There is something fitting about an article on artificial intelligence (AI) tools in genealogy that was itself largely written with AI assistance. The series of articles you have been reading on this blog — covering Y-DNA testing, autosomal DNA, whole genome sequencing, genealogy software, and more — was researched and drafted using Claude, an AI assistant developed by Anthropic, with human review, editorial judgment, and fact-checking throughout. That experience informs this article directly: AI tools are not a future development in genealogy. They are here now, actively reshaping how research is conducted, and serious genealogists who understand what these tools can and cannot do will be better equipped to use them effectively.
This article surveys the current AI tool landscape for genealogists — what the major platforms are doing with AI, which specialized tools deserve attention, and how large language models (LLMs) fit into a serious research workflow. We also address the limitations and risks honestly, because AI tools used carelessly can introduce errors into genealogical research just as easily as they can accelerate it.
A Brief History: From Hints to Generative AI
AI has been embedded in genealogy platforms longer than most researchers realize. The hint systems that AncestryDNA, FamilySearch, and MyHeritage use to suggest potentially matching records and family trees are driven by machine learning algorithms that compare your tree’s data against billions of records and millions of other users’ trees. When Ancestry’s ThruLines tool hypothesizes a path of relationship between you and a DNA match, or when MyHeritage’s SmartMatch surfaces a profile from another user’s tree that may represent the same person as someone in yours, those are AI systems at work — not simple database lookups.
The more visible AI revolution in genealogy began in 2021 when MyHeritage launched Deep Nostalgia[1], bringing ancestor photos to life through AI animation — a feature that attracted widespread attention well beyond the genealogy community. By 2023, handwriting recognition technology from companies like Transkribus reached accuracy levels that made mass transcription of historical documents feasible. The landscape shifted again in 2024–2025 with the widespread adoption of large language models. By 2026, the question is no longer whether AI belongs in genealogical research — it is which tools to use, for which tasks, and with what level of scrutiny.
AI Built Into the Platforms You Already Use
Ancestry has integrated AI at multiple levels of its platform. The AI Research Assistant provides personalized research suggestions based on patterns and gaps in your tree — identifying which ancestors lack death records, which locations have unlinked records available, and which DNA matches have not yet been explored. Ancestry.com has introduced a new beta tool for transcribing records stored in user galleries — Image Transcript — which uses AI optical character recognition (OCR) to convert scanned documents directly into searchable, editable text. AncestryDNA’s SideView feature uses AI to separate a tester’s DNA into contributions from each parent, even without a parent having tested — a significant analytical step made possible by machine learning applied to population genetics data.
FamilySearch has invested heavily in AI-driven record accessibility. AI and handwriting recognition now let users search nearly 2 billion genealogically significant records that had previously been available only as browse-only images — meaning records that once required page-by-page visual scanning can now be searched by name. More searchable image collections are being added regularly to Full-Text Search, and more handwritten languages[2] are planned. FamilySearch also operates an AI Help chatbot and an AI Research Assistant within its platform, and its volunteer indexing system — which turns scanned records into searchable text — now uses AI to perform an initial pass before human volunteers review and correct the results, dramatically accelerating the process.
MyHeritage has arguably the most extensive suite of AI features of any genealogy platform. The AI Biographer generates detailed narrative life stories for ancestors based on the facts in your family tree, incorporating historical context to place each ancestor’s life in the world they actually inhabited. MyHeritage expanded their AI Biographer tool in 2025 to generate more detailed narratives using historical context databases. The tool now incorporates social history information, helping researchers understand the world their ancestors lived in. The Photo Enhancer restores clarity to damaged or degraded images; the Colorization tool converts black-and-white photographs to color; and Deep Nostalgia animates still portraits. These visual tools have introduced many families to their ancestors in genuinely emotional ways, and they have brought MyHeritage significant recognition beyond the core research community.
RootsMagic 11 takes a thoughtfully different approach with its AI Prompt Builder, covered in detail in our RootsMagic article. Rather than generating AI content directly within the application, RootsMagic reads the documented facts in your database and constructs a precisely engineered prompt that you can paste into any external AI tool — Claude, ChatGPT, or others — to generate a biography, research plan, or narrative grounded in your actual data. This approach keeps the researcher in control and avoids the AI fabrication problem that plagues less careful implementations.
