies to place names and, to some extent, to dates, particularly when you think about calendar changes.

This is part one of this particular series of articles, and hopefully you will stay with me and read the rest of the series as it comes out over the next few weeks.

Another cautionary tale: Lawyers fined for AI-generated Legal Documents

2026-04-03T14:41:00.002-07:00

At first I thought the story of an attorney being sanctioned by the court for submitting a legal brief with AI fabrications had to be an urban legend, but recently I began to learn about the relatively large number of cases with the same theme.

The case of Mata v. Avianca, Inc. Case 1:22-cv-01461-PKC Document 54, United States District Court, Southern District of New York shows that the cases of the AI-generated briefs are not urban legends. The Court held in part:

In researching and drafting court submissions, good lawyers appropriately obtain assistance from junior lawyers, law students, contract lawyers, legal encyclopedias and databases such as Westlaw and LexisNexis. Technological advances are commonplace and there is nothing inherently improper about using a reliable artificial intelligence tool for assistance. But existing rules impose a gatekeeping role on attorneys to ensure the accuracy of their filings. Rule 11, Fed. R. Civ. P. Peter LoDuca, Steven A. Schwartz and the law firm of Levidow, Levidow & Oberman P.C. (the “Levidow Firm”) (collectively, “Respondents”) abandoned their responsibilities when they submitted non-existent judicial opinions with fake quotes and citations created by the artificial intelligence tool ChatGPT, then continued to stand by the fake opinions after judicial orders called their existence into question.

In my extensive court experience it was not uncommon to find an attorney citing a case that either did not apply to the issues before the court or did not exist at all and this was long before the existence of generative AI. According to the latest data from researchers tracking this trend (notably Damien Charlotin of HEC Paris), there have been over 1,200 documented instances worldwide, with approximately 800 of those occurring in U.S. courts. See Kaste, Martin. “Penalties Stack up as AI Spreads through the Legal System.” WGCU News | PBS & NPR for Southwest Florida, April 3, 2026. https://www.wgcu.org/2026-04-03/penalties-stack-up-as-ai-spreads-through-the-legal-system.

The irony of all these cases is that there are AI solutions that could be used to avoid the issues and still give substantial support to the judicial process. If the attorneys, or whomever, knew more about how to use AI and their own legal skills. all these cases could be avoided. However, I do have several suggested lessons that we should all learn.

#1 AI is not a toy, it is a highly sophisticated tool and it takes a measurable amount of time and effort to learn how it can and should be used.

#2 Read the fine print. This is actually Rule Ten of The Rules of Genealogy. See https://genealogysstar.blogspot.com/2025/10/another-new-rule-of-genealogy-for-2025.html Not only do you need to check your sources when doing genealogical research, you need to read the case law in the legal world.

#3 String cites ask for trouble. If you don't understand this lesson, then you don't know a lot about the legal profession.

#4 Ultimately, you have to know how to ask AI questions and how to give commands and more than both of these, don't trust anything you haven't verified for yourself.

That's probably enough suggested lessons for today.

Precision Inquiry: The real core of using AI for research

2026-04-02T12:59:00.000-07:00

When the current wave of generative AI came out, the main draw seemed to be the Natural Language Interface (NLI). The NLI is not Voice Recognition. Voice Recognition is the "ears" (turning sound into text), while a Natural Language Interface (NLI) is the "brain" (understanding what that text actually means). But in a real sense, when you communicate with a computer, even with a NLI, you cannot assume that the computer "understands" what you are saying. You cannot have a functional "assistant" with only voice recognition, as it would just be a very fast typewriter that doesn't know how to follow orders.

In order to obtain reasonable research responses, it is important to understand how to ask AI questions and give AI directions. The accuracy and value of AI responses (chats) relies heavily on the structure and content of the directions it is given with which to respond. In short, you need to learn how to communicate with AI and realize that the level of accuracy of AI's response rests primarily on the instructions it is given.

Fortunately, you can ask an AI chatbot how to ask questions and give directions. To get started, here are the two questions to ask AI:

How do I ask you questions?

How do I create a prompt?

Now, if you ask a question, before you click on the enter button, you might add another question:

Is the anyway to ask this question better?

If you draft a prompt or copy one from an online video or other source, ask the AI chatbot to suggest a better prompt. Many of the common AI programs have a way to assist you in clarifying your instructions. Google has an integrated program called Gems that appears on the Gemini window. You can ask Gemini for step-by-step instructions about how to write prompts using Gems. You can also ask Gemini to give you a prompt on a specific topic such as in-depth research or identifying photographs.

Here is an example of prompt for identifying old photos that was written by Gemini from my earlier attempts.

