For years I’ve subscribed to Czernowitz-L, a mailing list for people researching families from Czernowitz and the rest of Bukovina. My Mittman family came from there. The list is where I hear that someone has indexed a set of records, scanned a city directory, or worked out which archive holds the census. It is some of the best advice I get. It also arrives one message at a time, and then it sits in my mailbox, where I can never find it again when I need it.
This month I tried something different. I gave my whole archive of the list to Claude, an AI assistant, and asked it to work out what was in there. The result is a new page under Places: Czernowitz (Chernivtsi), Bukovina, Ukraine. This post is about how that went, because I think it’s a good way to use AI in family history: not to do the research, but to organize what you already have.
What went in
I exported the mailbox from my mail program as a single file. It held 577 messages from 111 people, dated from November 1998 to this month, though most are from the last three years. I didn’t sort or clean anything first.
What the AI did with it
First it made an inventory. It read every message and pulled out every web link: 6,866 links to 293 different websites. Then it sorted them. Most were not sources at all. They were the list’s own footers, antivirus notices, meeting invitations, news articles, book listings and travel photos. It set those aside and told me what it had left out and why.
What remained were 43 sites that hold records or documents. For each one it wrote a short entry: what the site has, which years it covers, who on the list recommended it and when, and any warnings members had raised. It grouped them into seven kinds, from indexes and databases to cemetery documentation. It also noticed nine sources that members had talked about without ever giving a link, such as the state archive in Chernivtsi, and listed those separately.
The part I use most came next. A list of websites is still just a list. The question I actually ask is “where do I look for a birth in 1905?” So the page has a table called Records by era. It runs from the Jewish birth, marriage and death books of 1856–1929, through the census books and city directories, to the wartime files and what came after, and says where each can be found.
Then it turned the inventory into a page for this site, with all 43 sources, the city’s names in six languages, a few facts at a glance, and a short timeline of the city’s history.
Something I had forgotten
Reading everything turned up more than websites. In a message from January 2008, a list member had looked up my family in the 1907 Czernowitz directory for me. It lists Leibisch Mittman, carpenter, at Landhausgasse 13, and Nuchem Klang, watchmaker, at Waggasse 4. I had that in my mail for eighteen years. It’s now on the page under Connections on this site, next to the birth record of Bernhard Mittmann, which names his father as Leibisch, a carpenter’s assistant.
What was still my job
- Deciding what the page is for, and what counts as a source. I wanted sites that hold records, not sites that talk about history.
- Checking. The source entries come from what list members wrote, so they are as good as the list. The history timeline is different. The AI wrote that from its own general knowledge, and the page says so: it is an outline I still need to check against published histories.
- The research itself. The page tells me where to look. It doesn’t look for me. The section for research notes is still empty.
Why I’d recommend it
Most of us have material like this: years of mailing-list messages, old correspondence with cousins, notes from a conference. We trust it, and we can’t find anything in it. Sorting it is slow, dull work, and it is the kind of work an AI assistant does well. It reads all of it, it doesn’t get bored at message 400, and it can say which message each fact came from.
A few things I’d pass on:
- Start with material you already know is good. The AI organizes what it’s given. It can’t make a weak source strong.
- Ask what it left out. The list of what was set aside told me as much about the result as the list of what was kept.
- Keep the credit with the people who did the work. The entries on the page name the list members who built or announced each resource. The AI sorted their work. It didn’t replace it.
- Treat anything the AI adds from its own knowledge as a draft, and mark it that way until you’ve checked it.
- Put the result somewhere you’ll keep it up. A page on the site can be updated when the next announcement comes through the list.
What’s next
The layout of the Czernowitz page is saved as an outline, so the next place page starts from the same sections: at a glance, history, records by era, resources, and connections to the families on this site. I have other places to do, and other mailboxes.
If you research Czernowitz or Bukovina families and know of a source I’ve missed, please let me know.
Leave a Reply