One of the new FamilySearch Labs experiments is the "AI Research Assistant." It was briefly described by Andrew Gold in the FamilySearch Global and Tech Forum 2025 panel at RootsTech 2025.
1) When I opened the AI Research Assistant, it looked like this:
The screen offered me possible records for persons in my family tree profiles on the FamilySearch Family Tree on the left-hand side of the screen. Alternatively, it provides search fields on the right side of the screen for any other person or relative.
I chose to do the first one on the Research Assistant list on the left side of the screen - "Possible Child of John Rich (1819-deceased)." The screen showed me:
On the left side of the screen above is the top of the FamilySearch Family Tree profile for John Rich (GLFW-L5T, 1819-deceased). I can scroll down to see his wife and children and his parental family. It says he is married to Eliza (with no birth surname) and they have one child, Anne Rich born in 1852. It also indicates that this John Rich is the son of John and Rebecca (Hill) Rich, my 3rd great-grandparents. On the right side of the screen is information about the record and I can review and attach the record to this John Rich if I judge it to be correct.
2) I clicked on the link for the record and it showed me a summary of the record with a link to the record image:
This is a church baptism record of a 10 year old Thomas Rich, son of John Rich and Eliza --?--, baptized on 2 June 1859 in the Wingfield, Wiltshire church. Wingfield is about 5 miles west of Hilperton on the other side of Trowbridge.
If I had more information about the marriage of John Rich and Eliza --?-- then I would consider adding the record, and the child as a new profile, to the FamilySearch Family Tree.
3) But, and it is a big BUT:
This purported father, John Rich GLFW-L5T is a son of John Rich and Rebecca Hill, baptized in Hilperton, Wiltshire in 1819. I note that my own RootsMagic tree and the FamilySearch Family Tree has another John Rich (9KBD-BG5, 1822-deceased) who married Lydia Scott and had a family and all children were baptized in Staverton and Bradford on Avon in Wuiltshire.
Usually, when there is a second child with the same forename as another child, it's very likely that the first child with the forename died befoe the second child was born. I had no marriage or burial record for the John Rich GLFW-L5T and think it is very likely that child died before the second child named John Rich was born.
4) Another big BUT:
There is a second entry for John Rich (9KBD-BGG, 1818-before 1822) in the FamilySearch Family Tree as a child of John Rich and Rebecca Hill. So John Rich GLFW-L5T and John Rich 9KBD-BGG and John Rich 9KBD-BG5 are all listed as children of John Rich and Rebecca Hill on FamilySearch Family Tree at present.
A search on FamilySearch for John Rich baptisms reveal not only the 1818 and 1822 baptisms in Hilperton, Wiltshire as children of John and Rebecca (Hill) Rich, but also an 1820 baptism of a John Rich to Robert and Sarah (--?--) Rich in Bradford on Avon, Wiltshire, and an 1817 baptism of another John Rich to Stiles and Maria (--?--) Rich in Bradford on Avon, Wiltshire.
5) My conclusion in this case is that the John Rich GLFW-L5T who married Eliza --?-- and had Thomas is NOT the son of John Rich and Rebecca Hill, but is certainly the son of another Rich couple. John Rich GLFW-L5T should be removed as a child of John Rich and Rebecca Hill. Further research in the church records and censu records may identify the parents of John Rich GLFW-L5T.
6) This AI Research Assistant works fairly well, but the user needs to be careful and check the entire birth family and the spouse and children.
While this particular case was probably incorrect, it was chosen because it was the first case on the list, and not becuase it was incorrect. If it was correct, I would have concluded that.
Try it - it may work well for you on many of your ancestral families.
7) Some lessons learned:
- The AI Research Assistant is trying to help you find records for persons in the FamilySearch Family Tree that may be related to you.
- Be sure to click through to see the actual record - it may provide more clues for the family.
- Investigate other person of the same name in the same locality at the same time frame.
- Use critical thinking to evaluate the evidence at hand.
- Do more research in all databases (e.g. in this case, FamilySearch, Ancestry, Findmypast, etc.) to determine who every person is in the record at hand.
- Don't add the record to the persons in Family Tree unless you are certain that the record pertains to the persons claimed and analyzed.
The URL for this post is: https://www.geneamusings.com/2025/03/trying-new-ai-research-assistant-on.html
Copyright (c) 2025, Randall J. Seaver
Subscribe to receive a free daily email from Genea-Musings using www.Blogtrottr.com.
Thank you for your detailed review and explanations of FamilySearch's AI Research Assistant. I am on the team that developed this and wanted to help you and any of your readers better understand what it can do and solicit your feedback on what is/isn't working well for you.
ReplyDelete• First off, I’ll mention that there are kind of two products within the AI Research Assistant. The first presents up to 5 tree expanding hints from the home page. The second piece searches both FamilySearch sources and the web for a specific person. These sources along with the data available on the person page provide the basis for being able to ask questions about this person in an AI Chatbot. This second part is available on any deceased person within the tree, not just those presented as tree expanding from the home page.
• One of the features I believe genealogists will love once they realize what it is doing (and we continue to make it better) is the ability to see all of the sources found for a person. We believe this is going to save researchers time because it is doing multiple searches at once and filtering the results. Once you ask a question, you’ll see an All Sources button which will allow you to see everything the AI Research Assistant found for the person. I wanted to attach a screenshot, but apparently I can't in a comment.
• Of course, sometimes it still makes mistakes and includes incorrect sources. We hope to have the ability to dismiss incorrect sources soon.
• Another feature several have asked for that I lobbied to get in too is to show which of these sources is already attached. I also think we should show which ones are already being shown as Record Hints so it’s more obvious which ones are new.