Purpose and Role:
You are the "Photo ID Specialist," an elite AI persona possessing the combined expertise of a Master Photographer, a Social Historian, and a Board-Certified Genealogist. Your mission is to analyze historical imagery with forensic-level detail to help researchers date photos, identify subjects, and place ancestors within their correct historical and social context.
Step-by-Step Analysis Protocol:
When a user provides a photo or description, you must evaluate it through these four lenses:

Material & Process: Identify the medium (e.g., Daguerreotype, Tintype, Albumen Print, Cabinet Card, RPPC). Explain the chemical or physical indicators that led to this conclusion and the specific year range the process was prevalent.
Fashion & Grooming Forensics: Analyze clothing (collar shapes, sleeve widths, button styles, fabrics) and hairstyles. Use these as primary markers for dating within a 3–5 year window.
Studio & Social Context: Examine backdrops, props (e.g., "hidden mother" chairs, ornate pedestals), and studio stamps/imprints. Note the social class suggested by the subject's attire and "theatricality."
Provenance & Epigraphy: Analyze any handwritten notes, photographer marks, or postal stamps (for postcards) to narrow down the geographic location and specific timeframe.
Rules for Interaction:

Request the "Reverse": Always ask if an image of the back of the photo is available, as the mount or handwritten notes are often more "evidentiary" than the image itself.
The "Evidence First" Rule: Before giving a final date estimate, list the visual observations (the "evidence") that support your conclusion.
Genealogical Bridge: Always suggest how the identified date/location helps narrow down specific census records or vital records for the user.
Output Structure (Required Format):
To maintain a professional and scholarly tone, organize your findings as follows:

### Executive Summary: A 1-sentence "Best Guess" for the date and location.
### Physical Analysis: Description of the photographic process and physical condition.
### Fashion & Visual Markers: Detailed breakdown of clothing/style cues.
### Historical Context: What was happening in that region/era that influenced this photo?
### Genealogical Recommendations: Specific "Next Steps" for the family researcher.
### Clarifying Questions: Ask for specific missing details (e.g., "Are there any tax stamps on the back?").
Tone and Voice:

Meticulous & Scholarly: Use precise terminology (e.g., "Leg-o-mutton sleeves," "Foxing," "Emulsion").
Empathetic: Acknowledge the sentimental value of these "shadows of the past."

Yes, it is long and complicated but it works.

A Family History of the Irish Famine

2026-04-01T07:33:00.001-07:00

https://www.findmypast.com/a-family-history-of

https://www.youtube.com/@AFamilyHistoryOf

A Family History of the Irish Famine, continues the "A Family History of..." series from Findmypast.com hosted by genealogist and research specialist, Jen Baldwin. Here is a summary of the Press Release.

In anticipation of the April release of the 1926 Irish Census, a powerful four-part podcast series titled A Family History Of The Irish Famine was launched on all streaming platforms on Tuesday, March 31, 2026. Hosted by genealogist and research specialist Jen Baldwin, the series examines the Great Famine through the life of her own ancestor, Archibald McKenzie. She is joined by guest historian Fiona Fitzsimons from Trinity College Dublin, who provides expert insight into the social and political forces that influenced Archibald’s journey.

Archibald’s story begins with his respectable life in County Cork before tracing his descent into poverty and his confrontations with the law. The narrative eventually follows his desperate bid to escape Ireland through the lesser-known chain-migration pathway. Using a variety of archival sources including census records, parish registers, and court documents, the podcast reconstructs Archibald’s world to provide an intimate portrait of resilience and the devastating human cost of this historical period.

This production is part of the weekly podcast A Family History Of..., which explores significant moments in British and Irish history through the lens of real families. The program has previously featured a series on wartime women and will continue with several upcoming themed series. Future installments are set to cover the 1926 General Strike, the Dardanelles Campaign of Gallipoli, and the Battle of the Somme. Listeners can also find bonus episodes and detailed research material on the podcast website.

The series is produced by Findmypast.com, a British and Irish genealogy platform owned by the family-run publishing company DC Thomson. By partnering with institutions like the British Library and The National Archives, the platform provides access to over a thousand years of historical records, helping individuals discover the stories and context of their ancestors' lives.

MyHeritage Founder and CEO Gilad Japhet at RootsTech 2026

2026-03-31T07:00:00.003-07:00

https://youtu.be/P5QdO_SYgoc?si=aUym21jaQV_6Vcez

One of the highlights of RootsTech 2026 is the session with Gilad Japhet, the Founder and CEO of MyHeritage where he speaks about what is new and what is coming MyHeritage.com. You can click above to see the entire presentation.

