Best AI Tools for Genealogy Research in 2026 (Tested Across 6 Stages)

Genealogy research hasn't gotten simpler. It's gotten faster — for the parts where speed actually matters.

In 2026, AI assists at every stage: finding records, reading old handwriting, interpreting DNA matches, restoring photographs, and animating ancestor portraits. The problem isn't finding an AI tool. It's knowing which tool belongs at which stage, and how much to trust what it produces.

This is the workflow I've tested — across multiple family trees, including one with roots split between rural Brittany in the 1880s and 1930s Warsaw. Six stages. Dozens of tools. Honest verdicts on what moves research forward and what generates plausible-sounding dead ends.

What to look for in an AI genealogy tool

Before the workflow, a frame for evaluation.

Good AI genealogy tools share three qualities. They show you sources — any tool that surfaces a family connection without a linked document is a hypothesis generator, not a research tool. They're honest about confidence — the best tools flag uncertain matches rather than presenting hallucinated connections as discoveries. And they support verification — they make it easy to get to the original document, not just the AI's summary of it.

Keep those criteria in mind as new tools enter this space. The marketing is often significantly ahead of the actual capability.

One more thing: AI systems produce plausible output, not accurate output. For genealogy specifically, the cost of accepting a wrong connection is high — it propagates into family trees, gets shared across platforms, and becomes increasingly hard to correct. A plausible AI suggestion is a lead. It becomes evidence only when you've verified it against a primary source.

How AI changes the biggest challenges in genealogy

Two challenges have historically slowed genealogy research to a crawl: accessing records, and reading them.

Most genealogical evidence is recorded in documents — civil registrations, church registers, census returns, military files, notarial acts. For centuries, these existed only in physical form. Accessing them meant traveling to archives. Reading them often meant learning a historical script that no one uses anymore.

AI has changed both constraints.

Large-scale digitization and AI indexing projects — primarily at FamilySearch and Ancestry — have made hundreds of millions of previously image-only records text-searchable. AI handwriting recognition tools have made historical scripts legible without specialist training. Reading a 19th century French notarial act — dense cursive on thin paper, in legalese, with regional abbreviations — used to require local expertise or months of practice. Now you can extract key names and dates from most documents in minutes.

Neither solution is perfect. AI transcriptions contain errors. Large portions of the historical archive remain undigitized. But the gap between what was accessible in 2015 and what's accessible now is enormous.

Stage 1: Start with FamilySearch and Ancestry

Before any specialized tool, before any AI assistant — start here.

FamilySearch is free. Over a billion indexed records, strong coverage across most of the world, and an AI indexing project that has made previously image-only documents text-searchable at a pace that accelerates every year. The collaborative world tree has errors — treat any connection from another user's unsourced tree as a lead, not a fact, until you've verified it yourself. The underlying record collection, though, is exceptional and worth exhausting before paying for anything.

Ancestry charges $25–50/month depending on plan. Their value lies in scale: 40+ billion records, and AI-powered "Hints" that surface potentially relevant connections from their record database and other users' trees. The AI is a hypothesis generator — every Hint needs independent verification. Once your tree has 50+ people across multiple generations, the volume of suggestions becomes genuinely useful for discovering records you'd never have found manually.

Findmypast is worth adding for British, Irish, and Commonwealth research. Their AI transcription work on historical newspapers — in partnership with the British Newspaper Archive — makes surname searching across historical press practical in a way it wasn't before.

The practical approach: start free, exhaust what FamilySearch has, then add paid access when you've hit a wall that only a specific record collection can solve.

Stage 2: Unlock old handwriting with AI transcription

This is where AI creates the most dramatic difference for most researchers.

A substantial portion of genealogically valuable records exist only as images of handwritten documents — church registers from the 17th and 18th centuries, military conscription lists, notarial acts recording land sales and wills, census returns in regional scripts. Before AI, reading these required either specialist knowledge or considerable time learning the script style.

Transkribus handles this better than anything else available. It supports hundreds of historical scripts — German Kurrent, old French cursive, Gothic hands, Latin ecclesiastical scripts. Pre-trained models cover the most common scripts; you can also train a custom model on a specific handwriting style, which is worth doing if you're working repeatedly with documents from the same region or the same notary's hand. Credit-based pricing; the free tier covers limited monthly transcriptions. For serious archive work, it's the tool I reach for first.

