How I used Claude AI to examine my work on one Danish family in the mid 1800s.
tl;dr I asked Anthropic's Claude to read 20 jpg files of handwritten church documents in Danish that I had placed on my computer in a subdirectory. He did, translated them to English, found the entries relevant to my family (which he learned from my GEDCOM file), and confirmed my previous work of extractions; found a few inconsistencies; found the place names for all the births, baptisms, confirmations and marriages and compared them with entries in GEDCOM.
This was extremely useful. Claude was able to do reading/translation/consistency check work/report writing which would have taken an experienced Genealogist (who could read Danish) over 100 hours of work, in less than an hour's work for me.
----------------------------------------------
My mother's maternal grandparents were Danish (her father was Italian); and my father's side is 100% Irish.
The Danes were immigrants to the US circa 1890-1910. Understanding and documenting the Danish branch of my family has been one of my greatest challenges.
Much of the most useful Danish sources from this period are the church documents (available and Indexed on Ancestry) of births and baptisms, confirmations, and marriages; these often contain full names of parents, with repetetive birthdates and townland names where they live.
These church documents are handwritten, in Danish, and I (who don't speak Danish) struggled for months pouring over the documents. The Indexes for these documents on Ancestry are already useful - but they are sometimes incomplete.
Furthermore, my Danish family has been challenging to me because the patronym changes every generation (e.g. Lars Penderson's son's last name is 'Larsen'); and I don't know Danish geography well, which made intercomparing place names challenging at times.
I wanted to see what I could do with Claude to double check my work, find any information I may have missed, find inconsistencies, and make suggestions for additions to my family tree.
For this test, I focussed on one Danish family - my 2nd great grandparents Mads Frederickson and Anna Hansen. Mads and Anna had 10 children.
Thus, attached to Anna's page are about 20 sources of church documents, in jpg format, handwritten in Danish. These documents are full pages which contain a lot of extraneous information from other families, which I have to read to find the information related to my own family.
From my Ancestry .com family tree, I downloaded to my local computer 20 jpg document copies of Danish, handwritten church records. I also downloaded my full GEDCOM file. I put these into what's called a "Project Directory", which Claude can read while in "Cowork" mode - a mode which allows it to interact with local files, manipulate the files, and create and populate new files.
This is what I prompted Claude with:
I would like to check the consistency of Danish church documents and census documents for the children of my 2nd great grandmother Anna Marie Hejne Hansen and my 2nd great grandfather Mads Frederickson.
Can you see the directory marked Anna Marie Hejne Hansen and Mads Frederickson?
Claude responded:
Yes, I can see the directory "Anna Marie Hejne Hansen and Mads Frederickson" inside your Genealogy folder. It contains 20 image files — the filename prefixes (48551 and 48623) suggest a mix of Danish parish register pages and census images from Ancestry/Rigsarkivet.
Before I dive in, a few quick questions to make sure I check the right things:
- Children list source — should I treat the GEDCOM file (XXXX Family Tree.ged) as the authoritative list of children to verify, or do you have a separate list of children in mind?
- Scope of "consistency" — are you primarily looking for:
- Matching birth dates / names / parents' names between church baptism records and census entries
- Children listed in one source but missing from another
- Spelling variations or transcription discrepancies
- All of the above
- Output format — would you like a Word document summarizing findings (a table per child with church record vs. census entries side-by-side, plus any inconsistencies flagged), or a more informal chat summary?
Also, the images are in Danish gothic/Gothic script — I'll do my best to read them, but church records from this era can be difficult. I may need to flag entries where I'm uncertain rather than guess.
After I answered these questions, in brief, what Claude then:
* Read all 20 jpg documents in handwritten Danish, translated them to English, and found the entries relevant to my family, ignoring the other entries.
* Compared the information found in the church documents to what was in my GEDCOM file, and found a few inconsistencies between the two, and some places where I had not inserted useful information.
* Claude found a few inconsistencies in the church documents which warrant a closer look by me.
* Wrote up a 12 page document, with tables, directly compared the gleaned information with the GEDCOM file contents, highlighting recommended changes (to the GEDCOM file, based on the gleaned information).
Just one example of new information this exercise provided, is this. Claude found an inconsistency between the documents and my GEDCOM file - specifically, one of the documents documented a child's re-marriage, and the written document contained the death date for that child's first husband, and where he died. That information was not indexed by Ancestry attached to the document (unsurprisingly). I previously had a wrong date/location - decades later, in the wrong country even - since I hadn't found any other documented death information. This corrected the record - my relative had been widowed, not divorced, prior to her second marriage.