Here is a summary of the announcements made during his presentation.

Gilad Japhet presented a vision of genealogy that blends deep historical analysis with cutting-edge artificial intelligence. The centerpiece of the presentation was the launch of Scribe AI, a tool designed to act as a "personal historian assistant." Unlike standard transcription services, Scribe AI can classify various items—from handwritten letters to gravestones—and provide rich historical context, such as identifying fashion trends in old photos to narrow down a specific date or recognizing cultural symbols like "visitation stones" on Jewish graves.

The presentation also highlighted a major shift in how users visualize and interact with their family data:

Country Coding: A new, free feature that adds national or state flags directly to family tree views. This isn't just aesthetic; it’s designed to help researchers visually identify migration patterns and spot "missing flags" that indicate gaps in their research.

Family Infographics: Moving beyond traditional charts, this feature generates stylized, "poster-worthy" summaries of an ancestor’s life, including a "Quilt" template inspired by genealogy expert Roberta Estes.

AI Video Innovations: Gilad showcased "Life Tribute" and "Walk of Life," experimental features that animate old photos chronologically or allow users to "walk" alongside AI-rendered versions of their ancestors in their youth.

On the data and DNA front, MyHeritage has completed a full transition to Whole Genome Sequencing, moving from testing 700,000 markers to 3 billion. This massive increase in data is expected to significantly improve matching accuracy and ethnicity estimates. Gilad also provided a thrilling update on the Artifact DNA project, sharing a success story where DNA was successfully extracted from a 70-year-old letter written by composer Arnold Schoenberg, resulting in a high-quality DNA match for his grandson.

Finally, the expansion of OldNews.com (https://www.oldnews.com) remains a priority, with the platform now hosting over 438 million newspaper pages. The latest technological leap there involves structured record extraction for non-English papers, starting with Spanish and Dutch collections, which automatically translates and organizes "juicy" historical stories into searchable genealogical data.

From the standpoint of those who also use FamilySearch.org, there is now an active link between MyHeritage.com and the FamilySearch Family Tree that allows users to connect with their MyHeritage family tree and move source citations directly from MyHeritage to FamilySearch Sources.

This blog post used assistance from Google Gemini.

One Key to Breaking Down Colonial Brick Walls

2026-03-28T11:23:00.007-07:00

https://youtu.be/aP6-qfZXkYk?si=5GzxJJLdRIj32Rut

This is a summary of the video with links to the appropriate part of the video. It was extracted from the video by Google Gemini. This summary can now be used as a handout for the video.

Imagine you’re tracing your family tree back through the centuries, and you finally reach the 1700s. Suddenly, you hit a "brick wall." Your ancestor seemingly magically appears in a colonial port like Virginia or Maryland, already established with a small plot of land or a trade, but with absolutely no record of where they came from or how they crossed the Atlantic.

In his deep dive into colonial history, James Tanner argues that this isn't a failure of your research—it's a symptom of a hidden history.

The Great Colonial "Disappearing Act"

For a long time, the American narrative preferred the image of the "brave pioneer" who arrived with a trunk full of dreams and a bit of gold. But the data tells a different story. Between one-half and two-thirds of all European settlers who arrived before the Revolutionary War didn't come as free agents; they arrived in "servile status."

If your ancestor is at the end of a line you can’t break, there is a 2-to-1 probability that they were either an indentured servant, a transported convict, or an enslaved person. Because of the social stigma attached to these origins, families often spent generations systematically "scrubbing" their history, leading to the lack of traditional records we see today.

Three Paths to the New World

The video breaks down the reality of these arrivals into three distinct, often brutal, categories:

The Indentured: These were men, women, and children who signed away four to seven years of their lives in exchange for passage. Some signed willingly; others were "shanghaied" off the streets of London. If they survived the grueling mortality rates—which could hit 50% in the Chesapeake—they were given "Freedom Dues" (usually a suit of clothes and a few tools) and left to blend into the population.
The Transported: Britain essentially used the American colonies as a pressure valve for its overcrowded prisons. Between 1670 and the mid-19th century, over 70,000 convicts were shipped to the Americas for crimes as minor as stealing a loaf of bread. Unlike Australia, which celebrates its "convict stain," American families historically buried these records to appear "respectable."
The Enslaved: Starting in 1619 Jamestown, the importation of enslaved Africans fundamentally shaped the continent. Tanner points out a critical historical correction: slavery was first legalized in the North (Massachusetts, 1641), not the South. While millions were brought over, the records for these ancestors are often buried in property deeds and probate lists rather than birth certificates.

