Is ChatGPT reliable for locating Connecticut genealogical records?

ChatGPT has become a popular topic of discussion in genealogy fields. Many genealogists are trying to use it locate historical context or determine where to look for ancestral records. But does it actually work?

I ran a test case using the free version as of 25 June 2025.

Some of this material is right. Some of it isn’t.

According to the Connecticut State Library’s Town and Counties guide, Essex did split from Saybrook in 1852. But you’ll note that it tells me to look in the land records of the Saybrook Town Clerk (actually the correct response) and then send me to the town clerk of Old Saybrook.

I try to prompt it to clarify.

And it now gives me a totally wrong answer: “Old Saybrook Town Clerk (they hold the early Saybrook volumes).”

ChatGPT has seen Old Saybrook more frequently than Saybrook, so as a result seems to try to predict Old Saybrook for any response that should contain Saybrook. In reality, the contemporary town of Old Saybrook actually split from contemporary Essex, which in turn had split from Saybrook. (See the state library guide for details…)

So, where are the Saybrook Town Clerk records? The town of Saybrook… which in 1947, changed its name to Deep River.

In other words, use ChatGPT with caution for this type of question!