r/AIGenealogy

Claude AI has excellent transcription and translation capabilities, and works with results from my Ancestry account. When combined with my GEDCOM file from Ancestry, it can hit discoveries out of the park.

Here's how I get the most out of it.

I use a Claude Desktop app on MacOS. I operate it in "Co-Work" mode.
I ask Claude to set up a project for Genealogy. Claude will create a dedicated directory "Genealogy" (within its "Projects" directory - the only directory it's permitted to read/write from).
I'll put my GEDCOM file in the project directory - downloaded from Ancestry, and direct Claude to read it "Read the GEDCOM file and extract all information from it. This is my family. I am astroproff born 1978." This gives Claude context to interpret the document.
Then, I take jpg historical document and tell Claude "Read this. My relative is Lars Hansen. Find my relative in it and extract all the information from the entry. Point out information which is new."

For example, here is a handwritten page from a Danish church record. It contained a possible relative of mine. I hand it to Claude, give it the above command. It churns for about 5 minutes:

This is Claude's response, in full - which, because I can't read Danish, and the handwriting is nearly inscrutible to me anyhow, is an enormous help:

=========================================================

Got it — entry 22. Here is the complete reading of his record, plus what it adds to what you already have.

Lars Hansen — birth/baptism record, Lindelse 1846 (entry No. 22)

Born: 28 August 1846
Baptized in church (i Kirken): 11 October 1846
Parish: Lindelse Sogn, Langelands Sønder Herred, Svendborg Amt — entered into the main parish register on 1 July 1847 by Pastor Toft
Parents' residence: Hersløv Mark (the outlying fields of Hersløv, a hamlet in Lindelse parish)
Father: Hans Jørgen Hejne Larsen — written with the prefix "Ungkarl" (literally "bachelor / young man"; usually marital status, but here paired with "hd." it most likely just signals he was a young husband)
Mother: Maren Pedersdatter — listed as "hd." (hustru = his wife)
Godparents / witnesses (Faddere):
- Hmd. (husmand — cottager) Hans Pedersen of Kjædeby
- Ungkarl Jørgen Madsen of Kjædeby
- Pigen (unmarried young woman) Caroline Madsdatter of Kjædeby
- Godmother: wife of Hmd. Morten Christiansen of Hennetved
Jævnførelses-Register cross-reference: 880 / 369 (this is where his entry is indexed in the parish's running comparison register — useful if you want to pull confirmation/marriage/death entries for him out of the same Lindelse books)

What's new vs. your GEDCOM

None of these people — Lars himself, his parents Hans Jørgen Hejne Larsen and Maren Pedersdatter, or any of the four witnesses — currently appear in your family tree. The two "Lars Hansen" individuals already in your file are different men (one b. 1836 in Kerte, Odense; one b. ~1816 in Mesinge, Odense). So everything in this record is net-new information.

A connection worth checking

This is the part to look at carefully. Your tree already includes Ane Marie Hejne Hansen (b. 5 Aug 1848, Helsned, Humble parish — the parish next door to Lindelse on Langeland), who married Mads Frederiksen and is the mother of Rasmus Madsen (the Sacramento immigrant). Her father is recorded in your GEDCOM only as "Hans," born ~1818 in Fyn, married to Dorthe Madsdatter.

The middle name "Hejne" is unusual and is the same one carried by Lars Hansen's father here. That raises a real possibility that Lars's father Hans Jørgen Hejne Larsen is the same "Hans" who later (after Maren Pedersdatter died) remarried Dorthe Madsdatter in Humble — which would make Lars Hansen a half-brother of Ane Marie Hejne Hansen, and therefore an uncle to Rasmus Madsen. That fits the chronology (Maren's child in 1846, Ane Marie born to Dorthe in 1848) and the geography (Lindelse → Humble, neighboring parishes). It would also explain why your "Hans" has no surname recorded — it had never been pinned down.

This is a hypothesis, not a confirmation. The thing that would prove it is the marriage record of Hans Jørgen Hejne Larsen to Dorthe Madsdatter in Humble or Lindelse, somewhere between 1846 and 1848, and/or a death/burial record for Maren Pedersdatter in Lindelse in that same window. Both would be in the same Rigsarkivet kirkebog series you already have references to (Reference 8008907761 / 8008908021 / 8009029901 for Humble and Lindelse).

Suggested next steps

Pull the Lindelse 1846–1848 deaths to look for Maren Pedersdatter
Pull the Lindelse + Humble marriages 1846–1849 for a Hans Jørgen Hejne Larsen → Dorthe Madsdatter
Look for Lars Hansen in the 1850 Lindelse folketælling — that will tell you if he was still alive, who he was living with, and whether the household has reorganized after a parental death/remarriage
Pull the Lindelse confirmation registers c. 1860–1861 (boys confirmed around age 14) — that's where you'd find Lars Hansen confirmed if he lived

Want me to draft a research log entry for this and save the hypothesis (and the new individuals) so you can pick it up next time?

==================================================

Since I've done this a few times now, I thought I'd write a quick "How to" for other genealogists.

This is the basic prescription.

You're going to need to download and have a *paid* account on Claude. Why paid? Because the free account just doesn't offer enough computation time.
Get used to using the Claue App on your desktop/laptop. Specifically, put it in Co-Work mode - which enables it to read files on your computer and set up a "Project" space for your genealogical work, like a directory called "Genealogy".
Download your GEDCOM file from Ancestry, and put it in your Genealogy project directory for Claude.
In Co-Work mode, instruct Claude to read your GEDCOM file, and look for the ancestor of interest by name, where the wall goes up. Let's call him "Edward Forgotten".
Tell Claude to search for Edward Forgotten on your family tree on Ancestry. com. Claude will probably need to access it through your browser, so grant permission to do so.
Finally, tell Claude to search for the documentation which will break down the wall. For example "Find Edward Forgotten's documentation near where he died, such as his marriage record, baptism record or birth record."

The response to this, if it is like mine, will be a description of that documentation, pulling out the key information - like exact date, location, parents, etc.

Really finally - READ all the discovered documentation. You want to be at least as familiar with this, as you are other documentation on the tree, and this is to confirm that claude did an accurate job of reading the documentation.

That's gotten me two generations of parents beyond my brick wall for my maternal grandmother's maternal grandmother in Denmark. So now, I have all the way back to her paternal grandfather.

The accounts cost money - so I won't say it's worth it. But I've found it extremely helpful in getting information that many hours of dilligent searching had not uncovered.

How To: Transcribing and Translating documents with Claude AI

Who better to ask for a prompt but AI itself?

How to break down a wall using Claude AI and Ancestry records