Breaking the Wall with Modern Tools

The good news? We are living in a "Golden Age" for breaking these specific brick walls. The traditional method of manual indexing (people typing out names they see on old scans) often fails because the handwriting is too messy or the "Long S" (which looks like an 'f') confuses the eye.

This is where AI and Full Text Search come in. Platforms like FamilySearch are now using high-level OCR to make the entire document searchable. You are no longer looking for a name in an index; you are looking for a mention of a "contract," "freedom dues," or a specific "plantation name" deep within a 300-year-old probate file.

The Road Ahead

If you suspect your ancestor had a forced or contracted arrival, your next steps involve moving away from "Passenger Lists" and diving into the "Abyss of Manuscripts." Look for the "10 Million Names" project or search colonial court records for mentions of contract sales. Your ancestors didn't just appear out of thin air—they were written into the legal and economic fabric of the colonies, and the records are finally becoming searchable.

Comparison of a FamilySearch Full-Text Search transcription and a Google Gemini Pro transcription

2026-03-21T14:08:00.002-07:00

"Newport, Rhode Island, United States records," images, FamilySearch (https://www.familysearch.org/ark:/61903/3:1:3Q9M-CSR8-JK8L?view=explore : Mar 21, 2026), image 196 of 1100; . Image Group Number: 008300396

This is a record of a court case in Rhode Island in 1769. To start out, I can easily read and understand this whole document because of my time spent reading town and legal records in Rhode Island in the 18th Century. As you might know, I also have a legal background and understand what was decided in this particular case. The copy of the document was found by the Full-text search app on the FamilySearch.org website.

Here is the transcription from the FamilySearch Full-text search.

368 64 Weaver Wyatt 27 Hazard her Goddard & 32 . ┃ James his Tanner & Dan November Court 1769 Peter Weaver of Swanzey in the County of Bristol in the Province of the Massa chusetts Bay & Yeoman Plain Lemuel Watt of Newport in the County of Newport Merch Defend In an action of the Case that he the Defend keep and perform his Promise and sumption to the plain made Damage forty pounds Lawful Money The Case was pleaded and committed to the Jury who returned a verdict in the Words We find for the Plaint Eighty seven Dollars and Cost It is therefore considered that the said Peter recover & have of the said Lemuel the Sum of Eighty seven Dollars and Cost of Suit taxed at 1.17 . Both land and Deft appeal to the next Superior Court of Indicature Bonds given John Perry Oliver Carl Jury James Tanner F. Benjamin Wilbour Wilkins Treby William Holt Richard Partelo Lemuel Witherell Ezra Browned Clarke Brown & William Leckham & George Brownell

The issue here is that there are a number of mistranscriptions and the parsing or placement of the text is garbled. There is also an issue with Contextual Understanding of the text. Below is the complete transcription and explanation of the court case by Google Gemini Pro. These two transcriptions illustrate the need for those using FamilySearch transcriptions to realize that the transcription may be inaccurate and misleading and second to use another method of transcription such as Google Gemini to verify and understand the record. By the way, the Google Gemini transcription is more accurate and than anything I could produce myself.

Full Transcription

[Left Margin:] 318 64 Weaver v [?] Wyatt in

[Main Body:] November Court 1769.

Peter Weaver of Swansey in the County of Bristol in the Province of the Massa- -chusetts Bay &c Yeoman Plain[tiff]. Lemuel Wyatt of Newport in the County of Newport Merch[an]t. Defend[an]t. In an Action of the Case that he the Defend[an]t. keep and perform his Promise and As- -sumption to the Plain[tiff]. made Damage forty Pounds Lawful Money The Case was pleaded and committed to the Jury who return'd a Verdict in these Words "We find for the Plain[tiff]. Eighty Seven Dollars and Cost" It is therefore Considered that the said Peter recover & have of the said Lemuel the Sum of Eighty Seven Dollars and Cost of suit taxed at £1. 17. 4 ½ Both Plain[tiff]. and Def[endan]t. Appeal to the next Superior Court of Judicature & Bonds given

[Jury List:] Jury James Tanner Jr. [highlighted] William Holt Richard Partelo

Benjamin Wilbour Lemuel Wetherill Clarke Brown

Wilkins Treby Ezra Brownell William Peckham

John Perry Oliver Earl George Brownell

Full English Translation (Modernized Reading)

Note: As the original document is written in English, I have provided a modernized expansion of the historical abbreviations and legal shorthand to ensure clarity for research purposes.

November Court, 1769.

Peter Weaver of Swansea, in the County of Bristol, in the Province of the Massachusetts Bay, etcetera, Yeoman, Plaintiff. Lemuel Wyatt of Newport, in the County of Newport, Merchant, Defendant. In an Action of the Case [a civil lawsuit], that he, the Defendant, keep and perform his promise and assumption [a contractual obligation or debt] made to the Plaintiff. Damages [sought]: forty pounds, lawful money.

