I have not used the high-end Artificial Intelligence tools that require a subscription, but I want to stay informed about them and how they can enhance, improve and amplify genealogy and family history work.
Recently, I asked OpenAI ChatGPT5 (the free version) to "Please describe complex use cases that ChatGPT For Work could perform for genealogists and family historians."
I posted 21 complex genealogy use cases in Complex Genealogy Use Cases for “ChatGPT Work” - Part 1 on 25 September 2026 and Complex Genealogy Use Cases for “ChatGPT Work” - Part 2 on 2 October 2026.
Linda Stufflebean commented on Part 2 that "Very interesting AI challenge with the Genealogy Research Manager. Hope you'll take that topic further!"
Consequently, I asked ChatGPT5 (the free version) to: “For item 21 (the genealogy research manager), would you please expand each bullet point with a short description? What actions would I have to take to support Work?"
The ChatGPT5 response was:
Randy, I think the Genealogy Research Manager might be the most useful of all the Work applications we've discussed. Rather than focusing on one task, you'd establish an ongoing research process in which Work helps you manage a difficult genealogical problem from the initial question through the final proof argument.
The key is that you would not turn over your genealogy research to an unsupervised AI. You would remain the researcher, supplying records and making decisions, while Work organizes the investigation, tracks progress, analyzes evidence, and prepares the finished products.
Let's walk through each part of the research manager concept and what Work would do, what you would need to do, and the Deliverables from each step:
The Genealogy Research Manager: How it would work
1. Research question: Define the problem and the desired result.
2. Existing evidence: Collect files, citations, notes and family-tree information.
3. Research and analysis: Search permitted sources, evaluate records and test hypotheses.
4. Research management: Maintain logs, evidence tables, contradictions and next steps.
5. Finished product: Produce a documented conclusion, research report and supporting files.
I'll use your Devier James Lamphier Smith parentage problem as the running example, although this approach would work for any of your Seaver, Carringer, Auble or other family-history investigations.
1. Define the research questionWhat Work would do: Establish the precise question the investigation must answer. It would distinguish the central problem from related questions and establish what would count as a satisfactory result.
For Devier, the central question is:
Who were the biological parents of Devier James Lamphier Smith, reportedly born in Jefferson County, New York, about 1839, and adopted by Ranslow and Mary (Bell) Smith?
Related questions might include:
Was Devier born in 1839 or 1842?
Was Lamphier his birth surname?
Was he biologically connected to Isaac Lanfear and Rosina Lown?
Could one of their daughters have been his mother?
What records might document the adoption or name change?
What you would do: Write a short research brief explaining what you know, what remains uncertain, and why the question matters. Include your existing conclusions, but label them as conclusions to be tested rather than established facts.
Deliverable: A one-page research plan with a clearly defined objective.
2. Collect the existing evidenceWhat Work would do: Inventory your existing research before beginning new searches. It would identify available documents, family-tree reports, previous conclusions, citations, DNA notes and unanswered questions.
It could also compare different versions of your research and flag discrepancies—for example, if one biography gives Devier's birth year as 1839 and another gives 1842.
What you would do: Gather the material you already have, preferably in a dedicated project folder. Useful items include:
Your Devier genealogical sketch and previous blog articles.
RootsMagic reports covering Devier and the Smith and Lanfear families.
Downloaded census, vital, court, land and probate records.
DNA-match notes and relevant descendant trees.
Existing source citations and research logs.
Correspondence or family statements containing relevant information.
You can upload files in manageable batches. You do not need to put your entire 75,000-person genealogy database into the project.
Deliverable: An evidence inventory, a summary of established facts, and a list of missing information.
3. Establish the research standardsWhat Work would do: Follow the rules you set for evaluating evidence. It would distinguish documented facts from inferences, identify source limitations, and avoid treating online family trees as proof.
For example, a census might establish that a household contained people of particular ages and sexes. It ordinarily would not, by itself, prove that a particular child was the biological child of a particular household member.
What you would do: Give Work a written research charter based on the Genealogical Proof Standard and your preferred citation practices.
I'd include these rules:
Cite every important factual conclusion.
Distinguish original records, derivative sources and authored narratives.
Preserve exact source details and record images.
Identify conflicting evidence rather than silently choosing one version.
Record searches that produced no relevant results.
Separate facts, reasonable inferences and speculation.
Actively test alternative explanations.
Never invent records, citations, quotations or relationships.
You should also specify whether you want citations formatted according to Evidence Explained.
Deliverable: A reusable set of research instructions for the entire project.
4. Build the research timelineWhat Work would do: Assemble a chronological account of the people, places and events relevant to the question.
