AI (Artificial Intelligence) is all the “rage.” Therefore everyone feels pressure either to use AI or explain why they don’t use AI. You would have thought that genealogists who study history from 200+ years ago would be more immune to the latest trends, but evidently we are not.

So here is my take on AI and genealogy. As I see it, AI can help in the three main tasks facing genealogists: Searching for records; Attaching records to a family tree or using records to add people to your pedigree; and Writing family histories, books, or research reports (that last is perhaps more common among professionals). So, where does it make sense to use AI?

The first thing to note is that we have been doing genealogy for long enough that AI is not recreating or redefining the field or best practices in it. We have too much human experience to let AI tell us what to do. At best, using AI speeds up what we do, at worst, it interferes or causes errors.

Let’s start with Searching. For most people, genealogy is synonymous with the search for records of ancestors. It is in reality only a part of the story (more about that later). The search for images, text or documents which might shed light on those missing ancestors is one area in which AI, properly managed, can speed up a search. As one example, FamilySearch uses AI tools to conduct full-text searches on over 2.5 billion (yes, billion) as yet unindexed historical records in the FamilySearch digital repository. (Check out a third-party description here). AI search tools can in fact speed up a genealogist’s search through old records.

But before we get too excited, remember a couple of facts. First, AI can only read what is on the internet, and many essential sources for genealogy have not yet been digitized to be discovered online. Second, AI was trained first on digital files, then typed documents, and has only recently been taught formal American cursive writing, not to mention variations in individual handwriting, other styles of cursive, or records in other languages. Yes, AI models will be taught to decipher handwriting in many languages, but that day is not yet.

So, it does seem efficient to include AI as a search partner, along with traditional offline methods.

Next, one might be tempted to use AI to build a tree, create entries on a pedigree chart, or even attach records to a tree. James Tanner, in his blog Genealogy’s Star, describes an instance where a computer generated record hint, if accepted, would have introduced many errors into his family tree. A person needs to create a pedigree based on found documents, add individuals to the tree only after carefully weighing the evidence. Do not leave this to AI, which has been known to hallucinate plausible stories. Just say no.

Finally, one might consider using AI to write a research report, a family history article or a book. However, because of this tendency to hallucination, everything that AI generates must be double checked by a person. In the end, it is more efficient to allow computers to aid in the searching, but leave the creation (of books and of pedigrees) to humans.

In my own work, I try to be more efficient by leaning on search engines powered by AI for finding records. But clients pay for a correct pedigree and an accurate report, and therefore, I do not use AI for either of those steps. My clients deserve a human intelligence to assemble a human family history.