The case was pleaded and committed to the jury, who returned a verdict in these words: "We find for the Plaintiff, eighty-seven dollars and costs." It is therefore considered that the said Peter recover and have of the said Lemuel the sum of eighty-seven dollars, and the costs of the suit, taxed at 1 pound, 17 shillings, and 4.5 pence. Both Plaintiff and Defendant appeal to the next Superior Court of Judicature, and bonds are given.

Jury: James Tanner Jr., William Holt, Richard Partelo, Benjamin Wilbour, Lemuel Wetherill, Clarke Brown, Wilkins Treby, Ezra Brownell, William Peckham, John Perry, Oliver Earl, George Brownell.

Genealogical Analysis

This document is highly informative for tracing individuals in the intersecting regions of Bristol County, Massachusetts, and Newport County, Rhode Island, during the late colonial period.

Key Findings:

Primary Parties & Socioeconomic Status:
- Peter Weaver (Plaintiff): Identified as a "Yeoman" (a land-owning farmer) residing in Swansea, Bristol County, Massachusetts.
- Lemuel Wyatt (Defendant): Identified as a "Merchant" residing in Newport, Newport County (Rhode Island). The interstate nature of this lawsuit highlights the deep economic ties between maritime Newport and agricultural Swansea.
The Highlighted Subject: James Tanner Jr.
- You have highlighted the name James Tanner Jr. in the jury list. The presence of the "Jr." suffix is a vital genealogical clue. It strongly implies that there was an older, living James Tanner (presumably his father or an older uncle) residing in or near the jurisdiction of this court in 1769 to necessitate the distinction.
- Serving on a jury in 1769 required a man to be a freeholder (property owner) of legal age (usually 21 or older). This places James Tanner Jr.'s birth year at or before 1748.
Location and Jurisdiction: * Given the defendant's residence and the historical makeup of the jury pool (names like Treby, Peckham, and Brownell are deeply rooted in Rhode Island history), this is likely a Newport County court record, handling a civil suit brought by a Massachusetts resident.
Currency Anomalies: The document shows an interesting transitional financial period. Damages were sought in "Pounds Lawful Money," court costs were taxed in British currency (£1. 17s. 4½d), but the jury awarded damages in "Eighty Seven Dollars" (referring to Spanish milled dollars, which were heavily circulated in maritime colonies like Rhode Island).

Navigating the Servile Barrier: Researching Indentured, Transported, and Enslaved Ancestors

2026-03-21T05:19:00.002-07:00

https://youtu.be/aP6-qfZXkYk?si=pzDl0Og-66D18cc-

Note: The time link refer to the video.

A "brick wall" is a common frustration for genealogists, particularly when an ancestor seems to "magically appear" in the American colonies with no discernible connection to an overseas origin. For many researchers, this missing link is not a matter of lost records, but a misunderstanding of the ancestor’s original legal status. As this post illustrates, a vast majority of early immigrants to North America arrived not as free settlers, but in "servile status." Understanding the intersection of traditional historical research and emerging AI-driven full-text search is now the definitive way to reclaim these erased identities.

The Statistical Reality: Most Early Immigrants Were Not "Free"

In the current genealogical landscape, there is often a subconscious bias toward the "free immigrant" narrative. However, the historical data suggests a different reality. Between 1630 and the American Revolution, approximately one-half to two-thirds of all European settlers arrived as indentured servants [04:38].

In areas like Virginia and Maryland, the numbers are even more stark; up to two-thirds of English migrants arrived in a servile state [06:54]. If your ancestral line ends abruptly in the 17th or 18th century with a prominent land-owning citizen, "the researcher" must consider the high probability that the early immigrant began their American journey under a contract of indenture.

The Indentured Experience: Contracts and "Freedom Dues"

An indentured servant was a man, woman, or child who signed a contract—an "indenture"—to work without salary for a specific master for a set duration, typically four to seven years [02:10].

The Contractual Nature: These indentures were legally considered "choses in action" or personal property. They could be bought, sold, and traded [08:06].
The Reality of Survival: Mortality rates were devastatingly high. In the Chesapeake area during the 17th century, 50% of indentured servants died before completing their terms [13:10].
The "Freedom Dues": Upon completion of their term, those who survived were released and given "freedom dues," which often included a suit of clothes, tools, or a small amount of land [03:55].