For Devier, this would include the Lanfear family's marriage and children, census appearances, Devier's reported birth and adoption, the Smith family's movements, and later records documenting his identity.
The timeline would help identify impossible dates, gaps and opportunities for further research.
What you would do: Review the initial timeline and correct errors. Supply dates or records that Work overlooked. When a date is uncertain, ask Work to preserve the uncertainty rather than assign a falsely precise date.
Deliverable: A dated chronology with a citation or evidence reference for each entry.
5. Maintain a research logWhat Work would do: Record each research action, including the repository, collection, search terms, dates covered, records examined, findings and follow-up tasks.
For example:
Research action |
Result |
Next step |
|---|---|---|
Examine 1840 census households in Jefferson County |
Potential Lanfear households identified |
Compare household composition with known family |
Search available local newspapers for Devier |
No relevant item found in the collection searched |
Try surname variants and other newspapers |
Review supplied probate records |
Possible family connections identified |
Obtain complete probate packet |
These are examples of log entries, not claims that those searches have already been performed.
What you would do: Because subscription sites such as Ancestry and MyHeritage may prohibit AI-agent use in their restricted areas, you would perform those searches yourself. Download the records where permitted, then give Work the files and search details.
At the end of each session, ask Work to update the log and save a copy to your project files. Verify that the saved version exists and contains the changes.
Deliverable: A continuing research log that helps prevent duplicated effort.
6. Conduct and coordinate new researchWhat Work would do: Recommend the next records to seek based on what is already known. It could identify repositories, suggest search variants, follow leads in accessible sources, and prioritize promising research avenues.
For Devier, possible avenues include:
Federal and state census records.
Jefferson County vital and church records.
Probate and guardianship records.
Land transactions involving Lanfear and Smith families.
Court records relating to adoption or name changes.
Historical newspapers and local histories.
Records documenting the lives of Isaac and Rosina's children.
Work should distinguish between a source it actually examined and a record set it merely recommends searching.
What you would do: Perform searches that require your subscription access, order records when appropriate, and upload the results. For publicly accessible sources, Work may be able to search directly, subject to the website's terms and technical limitations.
You would also decide whether the cost of obtaining an original record is justified.
Deliverable: A prioritized research queue with specific searches, their purpose and their expected evidentiary value.
7. Maintain an evidence matrixWhat Work would do: Organize evidence by research question and hypothesis rather than merely listing records.
For instance:
Evidence |
Supports |
Limits or conflicts |
|---|---|---|
Devier's reported birthplace |
A Jefferson County connection |
Does not identify his parents |
Lanfear family composition |
Potential connection through a daughter |
Household structure alone cannot prove maternity |
Autosomal DNA matches |
Potential biological connection to the Lanfear family |
Shared DNA alone may not identify the exact ancestral path |
Adoption or name-change evidence |
Could clarify Devier's identity |
Must be examined for the relationships it actually documents |
The matrix would be updated as new records arrive.
What you would do: Check whether Work has interpreted each item correctly and whether the source actually supports the claim. Correct mistaken identifications and supply missing context.
Deliverable: An evidence matrix that makes it easier to see which conclusions are supported, contradicted or still unproven.
8. Test competing hypothesesWhat Work would do: Compare plausible explanations instead of focusing exclusively on the leading theory.
For example:
Devier was a biological child of Isaac Lanfear and Rosina Lown.
Devier was the child of one of their daughters.
Devier belonged to another Lanfear family.
The DNA connection arose through a different ancestral path.
Work would identify the evidence that each explanation predicts, then compare those predictions with the actual records.
What you would do: Tell Work when a hypothesis is missing, implausible or inconsistent with your prior research. Ask it to make the strongest case against your preferred explanation, not just the strongest case for it.
You would also decide when the evidence justifies eliminating a hypothesis.
Deliverable: A documented comparison of competing explanations, including unresolved weaknesses.
9. Integrate DNA evidenceWhat Work would do: Analyze the match information you supply, group matches by hypothesized ancestral lines, and compare their family trees with the documentary evidence.
It could help distinguish several independent descendant lines from multiple matches who all descend from the same recent branch.
What you would do: Export or provide appropriate match data, shared centimorgan amounts, documented ancestral lines, and any chromosome or triangulation information available to you. Protect other people's private information and follow the DNA service's terms.
Review Work's proposed relationships carefully. Shared DNA can support a family connection, but it does not automatically establish a specific parent-child relationship.
Deliverable: A DNA evidence summary integrated with the documentary research, including limitations and alternative explanations.
