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 Work could perform for genealogists and family historians."
(AI Google Gemini Notebook Inbfographic - Complex Genealogy Use Cases for ChatGPT Work)
The ChatGPT5 response was (split into two parts - this is Part 1) is:
The big difference is that ordinary ChatGPT answers a genealogy question; ChatGPT Work can take on a genealogy project—researching, navigating websites, handling files, comparing evidence, producing a finished artifact, and continuing through many steps.
OpenAI describes Work as an agent that can work across apps, files, and the web and stay with complex projects for hours.
For a serious family historian, that opens up some pretty interesting possibilities.
1. A complete “brick-wall investigation”
Imagine giving Work this assignment:
“Investigate the parentage of Devier James Lamphier Smith, born about 1839 in Jefferson County, New York. Use my research files, my family tree, census records, probate, land records, newspapers, court records, and DNA evidence. Identify every plausible parentage hypothesis, test each one against the evidence, document negative evidence, and produce a research report with citations.”
Work could potentially:
- Search your genealogy files for everything you've already discovered.
- Examine your tree and existing notes.
- Search relevant online repositories.
- Open record images and transcriptions.
- Build a chronological timeline.
- Compare households in the 1830 and 1840 censuses.
- Investigate the Lanfear families in Jefferson County.
- Examine land transactions and probate.
- Compare the evidence against your DNA-match hypotheses.
- Keep track of which sources support or contradict each hypothesis.
- Produce a research report rather than merely a pile of search results.
That's a fundamentally different workflow from asking, “Who were Devier's parents?”
2. “Research what I already have before searching”
This may actually be one of the most useful applications for you.
You have decades of accumulated genealogy material. Work could be given a folder containing:
- genealogy sketches
- PDFs
- census images
- obituaries
- correspondence
- research notes
- newspaper articles
- photographs
- RootsMagic reports
- previous blog posts
- DNA research notes
Then:
“Before doing any new research, determine everything I already know about Sarah Fletcher Kemp. Organize it chronologically, identify contradictions, identify missing evidence, and list research questions.”
Work can work across files and connected file services rather than treating each document as an isolated upload.
For someone with a 75,000+ person tree and thousands of documents, this could be enormously valuable.
3. An automated Genealogical Proof Standard investigation
You could give Work a person and say:
“Evaluate this person according to the Genealogical Proof Standard. Separate original from derivative evidence, analyze source quality, identify conflicting evidence, search for reasonably exhaustive evidence, and tell me what additional research is needed.”
It could create something like:
- Research question: "Who were the parents of Sarah Fletcher?"
- Evidence located:
- 1802 birth evidence
- 1820 census
- 1830 census
- marriage record
- children's records
- probate
- land records
- cemetery evidence
- newspaper evidence
- Evidence supporting hypothesis A
- Evidence supporting hypothesis B
- Evidence contradicting A
- Unresolved questions
- Next five searches
That's much closer to having a research assistant than a chatbot.
4. Census reconstruction across multiple decades
Give Work a family and ask:
“Reconstruct this family's household from 1790 through 1880. Find every census household, identify probable members, compare ages and birthplaces, identify neighboring households, and produce a family timeline.”
It could work across multiple census years and websites, rather than you manually opening dozens of pages.
Even better:
“Look at the households surrounding Henry Carringer in 1850, 1860 and 1870. Identify neighbors who may have been relatives, associates, or migration companions.”
That gets into cluster research, which is where genealogy becomes particularly interesting.
5. FAN-club research at scale
This is one I'd especially like for genealogy.
Instead of researching only: Henry Carringer
Work could research:
- Friends
- Associates
- Neighbors
- witnesses
- sponsors
- employers
- landlords
- business partners
- nearby families
- people appearing in probate
- people appearing in land transactions
Then ask:
“Determine whether any of these people provide evidence about Henry Carringer's origins.”
Work could build a cluster map and identify connections that you might otherwise overlook.
6. Land-record research
Land records are notoriously tedious.
A Work assignment could be:
“Find every land transaction involving Henry Carringer in Mercer County, Pennsylvania between 1825 and 1855. Extract grantor, grantee, dates, acreage, locations, witnesses, adjoining landowners and consideration. Put everything into chronological order and identify relationships.”
The result could be a spreadsheet such as:
Date |
Grantor |
Grantee |
Acres |
Location |
Witnesses |
Significance |
|---|---|---|---|---|---|---|
1831 |
… |
Henry Carringer |
40 |
Sandy Lake |
… |
First appearance |
1837 |
… |
Henry Carringer |
25 |
… |
… |
Adjacent tract |
1849 |
Henry Carringer |
… |
15 |
… |
… |
Possible relocation |
That turns hours of clerical work into an organized research dataset.
7. Probate investigation
Give Work:
“Investigate the probate of John Smith of Jefferson County, New York. Identify every person named in the probate documents, classify each person's relationship to John if stated or implied, and compare the names with the 1830 and 1840 census.”
It could potentially:
- locate the probate file
- download documents
- OCR handwritten pages where possible
- transcribe them
- identify names
- compare names against your tree
- construct a relationship hypothesis
- flag discrepancies
And importantly, you could tell it:
Do not infer a relationship unless the evidence supports it.
That's exactly the sort of constraint that makes AI much more useful for genealogy.
8. Newspaper mining
This one could become addictive.
Suppose you have 500 newspaper references to Seavers.
You could say:
“Search these newspapers for members of the Seaver family from 1850–1920. Extract every article that mentions a Seaver. Classify each as birth, marriage, death, accident, employment, travel, legal matter, social event, military service, property, or miscellaneous. Eliminate duplicates.”
Then:
“Identify events that are not currently represented in my genealogy database.”
That's a very powerful gap-finding operation.
9. Automatically turning research into source-based biographies
You could give Work:
- census records
- vital records
- newspapers
- military records
- land records
- photographs
- your genealogical sketch
and say:
“Write a 2,000-word biography of Isaac Seaver. Every factual assertion must be traceable to one of the supplied sources. Separate documented facts from reasonable historical context. Do not invent dialogue or undocumented events.”
Then another instruction:
“Now create a second version written as a historical narrative, clearly labeling reconstructed scenes as plausible rather than documented.”
That would fit beautifully with your ancestor-story workflow.
10. “Find the holes in my tree”
This is a particularly interesting Work project.
“Analyze my family tree for people who have a birth event but no birth source, a marriage event but no marriage source, a death event but no death source, or census gaps of ten years or more.”
Then: “Rank these research tasks by the number of descendants potentially affected.”
Not a ranking of people—rather, a research-priority list based on objective criteria you specify.
Work could then attack those problems one at a time.
Part 2 will follow shortly with ten more examples.
No comments:
Post a Comment