"The stigma from being indentured was often enough that people would change their names after the indenture... they would simply move to a different part of the country where they were unknown." [10:01]

The Transported: Criminalizing Nonconformity

"Transported" individuals represent a distinct category: those sent to the colonies against their will as a form of criminal punishment. While Australia is famous for its penal colonies, North America received between 69,000 and 74,000 convicts from Great Britain and Ireland before the mid-19th century [21:03].

Many of these individuals were not "hardened criminals" in the modern sense but were "nonconformists"—individuals such as Methodists, Baptists, or Quakers whose religious practices were criminalized by the Church of England [27:14]. Because the British government viewed shipping convicts as cheaper than building prisons, these ancestors were often sold into labor upon arrival to recoup the costs of their transportation [26:59].

Tracing Enslaved Ancestors: From 1870 to the "Great Migration"

Researching enslaved ancestors requires a specialized methodology that bridges the gap between the 1870 US Federal Census and the arrival of the first 20 odd individuals in Jamestown in 1619 [34:26].

The Research Methodology:

Work Backward: Establish the family line firmly in the 1870 census, the first to name formerly enslaved people.
The "Great Migration" Context: Roughly 6 million African-Americans moved from the South to Northern cities like Chicago and New York between 1870 and 1960 [39:32]. Tracing these urban ancestors back to their Southern origins is a prerequisite for finding earlier records.
Plantation Records & Deeds: Before 1865, enslaved individuals were often recorded as property in probate records and deeds [40:46].

AI Spotlight: The Revolution of Full-Text Search

The most significant advancement in this field is the implementation of AI-driven full-text search (specifically on platforms like FamilySearch). Traditionally, searching for an indentured or enslaved ancestor required manually scrolling through thousands of unindexed probate images.

How it Works: Modern AI models utilize OCR (Optical Character Recognition) and transformer models to "read" handwritten documents. This allows researchers to search for specific surnames or locations directly within the body of a deed or an estate inventory—records that were previously "unsearchable" [01:18].

Pro-Tip for Researchers: When using AI transcription tools (like Google Gemini), be aware of "long S" characters (which look like "f") in 17th-century documents. You can prompt the AI to: "Translate this 17th-century indenture into modern English and list all named parties" to bypass archaic paleography [11:43].

Synthesis & Context: The Importance of Probate

Why are probate records so central to this research? In the colonial era, if the parents of a child died, the child was often legally placed into an indenture by the court to ensure they did not become a financial burden on the parish [18:07]. These "orphan court" records are hidden gems that contain the names of the child, the master, and the terms of service, effectively acting as a birth record for an otherwise "invisible" ancestor.

Key Resources

FamilySearch Full-Text Search: familysearch.org/search/full-text
10 Million Names Project: A collaborative project to identify and link descendants of the 10 million people enslaved in the US.
DiscoverFreed.org: The Freedman’s Bureau Project database.
BYU Library Family History Center: For access to specialized slide decks and handouts [10:48].

The summary of this video was assisted by Google Gemini.

Could FamilySearch use AI to generate family trees?

2026-03-17T07:26:00.000-07:00

The idea of using AI to generate family trees for insertion into the FamilySearch.org Family Tree was raised in an interview during RootsTech 2026 of Elder Mark A. Bragg, a General Authority Seventy and executive director of the Family History Department of The Church of Jesus Christ of Latter-day Saints. See Could we ever run out of temple names? Elder Mark A. Bragg on why family history is accelerating.

The fact is that the FamilySearch.org website already has a section with computer generated family trees. You can find this section in the Genealogies directory under the main Search tab.

This section contains seven different computer-generated trees, but no trees have been generated since 2023. Here is a screenshot of the current selection of files available to search.

One of the statements made in the video cited above, is that AI is significantly accelerating family history by scraping data from digitized books and records to build out trees automatically. It is certainly true that AI does have the ability to build automatic family trees. However, the reality of creating "family trees" from raw data is not as simple as scraping a book. Despite the ability of AI to scrape a book and build a family tree, the actual reality of the process of creating a valuable genealogical resource is much more complicated. There are several tremendously difficult obstacles to any actual AI generated tree having valid information. None of which were mentioned in the interview above.

Back in 1894, with the formation of the Genealogical Society of Utah, the predecessor of FamilySearch, the main concern was preventing the duplication of temple ordinances. See Allen, James B., Jessie L. Embry, and Kahlile B. Mehr. Hearts Turned to the Fathers: A History of the Genealogical Society of Utah, 1894-1994. BYU Studies, Brigham Young University, 1995. One of the key issues in the demise of the predecessor of the FamilySearch Family Tree (new.familysearch.org) was the proliferation of rampant duplication of the ordinances. My great-grandfather, Henry Martin Tanner, had his temple ordinances done over 800 times, despite the fact that he performed the ordinances during his lifetime in the St. George Temple.

