Can AI Change the Game for Your Family History Research?

Artificial Intelligence (AI) was the buzzword for 2024. Being a little conservative, I have been hesitant to experiment with AI. My first foray into using AI was with ChatGPT. I asked it to transcribe a handwritten will I inserted in their chat box. I do not know where AI got some of the names. It also made assumptions. Two of the obvious errors were:

1. That the daughter’s surname was the same as the deceased, even though it identified her as “the wife of …”

2. One of the witness’s names was left out, and another erroneous name was inserted as the witness.

My next experiment was using Lucid to create a DNA match chart showing each generation’s connections. This produced a chart with the generations connected, but it was not aligned with the generations. The boxes were spread over the paper. I spent quite some time trying to get the boxes lined up and in a generational order. I realized I did not know how to ask for what I wanted AI to produce.

It’s all in the prompt. Two more webinars and another attempt brought me closer to what I was looking for. Writing a prompt is probably not my forte as a beginner with AI. So, my next experiment was suggested by Andrew Redfern on 10 Practical Uses of AI in Family History. Let ChatGPT write your prompt. I asked ChatGPT to “Create only a prompt for a family tree illustrating the Most Recent Common Ancestor for two individuals, match 1 and match 2 that are third cousins.” ChatGPT came back with, “An illustrated family tree diagram showing the Most Recent Common Ancestor (MRCA) of two individuals, labeled Match 1 and Match 2, who are third cousins. The tree should clearly depict multiple generations, starting from the MRCA at the top, branching down to their children, grandchildren, and great-grandchildren. Match 1 and Match 2 should appear on the same generational level, highlighting their shared lineage. Use a clean and structured layout with connecting lines to show relationships clearly.” This was a giant leap forward. I copied this prompt into Lucid to create the chart. It is not perfect, but I got a horizontal chart, not a vertical one, as I requested. And with a couple of tweaks, I was able to produce the following:

DNA Match template created with AI at Lucidchart.

I attempted another prompt on a transcription. The first iteration it had not read one of the names correctly. I told AI what the name should be. It thanked me for the correction and rewrote the transcription with the name corrected.

I next asked ChatGPT to “Organize in a spreadsheet with Columns for dates, locations, and relationships, from births, baptisms, marriages, deeds, census, Social Security, military, death, and additional notes.” After adding color, header, and footer, and duplicating line items to allow for multiple entries for some event types, here is what I got.

I was happy with my new template. I am still learning, but I see the advantages in some instances if you take the time to experiment. It may add time to your research by freeing you from some of the time-consuming tasks of creating templates.

Takeaways: