If you’ve been working on your family tree for any length of time, you know it can be exciting, but sometimes frustrating. You hit a wall. A record’s hard to read. You wonder if you’re even following the right person. We’ve all been there. But now, there’s a tool that can help our genealogy research, AI, or artificial intelligence.
AI won’t do the thinking for you—that part is still up to you. But it can take on time-consuming tasks like transcribing and abstracting documents. That means you can spend less time deciphering handwriting and more time discovering your ancestors and their stories.
In this blog post, you’ll learn three transcribing tools that use AI and how they compare.
I’m sure by now you’ve heard of at least one AI tool such as ChatGPT or Claude. I’m going to throw one more in the mix – Transkribus. Let’s see how these three stack up against each other in transcribing and abstracting a genealogy document.
Why Transcribing/Abstracting Is Important
Transcribing is the process of turning handwritten or printed words into clear, typed text. It is a word-by-word copy of the original. In genealogy, we often transcribe:
- Wills
- Deeds
- Church records
- Immigration documents
- Letters and diaries
- Family Bible entries, and more
Every detail in these documents can hold a clue—names, dates, relationships, places, and even who the neighbors were. But old handwriting, foreign languages, and outdated wording can make these clues hard to spot. Transcribing turns that difficult text into something readable, searchable, and easier to understand.
Abstracting goes a step further by pulling out the most important information so you can see key facts easily.
Whether you transcribe by hand or use AI to do it, READ EVERY WORD so you don’t miss any of the details. You don’t want to overlook something that could break down a brick wall in your family tree.
Example
I’ve used all of these tools to transcribe deeds and wills, but I wanted to see how it would handle something I don’t like to transcribe – an inventory. An inventory is a detailed list of a deceased person’s belongings, included as part of their estate or probate records.
The example for this demonstration is from my 3rd great-grandmother, Harriet Mynatt Churchman. She died in December 1875 in Tennessee. Although this document is about her, it actually came up in a search for my 2nd great-grandfather, John Dunahoo, in the FamilySearch full-text search. If you’ve never used it before, read my blog post (it’s a game changer):
The Exciting World of AI: FamilySearch Full-Text Search
The beauty of this AI tool on FamilySearch is the ability to find people who are not indexed in records. John, who was Harriet’s son-in-law, shows up in the inventory of her estate as someone buying items of her property. FamilySearch does provide a transcription, but in some cases, like this one, it’s more of a mess than a help.
Let’s take this part of Harriet’s inventory and see what Transkribus, ChatGPT, and Claude can do.
Transkribus
Transkribus is a website that has a free version for a number of monthly transcriptions. While my experience is limited, it had a hard time reading this document:
In all fairness, it does a great job with letters, deeds, or wills. But I wanted to see how it would handle a list like this one. It does allow you to open up the transcription in an editor so you can make any necessary changes.
ChatGPT
ChatGPT has a free version, which is the one I use. You can upload an image of the document (by clicking the +) and give it a prompt as to what you want.
Not bad! It had trouble with my ancestor’s name, John Dunahoo, and a few initials, but all in all, a good job. I asked it to correct the mistakes and give it to me in a cleaner table format:
I can copy and paste this into my research report easily. It’s best to give small pieces of data, so it’s not overwhelmed. It can get confused and make things up.
Claude
Claude is our last tool to look at today. It also has a free version. Just like ChatGPT, it requires prompts, so I repeated the exact instructions:
While Claude did a better job reading my ancestor’s name, it had a tougher time with the initials of others, and it gave November as the month instead of December. I would continue the conversation to clean up those mistakes and provide this information in a table format if needed.
Abstractions
Next, I asked for an abstract of the transcriptions:
ChatGPT:
Claude:
Both did a good job of deriving key information in a way that is easy to read. Also, notice that Claude did a bit of analyzing, naming R.L. Churchman as a possible relative. It’s just a matter of preference as to which you like better.
Tips for Better Results
- Use the clearest image you can. Blurry or dark scans make it harder for AI to do its job.
- Check every transcription. AI is fast, but it still makes mistakes. Always compare to the original.
- Learn common terms. If you work with the same type of record often (like wills), learn a few common phrases to help you double-check the AI’s work.
- Keep both versions. Save the original image, your transcription, and the AI summary or translation. That way you can always go back and recheck.
Finally
AI isn’t replacing us as researchers. But it’s becoming a helpful assistant.
When it comes to reading old records, transcribing, and abstracting, AI tools can save time, reduce frustration, and help you pull out key facts more quickly. That means you can spend more time making connections and less time figuring out faded handwriting.
If you haven’t tried one of these tools yet, I encourage you to give it a shot. Start with a document you’ve already transcribed and compare results. You might be surprised at how accurate and helpful it can be.
Have you used an AI tool to help with transcriptions or translations? I’d love to hear your experience. Drop a comment below or send me a message.
Genealogy tip: YouTube has many wonderful videos on this very subject. Search “AI Genealogy” to learn more about how artificial intelligence can aid you in your genealogy research.
While I love using AI in my research, you may not know much about it yet. If you’d like a human to research for you or with you, give me a call. You can schedule a free 30-minute consultation below.