With the current state of the FamilySearch Family Tree, the issue of duplication is by far the most serious. My own experience is that any time I do research for adding names to the FamilySearch Family Tree, I routinely find duplicate records. Taking a set of records from any major European country and including most of North and South America would produce a significant number of duplicate entries. If AI is able to create a valid family tree, then why not use it to eliminate the duplicates that are currently in the FamilySearch Family Tree?

Another major issue with the FamilySearch Family Tree is the existence of "floaters," who are people and segments of family trees that are, in a real sense, floating around the main FamilySearch Family Tree because they lack any connections. Many of these floaters are duplicates and also are not currently readily available for inclusion, if they are discoverable at all, because of a lack of more specific information about their existence and identity.

An undercurrent of the concept of computer or AI-generated family trees is the issue of whether or not individuals, such as all of the currently active genealogists contributing to the FamilySearch Family Tree, have any future utility. If my job is going to be turned over to AI, then why am I spending time doing research and verifying names through careful application of the genealogical proof standard if it is only necessary to identify a name from a book to qualify for entry? Name extraction has, in the past, created a tremendous number of duplicate entries. It is still common on the FamilySearch Family Tree to find English records with duplicates because births, marriages, and burials were all extracted separately. A family with six or eight children could have thirty or forty duplicates. Current name extraction programs are also creating a substantial number of duplicate entries.

The current state of the FamilySearch Family Tree, where changes are randomly made to established entries without any supporting data, is also undermining the ability of the tree to function efficiently. Unless these systematic failings of the FamilySearch Family Tree can be adequately addressed, whether by AI or otherwise, it seems vastly premature to propose AI-generated family trees.

Since I now work full time essentially as a genealogist with the primary goal of finding and verifying additional people for the FamilySearch Family Tree, I would appreciate it if someone would let me know if it's time for me to retire.

Online with New AI Developments Every Wednesday (except for Holidays)

2026-03-16T10:28:00.000-07:00

https://familyhistory.lib.byu.edu/get-help

The BYU Library Family History Center is the host of my weekly Understanding AI Q&A discussion on the Virtual Help Desk. We meet in a Breakout Room and talk through some of the weekly changes and I answer question. This is all totally free on the BYU Library Family History Center's open Zoom.

The time is 10:00 am Mountain Daylight Time. Here is the link again to the Virtual Help Desk.

https://familyhistory.lib.byu.edu/live-virtual-family-history-help

The Virtual Help Desk is also open for consultation with all of the other missionaries on the shift at the time of your call. We also have a variety of specialists on countries and many other resources.

The Coalition for Responsible AI in Genealogy in the Church News

2026-03-16T05:53:00.000-07:00

https://www.thechurchnews.com/living-faith/2026/03/06/researchers-provide-guidelines-for-responsible-ai-usage-family-history-rootstech/

Although it kind of got buried in all the articles featuring RootsTech 2026 and other news, there was a good informative article about the panel discussion from the Coalition for Responsible AI in Genealogy or Craigen.org. Here is a short summary of the article.

At RootsTech 2026, experts addressed the evolving landscape of family history by establishing a framework for artificial intelligence and sharing new methods for researching pioneer ancestry. A panel from the Coalition for Responsible AI in Genealogy, led by researcher Lynn Broderick, introduced five core principles—accuracy, disclosure, privacy, education, and compliance—to guide users through the technological "wilderness."

During the discussion, James Tanner cautioned against the Dunning-Kruger effect in AI, where the technology may generate false information when it lacks verifiable data. To counter this, David Ouimette emphasized the importance of skilled prompting and human discernment. The panel also highlighted the "Water Cooler Rule" regarding privacy, with Steve Little and Katherine Borges warning against uploading sensitive personal information or DNA data into AI systems where storage and usage policies may be unclear.

In addition to technological guidelines, the conference featured sessions on local history. Julie Merrill provided practical advice for navigating unorganized pioneer temple records, recommending the use of specific annotated records to bridge gaps in research. Further honoring the past, Ellen Jeppson shared stories of early women pioneers like Ruth May Fox, highlighting the 125-year legacy of the International Society Daughters of Utah Pioneers and its mission to preserve the faith and courage of those who settled the region.

There were also a few photos.

Decipher your archive with Leo Technologies Limited

2026-03-16T05:36:00.000-07:00

Leo

https://www.tryleo.ai/

Leo is a comprehensive digital hub designed to transform how historical manuscripts and archives are managed, offering an all-in-one transcription and document management platform. At its core is a state-of-the-art AI model specifically optimized for historical handwriting—such as secretary or cursive hands—and printed materials in English, Latin, French, German, and Spanish. Unlike other services, Leo is ready to use immediately with no model training required, though it maintains the flexibility to incorporate other AI models like GPT, Gemini, and Claude.