For lighter needs — 20th century letters, typed documents, relatively legible census entries — Google Lens and Apple Live Text handle transcription for free. Neither handles 17th century Latin, but both are useful for more recent family documents.

A practical note: always verify the transcription against the original image. AI transcriptions of old handwriting contain errors, particularly on names. The transcription is a starting point for your reading, not a replacement for it.

Stage 3: Use AI assistants as research advisors — not data sources

This distinction prevents most of the serious mistakes people make with AI in genealogy.

Large language models — ChatGPT, Claude, Perplexity — don't know what happened to your specific ancestors. They don't have access to genealogy databases. Ask one to confirm a specific family connection and it will produce a plausible-sounding answer based on what seems likely, not on actual records. This is the hallucination problem, and it's particularly dangerous in genealogy because the outputs look credible.

What AI assistants are genuinely useful for:

Research strategy. "What records would have documented a Polish Catholic farmer born in 1887 in Galicia? What survived World War II?" A good AI assistant gives you a research roadmap based on historical knowledge. That roadmap is often excellent.

Document interpretation. "What does this Latin abbreviation mean in a 19th century baptismal record?" Or: "I have 15 DNA matches that all cluster around one region — what does that pattern suggest?" AI assistants are good research advisors at this level.

Translation. AI-powered translation for records in languages you don't read — not perfectly, but well enough to extract key names, dates, and places from most documents.

The rule: use AI assistants as a knowledgeable advisor, never as a data source. The moment you ask one to tell you something specific about your ancestors rather than helping you find and interpret real sources, you've moved from research to storytelling.

Stage 4: Analyze DNA evidence with the right tools

DNA genealogy works best as triangulation. One test alone tells you less than multiple tests from the same lineage, compared against each other.

AncestryDNA has the largest database — over 25 million people — which maximizes the statistical value of matching. Their ThruLines feature uses AI to suggest how you and a DNA match might be related, based on the family trees both of you have built. Treat ThruLines as hypothesis generation: useful for identifying where a cluster of DNA matches might converge in your tree, not as confirmed relationships.

GEDmatch accepts raw DNA uploads from any testing company. It's the platform of choice for cross-company analysis and provides chromosome browsers and admixture calculators unavailable elsewhere. The free tier covers most research needs. Uploading your DNA data to GEDmatch after testing with Ancestry or 23andMe is one of the highest-value free actions in genealogy — it dramatically expands your match pool at no additional cost.

DNA Painter helps you build a chromosome map showing which DNA segments you share with which matches. For complex cases — identifying an unknown grandparent, resolving a non-paternity event — the visual segment mapping is invaluable. Free tier available; the paid tier ($65/year) adds AI-assisted clustering.

On 23andMe vs MyHeritage for European ancestry: MyHeritage edges out 23andMe for matching European relatives, particularly Eastern European, French, and Mediterranean ancestry. If your research focus is European, MyHeritage's Theories of Family Relativity feature — which combines DNA matching with historical record connections — is worth the subscription.

The DNA genealogy tree, built in parallel with the documentary family tree, gets stronger as the biological and archival evidence converge. When a DNA match confirms a documentary connection you've independently verified in records, that convergence is solid. When DNA and documents diverge, you have a mystery worth pursuing.

Stage 5: Restore and enhance old family photographs

Physical photographs degrade. AI tools reverse some of that.

Remini is the tool I recommend first for restoration. It's particularly effective on photographs with significant degradation — very small prints, poor exposure, damaged emulsion — conditions common in family photography from before 1950. Available on mobile and web; free tier with limited monthly enhancements.

MyHeritage Photo Enhancer is designed specifically for family photographs and produces consistently strong results for portrait-style images. Well-integrated into the MyHeritage platform if you're already using that for record research and DNA analysis.

For large-scale restoration work — digitizing and restoring several hundred family photographs — Topaz Photo AI offers the most control and highest quality output. Desktop application, one-time purchase ($199). It combines AI upscaling, sharpening, and noise reduction in a single workflow. The investment makes sense if photo restoration is a significant part of your project.

