The use of artificial intelligence (AI) for genealogy research is still in its infancy but gaining momentum as each new tool is released, including ChatGPT and Google Translate. These tools do not replace the need for a professional genealogist to analyze the data but can speed up the research process by finding information more rapidly.
We asked our research team how genealogists can apply AI to their genealogy research projects. Here are some of their key learnings and cautions.
First, How AI Tools Work
The AI tools used most often by genealogists rely on NLP – Natural Language Processing. NLP enables machines to interpret, understand, apply, and generate human language. Researchers can use NLP effectively to extract data from historical documents for analysis. Documents most readily analyzed include birth and death certificates, census records, marriage licenses, and even newspapers. The NLP algorithms can extract names, dates, and locations from these records, saving a researcher valuable time.
AI tools also have machine learning algorithms that help to predict relationships. Large datasets of family history information can train the algorithm to make predictions about family relationships or to find missing information. The algorithm may then be able to fill in gaps in a family tree or predict how likely two people are related to one another based on their DNA test results.
Using ChatGPT for Genealogy Research
ChatGPT is an artificial intelligence chatbot developed by OpenAI. When answering a query, it can create a response that appears to be human-created text. It is also able to write and provide data on a variety of subjects.
One of the challenges of using ChatGPT for genealogy research is the need for more accuracy and the requirement to fact-check. A few tests run by Legacy Tree Genealogists found that while ChatGPT can produce information, it occasionally adds a bit of fiction.
“I found when I asked ChatGPT to write me an early history of Smith County, Tennessee, with sources, that the information was mostly accurate, but the sources were made up. That’s part of the problem–it’s an LLM (large language model), meaning it guesses at the next most likely word or character rather than being an actual research AI,” said one genealogist.
Another Legacy Tree researcher said, “I used ChatGPT with a prompt ‘Tell me about my second great-grandfather Blazuius Britanik.’ It gave me a totally plausible response but was also completely incorrect. I’m curious if AI will be putting out more bad information that we’ll have to be aware of as genealogists.”
ChatGPT can be very helpful when asked carefully-worded and specific questions, but it can also provide inaccurate answers that a researcher must check.
When we asked ChatGPT how we could use the platform for genealogy research, it responded with:
1. Ask specific questions: You can ask ChatGPT specific questions about your family history, such as “Who were my great-grandparents?” or “Where did my ancestors come from?” ChatGPT can use its vast knowledge base to provide you with answers or suggestions for further research.
2. Ask for research tips: ChatGPT can also provide you with research tips and suggestions for further research. For example, you can ask “What are some online resources for genealogy research?” or “What are some tips for tracing your family tree?”
3. Use ChatGPT to transcribe handwritten documents: If you have old handwritten documents, such as letters or journals, you can use ChatGPT to transcribe them. Simply take a photo or scan the document and upload it to ChatGPT. ChatGPT can use its optical character recognition (OCR) capabilities to transcribe the document. *This is an example of the inaccuracy of ChatGPT. At this time, ChatGPT is not able to ‘read’ handwritten documents or have photos or images uploaded, yet it gave us this information.
4. Use ChatGPT to translate documents: If you have documents in a language you cannot read, ChatGPT can help you translate them. Simply upload the document to ChatGPT and ask for a translation. *Another example of the inaccuracy of ChatGPT. At this time, ChatGPT does not allow document uploads in the traditional sense, but you can copy and paste text into ChapGPT for translation.
5. Use ChatGPT for record searches: ChatGPT can also help you search for records, such as census records or immigration records. Simply ask ChatGPT to help you find a specific record and provide any details you have about the person you are researching.
ChapGPT Takeaway:
ChatGPT is a great tool to help formulate research strategies and research historical context. However, all information must be sourced and checked for accuracy, and it is not a replacement for an experienced genealogist.
Using Google Translate for AI Genealogy Research
This AI tool is an excellent help to genealogists, especially when translating records from one language to another.
An experienced researcher at Legacy Tree has found Google Translate to speed up the work. “In my research work, I used Google Translate to take a photo of text written in another language, and I’m shocked at how good it is sometimes. Of course, we still verify, but it saves us so much time from translating documents that might not be for the right person.”
Google Translate Takeaway:
Google Translate is a good tool for starting a translation or determining if a document has information about the person or family you are researching. Verifying the translation with a professional is always necessary for accurate interpretation.
How Major Genealogy Companies Use AI
MyHeritage has also developed a suite of AI-powered genealogy research tools, such as identifying ancestors in historical photographs and automatically colorizing black and white photos. This has led to the creation of photo animation and engaged many new genealogy users through the interactive AI Time Machine.
Ancestry successfully used AI handwriting recognition software to transcribe documents like the 1950 Census records. They also use suggested records in their search, identifying whether a record looks similar to another person in a different record to help merge duplicates or find other relatives.
FamilySearch is currently training the algorithm to index handwritten documents. Current research is focused on the 1400s at speeds unmatched by human researchers. In 2022, FamilySearch released the 1950 US Census by creating an initial index from digital census images that volunteers later checked for accuracy.
If you’d like help with your genealogical research, you can get a free quote here!
Latest posts by Legacy Tree Genealogists (see all)
- Using AI Effectively for Genealogy Research - April 27, 2023
- RootsTech 2023 Recap: Uniting and Learning In Genealogy Family Research - March 10, 2023
- Legacy Tree Genealogists RootsTech 2023 Schedule - February 9, 2023
When I ask CHAPGPT how to upload a handwritten document for tanslation, it tells me that you can’t upload documents?? What am I missing? Thanks
Hi Vicky. We have made an update to the article above that addresses this question. While ChatGPT answered that yes, it could translate handwritten documents, it actually does not allow for the upload of documents and does not have the ability to ‘read’ handwriting. There are other tools available that assist with handwriting analysis and can be found with a quick google search. But unfortunately, ChapGPT is not capable of doing this currently.
I have used ChatGPT to answer some historical questions during my researches. For instance, I wanted to know how far apart two people would have been if they were living on two particular named streets in Hobart in the 1880s. It gave me useful and contextual information that I could then verify by consulting some old maps. I have found that kind of question much easier to answer with ChatGPT than with Google.
I also asked it to help me learn about relationships and dna. For instance, I had some DNA relatives that were both related to my sample but not to each other, and I got it to postulate some explanations for that in the context of their known relationships.
I used Google’s Bard to ask about myself, it totally made up an answer about a guy in England with my name who supposedly worked for the same company as me, but was born in England, went to Oxford and died in 2014! I could not find any such person!
That is such a great thing to try! Turns out AI isn’t a substitute for real research! Thanks for sharing, Doug.
i have a question..i did a search on surname Cross in 1998 and i do not know how to interpret the 2 main records returned fron the search..here they are…
Cross 1798 1860 Cherry Valley ,NY USA baldy1
Cross 1798 1923 Cherry Valley,NY,>Maryland,OtsegoCo,NY USA baldy1
i know baldy1 is the researcher..i believe the second date is when the record was updated but what does the first date imply? birth or death? and what do the
locations imply?
Hi Don! Such great questions. Our team would be happy to help you. You can schedule a consultation here.
Hi Don! Such a great question. We’d love to help you out and get more information from you.You can schedule a consultation here.
I have used Google translate with images of German text from the 1800s. While it is not completely accurate, it is very helpful. It would be awesome if there was a way to have users who know how to read all the variations of old script to be able to upload the correct translation and create an AI index/dictionary for all the different handwriting samples. I would love to be able to contribute my knowledge and abilities to such an awesome work!
Agreed! It would be so helpful to have AI that could accurately translate variations of old German script. Who knows, maybe someone is already working on it!