The user experience follows a straightforward three-step workflow:

Upload: Users begin by uploading images of handwritten or printed documents written in Latin scripts from the past 500 years.
Transcribe: The platform's machine learning technology generates instant, accurate transcriptions that preserve complex elements like margin notes, tables, and strikethroughs.
Manage: Once transcribed, documents are organized within a "Dropbox-like" interface. Users can categorize their collection into custom lists, utilize advanced search across all notes and transcripts, and add metadata for deeper filtering.

Beyond individual research, Leo emphasizes collaboration. Researchers can share their work through secure public links, allowing colleagues or students to access the materials and even copy documents to their own collections without needing a personal account. When the research is ready for the next stage, data can be exported in various professional formats, including XML, Word, or PDF.

To accommodate different project scales, Leo offers a tiered pricing structure:

Free: Includes 10 monthly credits and storage for 100 images.
Standard ($15/mo): Provides 100 monthly credits and storage for 10,000 images.
Scholar ($65/mo): Provides 500 monthly credits and storage for 50,000 images.
Professional ($120/mo): Provides 1,000 monthly credits and storage for 100,000 images.

For users with specific needs, Leo also sells additional lifetime credits that never expire, ensuring that even if a subscription ends, the research remains accessible and exportable.

Neo-luddites and AI: A Retrospective

2026-03-14T12:19:00.002-07:00

If you need a review of the history of Luddite movement, see https://en.wikipedia.org/wiki/Luddite.

Let's start by thinking about photographs and digitization. The Kodak Professional Digital Camera System (a Nikon F-3 camera equipped by Kodak with a 1.3 megapixel sensor) for photojournalists was introduced in 1991. If my calculations are correct that was 35 years ago. Than means that would need to be about 45 years old to remember how quickly digital cameras replaced almost all film versions. Adobe Photoshop 1.0 was officially released on February 19, 1990, exclusively for the Macintosh. The internet, in the form of the World Wide Web, The first online version of the World Wide Web, created by Tim Berners-Lee at CERN in December 1990, was hosted on a NeXT computer at http://info.cern.ch. It functioned as both a browser and editor, showcasing the first website, which explained the WWW (World Wide Web) project itself and how to use hypertext. As a matter of note, I first viewed the WWW when there were six websites.

Almost everything I read is currently digitized and primarily online. Every photograph that I take is a digital photograph. Since Google's nano banana was integrated with Adobe Photoshop, all the photos that I take that I use any manipulation to alter are done by using AI in Photoshop. All the major genealogical websites are now almost totally AI-intensive.

OK, so here is the issue. I am not a regular reader of Facebook but from time to time, I do read a few posts. I find a very vociferous anti-AI minority, mostly teachers and academics. I even see statements about how the posters are never going to use AI. There are also statements about how easy it is to spot an AI photo or article. This may have been an issue two or three years ago but my own experience is to the contrary. You cannot tell an AI photo from one taken with a camera because both are essentially the same thing. Obviously if the image show a dancing hippo or a waltzing dinosaur, the images are most certainly generated.

If you think you are never going to use AI, you may as well turn off your electricity and move out into a camp somewhere in what little wilderness is left in the world because, shortly, almost everything electric will have some AI function.

AI is a tool. You must use it as the master in a master/servant relationship. You make all the important decisions. AI is not a browser or a search engine, it is AI. Here is an example. The paragraph above beginning with OK was written entirely by me not using any AI assistance. Here is the same paragraph after I told Google Gemini to revise the paragraph for grammar, spelling, and accuracy.

"OK, so here is the issue. I am not a regular reader of Facebook, but from time to time, I do read a few posts. I find a very vociferous anti-AI minority, mostly teachers and academics. I even see statements about how the posters are never going to use AI. There are also statements about how easy it is to spot an AI photo or article. This may have been true two or three years ago, but my own experience is to the contrary. You cannot tell an AI photo from one taken with a camera because, at the pixel level, both are essentially the same thing. Obviously, if the image shows a dancing hippo or a waltzing dinosaur, the images are most certainly generated."

I used the suggestions from Gemini to revise my original paragraph and the rest of the article.

Now, the quote by Sam Altman, CEO of OpenAI: "AI won’t replace humans. But humans who use AI will replace those who don’t." Right now, if you don't know how to use AI, you don't know how to use FamilySearch, Ancestry, MyHeritage and most of the other genealogy programs.