10. Review negative and conflicting evidenceWhat Work would do: Examine what was not found as well as what was found.
Examples include a missing person in a census, an absent probate reference, conflicting birth dates, or the failure to locate a record in a particular collection.
It would document the scope of each search and assess whether the absence is meaningful. A missing name in an incomplete index, for example, is much weaker evidence than absence from a demonstrably complete record set where the person would ordinarily be expected to appear.
What you would do: Supply the details of your searches, including collections and date ranges. Verify whether Work has accurately described the search and whether the records are sufficiently complete to support an inference from absence.
Deliverable: A negative-evidence register and a list of contradictions requiring resolution.
11. Maintain the project and its working filesWhat Work would do: Keep the evidence inventory, timeline, research log, hypotheses and working report coordinated as the project develops.
Ideally, every new record would prompt the relevant documents to be reconsidered. Work could identify which previous conclusions need revision.
What you would do: This is an important practical point: don't rely on a conversation alone as the permanent repository of your research.
Save the important outputs as files in your project folder. Periodically ask Work to reconcile the research log, evidence matrix and report, and check that the saved files are current. If Work loses context or reaches a usage limit, those files provide a way to resume without starting from scratch.
Deliverable: A portable, backed-up research dossier that you control.
12. Prepare and review the finished reportWhat Work would do: Turn the accumulated research into a coherent genealogical report with the research question, background, methodology, evidence, analysis, conclusion and source citations.
The conclusion might establish parentage, identify the most likely hypothesis without claiming proof, or explain why the question remains unresolved.
What you would do: Review the report against the original records. Check the citations, verify key quotations and dates, challenge unsupported claims, and decide whether the evidence meets your proof standard.
After you approve it, you could adapt the report for RootsMagic, a Genea-Musings article, a family-history book or a genealogy society presentation.
Deliverable: A documented research report, a final research log, an evidence matrix and a list of recommended future searches.
Your practical workloadYou don't need to do all of this at once. Here's the division of labor I'd aim for.
Your work: the irreplaceable part
Define the question, provide your existing research, search restricted subscription sites, obtain records, correct errors, judge evidence and approve the conclusion.
Work's role: the time-consuming support
Inventory files, transcribe supplied records, build timelines, maintain logs, compare evidence, investigate accessible sources, challenge hypotheses and draft reports.
My recommendation is to start small, even though the eventual project may be extensive.
Project startup checklistCreate a project folder
Use a dedicated folder for Devier's parentage investigation and its working files.
Write the research question
State what you want to establish and what is currently uncertain.
Gather existing research
Include your best genealogical sketch, relevant RootsMagic reports, DNA notes and key source documents.
Ask Work for an initial audit
Have it inventory the evidence and identify gaps before conducting new research.
Review and approve the plan
Correct the initial findings and agree on the first research priorities.
Copy checklist
One final recommendation:
Have Work produce an initial assessment before it starts searching for more evidence. You have already done substantial research on Devier, including the Lanfear DNA hypothesis. The first objective should be to organize and critically assess what you already know, not to rediscover it all.
That gives you a baseline against which every subsequent discovery can be measured. It's also the best way to find out whether Work is genuinely adding value to your genealogy rather than simply generating more pages of prose.
================================
The Google Gemini Notebook video overview of this blog post is on my YouTube channel here, and below:
I have collected in one form or another the existing research information, relevant reports, DNA notes, and key source documents for the Devier Smith research question. However I need to put it all into one file folder and provide access to the file folder (not my entire computer file set) to Work.
Work recognizes (because I told Work earlier!) that I will have to perform the searches on the genealogy websites (Ancestry, FamilySearch, MyHeeitage, American Ancestors, etc.) myself. Then I would pass the results to Work, because of the restrictions on use of AI while within the website subscription walls.
Frankly, the only thing stopping me from doing this project using Work is time (it is another project that can be performed by rearranging my priorities) and a $20/month subscription to ChatGPT Plus to use Work. That is possible but I'm not going to do it at this time. I wish I was younger with more energy.
Links to my blog posts about using Artificial Intelligence are on my Randy's AI and Genealogy page. Links to AI information and articles about Artificial Intelligence in Genealogy by other genealogists are on my AI and Genealogy Compendium page.
The URL for this post is:
Copyright (c) 2026, Randall J. Seaver
Please comment on this post on the website by clicking the URL above and then the "Comments" link at the bottom of each post. Share it on X, Facebook, or Pinterest using the icons below. Or contact me by email at randy.seaver@gmail.com. Note that all comments are moderated, and may not appear immediately.
Subscribe to receive a free daily email from Genea-Musings using www.Blogtrottr.com.
No comments:
Post a Comment