🤖 Large Language Models as Research Assistants
General-purpose large language models — including Claude (Anthropic), ChatGPT (OpenAI), and Gemini (Google) — are not genealogical databases. They do not know what happened to your specific ancestor, and asking them to generate specific genealogical facts is one of the fastest ways to introduce fabricated information into your research. This is a critical point that cannot be overstated: LLMs are trained to produce fluent, plausible-sounding text, and they will generate specific-sounding details — dates, names, places — that have no basis in reality if asked to fill in gaps they don’t actually know the answer to.
Used correctly, however, LLMs are genuinely powerful genealogical research assistants. Their appropriate uses include: explaining research methodology (what records would document a German immigrant in 1880s Indiana?); interpreting legal or Latin terminology in historical documents; suggesting research strategies for specific geographic regions or time periods; translating foreign-language records; helping draft research reports, family histories, or correspondence with archives; analyzing the logic of a research problem and suggesting lines of inquiry; and reviewing a draft research conclusion for logical consistency. The rule is simple: use LLMs for strategy, interpretation, and writing — not for generating facts about specific individuals. Using an AI assistant as a knowledgeable research advisor — not a data source — significantly accelerates the research process for beginners and experienced researchers alike.
This is precisely how we have used Claude in developing the DFA blog series — as a research strategist, writer, and editor, working from factual sources that were independently verified, not as a generator of genealogical facts.
Specialized AI Tools Worth Knowing
Transkribus (transkribus.org) — Transkribus remains the gold standard for transcribing[3] historical handwriting. The free tier lets you upload pages and apply trained AI models for Gaelic, English, German, Latin, and more. Genealogists can decode old parish records, wills, and letters far faster than manual transcription. Transkribus uses a technology called Handwritten Text Recognition (HTR), which is fundamentally different from standard OCR — it can be trained on specific handwriting styles and historical scripts, making it significantly more accurate than general-purpose text recognition tools for difficult documents. For Dillman Family Association (DFA) members working with German church records, Swiss civil registers, or 18th-century American documents, Transkribus is the most powerful transcription tool currently available. A free tier is available; paid plans start at approximately €24 per month for higher usage volumes.
Archive Studio — A free, open-source tool that connects directly to AI model application programming interfaces (APIs) for document transcription. Archive Studio achieves high accuracy by transcribing with one large language model and correcting with another — a two-stage approach that catches errors the first model introduces. Particularly useful for researchers who want fine-grained control over the transcription process.
HandwritingOCR.com — A web-based specialized tool supporting over 300 languages and historical scripts. A practical alternative to Transkribus for researchers who need quick transcription without creating a full account or training a custom model.
Leo — An AI web application fine-tuned specifically for paleography — the study and decipherment of historical handwriting. Where general-purpose LLMs struggle with pre-modern scripts and heavily abbreviated historical documents, Leo’s specialized training makes it more effective for challenging materials.
Kindex (kindex.org) — An online platform for AI-assisted indexing of genealogical records, featuring an Auto-Index Service that can process uploaded document images and extract structured genealogical data. Particularly useful for researchers working with large collections of similar document types.
Google Lens — Often overlooked as a genealogy tool, Google Lens provides on-device AI recognition[4] that can translate foreign-language tombstone inscriptions, decipher signage in historical photographs, identify objects in old family photos, and provide instant OCR of handwritten text when pointed at a document. Available free on any smartphone. For fieldwork at cemeteries and archives, it is a practical and immediate tool.
DeepL — A translation tool widely regarded among researchers as producing higher-quality translations than Google Translate for European languages, particularly German, French, Dutch, and Scandinavian languages. Available free for shorter texts; a paid Pro version removes length limits. For DFA members working with German-language church records, civil registers, or emigration documents, DeepL is a significant improvement over general-purpose translation tools.
AI and Photo Restoration
Photo restoration and enhancement represent some of the most emotionally resonant applications of AI in genealogy, and the tools available in 2026 are genuinely impressive. MyHeritage’s Photo Enhancer and Colorization tools are the most accessible, integrated directly into a platform many researchers already use. For standalone image restoration, Topaz Photo AI and Remini offer higher levels of control and quality for researchers willing to work outside an integrated platform. Deep Nostalgia’s animation feature remains popular for sharing ancestor portraits with family members who might not otherwise engage with genealogical research — a legitimate outreach use even for researchers who prefer not to animate historical photographs for their own work.
