Using AI in Family History

AI can help genealogists search records, read handwriting, translate documents, summarize sources, analyze information, and organize research. But AI cannot replace genealogists. It does not know whether a record truly belongs to your ancestor, and it cannot evaluate evidence the way a trained researcher can.

Use AI with care

The Coalition for Responsible AI in Genealogy has published Guiding Principles for Responsible AI in Genealogy, organized around five major responsibilities: accuracy, disclosure, privacy, education, and compliance.1 These principles are a useful foundation for any genealogist using AI tools because they focus on trust: verifying information, being honest about AI's role, protecting private information, learning how the tools work, and respecting legal or contractual limits.

For genealogists, AI should not be used as a shortcut around evidence. It can help find possibilities, process difficult material, or suggest ways to think about a problem, but the researcher still has to check the original sources, evaluate conflicts, and explain the reasoning.

Can I verify, explain, and support this conclusion with the evidence?

If the answer is no, the work is not ready. AI tools can be persuasive even when they are wrong, and polished output can make unsupported conclusions look stronger than they are.

  • Check names, dates, places, and relationships against the original source.
  • Ask AI to explain uncertainty instead of forcing a single answer.
  • Keep citations and source images connected to your conclusions.
  • Disclose meaningful AI assistance when it affects transcription, summaries, analysis, drafting, or images.
  • Do not enter sensitive information about living people unless you understand the tool's privacy terms and have a valid reason to use it.
  • Do not use AI-generated biographies or proof statements as evidence.

How much did AI help?

The AI Assessment Scale (AIAS) was created for education to describe how AI may be used in student work.2 Later versions continued to refine that framework as AI capabilities expanded.3

Genealogy needs a different emphasis. The key issue is not simply whether AI was used, but what part of the research AI performed and what the genealogist personally verified.

What role did AI play in the research, and what did the genealogist personally verify?

Why use G-AIAS?

  • It helps the genealogist intentionally choose what role AI should play in genealogy.
  • It avoids an all-or-nothing approach. The choice is not simply “AI” versus “no AI.” There are meaningful levels in between.
  • It also separates the roles AI can play. Finding a record, transcribing it, analyzing evidence, and running a multi-step workflow are very different uses of AI.
  • It shows where human judgment is especially important when deciding when to use one level or another.
  • It keeps responsibility with the genealogist. The scale makes clear that greater AI involvement does not mean less responsibility for checking sources and defending conclusions.
  • It makes expectations clear for students, instructors, and researchers when applying the scale to an assignment or project.

Proposed Genealogical AI Assessment Scale — G-AIAS

G-AIAS describes how much of the genealogical research process is shared with AI. A higher level does not mean better research; it means AI is taking part in more of the work and the researcher has more to audit.

Levels 2A and 2B are parallel, not sequential. Discovery helps find sources. Processing helps work with sources already found. A genealogist may use either one, both, or neither.
Level 1 Human-Led Research (No Direct AI Assistance) The genealogist performs the research task.
Level 2A AI-Assisted Discovery AI helps find it.
Level 2B AI-Assisted Processing AI helps work with it.
Level 3 AI-Assisted Analysis AI helps think through the evidence.
Level 4 AI-Integrated Research AI participates across the research process.
Level Name What AI may do What the genealogist must do Typical examples
1 Human-Led Research
(No Direct AI Assistance)		 AI does not directly perform the research task.
?
What does “No Direct AI Assistance” mean? AI may still be built into tools you use. Level 1 means you do not rely on AI-generated summaries, interpretations, recommendations, or analysis to perform the research task. In Google, adding -ai to the end of a search can often suppress the AI Overview; you can also choose the Web filter for text-based results. These are practical ways to keep the search human-led, but if an AI Overview still appears, ignore it and evaluate the linked sources yourself.
Search, read, interpret, analyze, correlate, and conclude independently. Manual record searching; making your own transcription; independently analyzing evidence or constructing a proof argument.
2A AI-Assisted Discovery Help find possible sources, records, collections, people, places, or research leads. Inspect the original sources and independently determine relevance, identity, and reliability. FamilySearch Full-Text Search; record hints; AI-enhanced search; an AI research assistant locating candidate records.
2B AI-Assisted Processing Help transcribe, translate, extract, summarize, or organize information from sources. Compare the output with the original, correct errors, and decide what the information means genealogically. HTR/OCR; transcription; translation; extracting names and dates; summarizing a probate file; converting records into a table.
3 AI-Assisted Analysis Help compare evidence, identify conflicts, suggest hypotheses, critique reasoning, or propose next steps. Evaluate the reasoning, weigh the evidence, resolve conflicts, and make the genealogical conclusion. Comparing records for identity; finding inconsistencies; testing hypotheses; suggesting research directions or a research plan.
4 AI-Integrated Research Participate across multiple connected stages of research, including discovery, processing, analysis, synthesis, or drafting. Direct and supervise the workflow, audit sources and reasoning, verify results, document provenance, and defend the conclusions. AI research agents; multi-step research assistants; workflows that search, process records, analyze evidence, and draft findings.

How to read the scale

  • Classify the use, not the tool. The same AI tool can operate at several levels depending on what you ask it to do.
  • AI output is not genealogical evidence. AI can help find, read, organize, or analyze evidence; the historical sources remain the evidence.
  • The genealogist remains responsible. More AI involvement means more work to verify and audit, not less responsibility.

One research problem, different levels

Imagine you are researching a probate case. The level changes according to what AI is asked to do.

2A Find it

AI helps locate possible probate records.

2B Process it

AI transcribes the will and extracts names, dates, and relationships.

3 Analyze it

AI compares the probate record with census, land, or other evidence and points out conflicts or hypotheses.

4 Integrate it

AI searches, processes records, follows leads, analyzes evidence, and helps draft findings across the project.

The rule that matters most

AI can assist with genealogy, but it cannot carry the proof burden.

The genealogist is still responsible for the sources used, the evidence accepted, the conflicts resolved, and the final conclusion.

Find resources by topic

Use the timeline to find presentations and resources connected to the AI task you want to learn next.

Open the Family History AI Resource Timeline

Notes

  1. Coalition for Responsible AI in Genealogy, “Guiding Principles for Responsible AI in Genealogy,” accessed September 12, 2026, https://craigen.org/. ↩
  2. Mike Perkins, Jasper Roe, Leon Furze, and Jason MacVaugh, “The AI Assessment Scale (AIAS),” The AI Assessment Scale, accessed September 12, 2026, https://aiassessmentscale.com/. ↩
  3. Mike Perkins, Jasper Roe, and Leon Furze, “The AI Assessment Scale Revisited: A Framework for Educational Assessment,” arXiv, December 12, 2024, https://arxiv.org/abs/2412.09029. ↩