Comments on the FamilySearch AI Research Assistant

FamilySearch.org has integrated several advanced artificial intelligence tools to transform how we approach historical research and family tree development. The FamilySearch AI Research Assistant is actually part of a multi-pronged AI ecosystem deployed on the site, shifting the platform from passive record hosting to proactive research collaboration. See AI Developments in Genealogy See also Introducing Tree Extending Hints from the AI Research Assistant

 When you sign into FamilySearch.org, the AI Research Assistant will appear on your startup page. There are immediately some initial questions. Is this an actual AI research assistant? How is it different from record hints? How accurate is it? Should I spend my time working on the suggestions or on my own research? None of these questions is particularly easy to answer. Let's look at how it works and then see if any of the questions are answered. 

The image above is a screenshot of my startup screen when I signed on to write this blog post. The first name on the list: Emma Gregersen G6JY-P5S is not at all familiar to me. It has been some time since I have done my own Danish research. Here is the entry I see when I click on that on the first selection. 

 I am familiar with the Danish census records, and so I decide to view the record. 

 I can now view the original document.

 I can see that it is very likely that Emma is the daughter of Ole Poulsen-Gregersen. I could proceed and attach her to the tree, but at this point I have absolutely no idea how I am related to her. This answers one of the questions that I posed earlier. This AI research assistant is nothing more than a finding aid. It has no interactive capabilities. I can't ask it any questions. I can't ask it for explanations. I am no longer actually in touch with the AI research assistant. If I use an AI chatbot such as Google's Gemini to create a research assistant, the assistant is interactive and will answer questions, etc. 

 I can click on Emma Gregersen's name and go to her profile page or person page. 

 I can then click on View Relationship and see if I am actually related

 I cut this off on my grandfather, but it appears that I am related through a line extending from my great-grandfather. This line was actually established by DNA from MyHeritage.com. My grandfather, Marinas Christiansen, was adopted, and we finally identified his biological parents. The chart shows that I am related to Emma only through her husband. This means that I am not likely related to Emma's parents. I have not researched this line, and I have no idea if each of the connections shown is accurate. Danish names are extraordinarily common, and therefore a name match is unreliable. Now I am faced with another of the questions I asked earlier. Am I going to keep doing research on this one person on this line, or am I going to return to do the research that I am presently doing on other lines? Another question would be, is this information correct? Check for additional information about her husband Lars Marinus Frederiksen G8H1-F2Z marriage was supposed to have happened in Copenhagen, Denmark in 1936. Lars was born, apparently, in Hals, Ålborg, Denmark. This is 359 km from Copenhagen. The chances of him actually marrying a person in Copenhagen seem remote. This is especially true because Emma is in the census record for Sundby, Mors, Denmark. Sundby is 130 km from Hals. A quick search in FamilySearch historical records shows tens of thousands of women named Emma Gregersen. There are no records that show this relationship could be even possible. Before I would be confident in accepting this suggested record, I would have to do considerable research back on the family line, but even this initial research shows that both these individuals are highly unlikely to have been related to my family line. 

At this point, I am fairly well convinced that any further research would be futile. The first question about record hints is now answered. This is less accurate than most of the record hints I see. Question: Should I change and begin doing research now in Denmark based on this record hint? The answer is clearly no. Lastly, how accurate is it? It appears to be entirely a same name, same person issue. Am I going to continue looking at the rest of the AI research assistant suggestions? Not likely. 

Just because you stick AI in front of Research Assistant does not make the results any more plausible and likely than a record match or record hint.