One note of caution: colorized and animated photographs are interpretations, not records. The colors applied by AI to a black-and-white photograph are statistically derived guesses, not actual historical colors. Colorized images should always be clearly labeled as such and never substituted for the original in any research documentation.
The Terms of Service Consideration
A recent and important development: MyHeritage updated its terms and conditions, adding language restricting the use of AI agents[5], bots, automated systems, or any other artificial intelligence technology to access, scrape, crawl, harvest, or collect data from the Website or Service — including family trees, historical records, photographs, and DNA results. Ancestry has similarly updated its terms of service to address AI data harvesting. These restrictions are aimed at automated bulk data collection, not at individual researchers using AI tools to assist their personal research. However, they are worth being aware of, particularly for technically sophisticated researchers building automated workflows that interact with genealogy platform APIs.
Honest Limitations and Risks
Hallucination remains a serious risk. LLMs generate text that sounds authoritative whether or not it is accurate. In genealogical research — where a fabricated date, name, or relationship can cascade into years of misdirected work — this risk is particularly consequential. Every specific fact produced by an AI tool must be verified against primary sources before it is treated as reliable.
Transcription errors require human review. Even the best AI transcription tools — including Transkribus at its most accurate — produce errors, particularly with damaged documents, unusual scripts, or heavily abbreviated Latin. AI transcription should be treated as a first draft that requires human review, not a final product.
Ethnicity estimates and AI matching are probabilities, not proofs. The AI systems driving ethnicity estimates, SmartMatch, ThruLines, and DNA matching produce probabilistic outputs — statistically likely answers, not genealogical conclusions. Every AI-generated suggestion requires the same documentary verification as any other research hypothesis.
Copyright and data privacy considerations apply. Using AI tools to process genealogical records raises questions about copyright (some records are held by commercial companies whose terms restrict reproduction) and data privacy (uploading documents containing other living people’s information to AI tools has privacy implications). Researchers should be thoughtful about what they upload to any AI service.
What This Means for Dillman Researchers
For DFA members, the most immediately practical AI tools are those that address the specific challenges of Dillman research: German and Swiss records written in Fraktur, Kurrent or Sütterlin script (the old German handwriting styles used through the mid-20th century), Latin-language church records, and the translation of 19th-century German emigration documents. Transkribus and DeepL directly address these challenges. FamilySearch’s Full-Text Search is increasingly making German-language records searchable that were previously accessible only by browsing image by image.
For researchers contributing to the Dillman DNA Project, AI tools for DNA analysis — clustering, segment analysis, and the kind of strategic research planning that LLMs support — can accelerate the work of connecting genetic evidence to documentary research.
The tools are better than they have ever been. Used with clear-eyed understanding of both their capabilities and their limitations, they make serious genealogical research faster, more accessible, and more productive than at any previous point in the history of the field.
Tools mentioned in this article: Claude (claude.ai, free and paid tiers), ChatGPT (chatgpt.com, free and paid tiers), Transkribus (transkribus.org, free and paid tiers from €24/month), DeepL (deepl.com, free for shorter texts, Pro from $8.74/month), Archive Studio (free, open-source), Google Lens (free, available on iOS and Android). Platform AI features at Ancestry, FamilySearch, MyHeritage, and RootsMagic are included with existing subscriptions or software purchases.
Sources
- MyHeritage Blog, “Introducing Deep Nostalgia,” 2021, blog.myheritage.com (accessed May 2026).
- FamilySearch, “Full-Text Search,” familysearch.org (accessed May 2026).
- Transkribus, “About Transkribus,” transkribus.org/about (accessed May 2026).
- Google, “Google Lens,” lens.google.com (accessed May 2026).
- MyHeritage, “Terms and Conditions,” April 2026, myheritage.com/terms-and-conditions (accessed May 2026).
AI Disclosure: This article was researched and written with the assistance of Claude, an AI assistant developed by Anthropic. The irony of using AI to write an article about AI tools in genealogy is not lost on us — and is, in fact, the point. All facts have been reviewed for accuracy, but readers are encouraged to verify information independently, particularly for time-sensitive details such as pricing and feature availability.
Discover more from Dillman Family Association
Subscribe to get the latest posts sent to your email.