A note on colorization: converting black-and-white portraits to color is now possible with reasonable accuracy. The results vary by photograph quality. A well-lit studio portrait from 1920 can be colorized convincingly. A damaged snapshot from 1890 will produce less reliable color. Colorized versions are interpretations — never replace the original with the processed version.

For more on photo restoration tools, restoring old photos with AI covers the current options in detail.

Stage 6: Bring ancestor portraits to life

More than 12,000 photographs have been animated on Incarn since launch — most of them ancestor portraits: grandparents, parents who died young, children who became adults before anyone thought to make a video.

Incarn takes a still portrait and produces a short video showing the face with natural, realistic movement. Not a loop. Not a caricature. Not the uncanny puppet-face effect that other tools produce on damaged old photographs. The animation works on grayscale images, formal studio portraits from the early 20th century, and photographs with moderate quality issues. It handles old photographs particularly well — the cases where other tools fail.

One client described seeing his mother's photograph from 1987 move for the first time in 38 years. Her grandchildren, who never met her, saw their great-grandmother breathe. That's what animation adds to research: not more information, but presence.

MyHeritage Deep Nostalgia produces looping portrait animations. Useful if you're already embedded in the MyHeritage ecosystem, and the loop format works for some presentation contexts. The motion range is more limited than what Incarn generates for individual photographs — both are worth trying on the same image if you want to compare.

The first animation on Incarn is free to try — no account required. After that, 1,99€ per photo. The animation generates in under 2 minutes.

For more on what's possible with animated portraits, bringing ancestors to life covers the emotional and practical dimensions.

How to verify AI findings and avoid the most common errors

This is the step most researchers underinvest in — and the one that determines whether AI research holds up.

For record finds: locate the original document image, not just the AI transcription. Transcription errors in indexed records are common, particularly on names and dates. The image is the primary source; the transcription is a search tool.

For tree connections: any AI-suggested family relationship requires independent documentary evidence before you accept it. ThruLines suggests; records confirm.

For DNA interpretations: a single DNA match is a hypothesis. Multiple matches from different platforms converging on the same ancestral couple — cross-confirmed with documentary evidence — is evidence.

For photo enhancements: enhanced, colorized, and animated photographs are interpretations of the original, not replacements. Keep every original. The processed version has value as a presentation tool; it isn't a genealogical document.

The standard of evidence in genealogy hasn't changed. AI makes reaching that standard faster. It doesn't lower it.

For a detailed comparison of specific tools across each category, the AI genealogy tools comparison covers product-level verdicts with more depth than this workflow guide.

Privacy considerations when using AI genealogy tools

Genealogical data is sensitive in ways that aren't immediately obvious.

DNA data uploaded to third-party platforms is subject to those platforms' privacy policies. Before uploading raw DNA to any service, read specifically how they handle law enforcement requests, insurance partnerships, and research data sharing. Policies vary significantly.

Family tree data that includes living people should be managed carefully. Most genealogy platforms allow you to mark living individuals as private. Verify how that designation works in practice — particularly for tree sharing and research collaboration features.

Photographs uploaded to AI enhancement or animation services typically remain yours under current terms of service, but enhanced versions may be used to improve the AI model. For photographs with personal or historical significance, check the terms before uploading.

What's next for AI in genealogy research

The direction is clear: more records accessible to more people, with AI handling more of the reading and connecting.

FamilySearch's AI indexing project continues making previously inaccessible documents text-searchable at a pace that wasn't possible before machine learning. DNA databases grow every year, which improves the statistical power of matching for everyone already tested — including people who tested years ago and found few useful matches.

AI translation tools are improving for the less common languages and scripts that appear in genealogical records: old Cyrillic, Ottoman Turkish, regional Iberian scripts, colonial Latin American notarial hands. Languages that previously blocked research for most non-specialists are becoming more accessible.

Photo animation and restoration have moved from novelty to research workflow. Families who found a photograph of a great-grandparent now have tools that weren't available even five years ago: restore it, colorize it, animate it, share the result with relatives who never knew that person existed.

The ancestors are there. The workflow to find them — and bring them forward — has never been more within reach.