Saturday Night Genealogy Fun

My Oldest Confirmed Record

This week's Saturday Night Genealogy Fun assignment asks:

Calling all Genea-Musings Fans:
It's Saturday Night again — time for some more Genealogy Fun!!

Here is your assignment if you choose to play along (cue the Mission Impossible music, please!):

1) What is "Your Oldest Confirmed Record?" What is one of the oldest documents you've personally found for any ancestor?

At first, I thought this would be an easy question to answer.

My immediate reaction was that the record was probably associated with either my HAMMOND or WHITE ancestors. Both families extend well into the 1600s in my RootsMagic database.

But how was I going to determine which record was actually the oldest?

With thousands of people and many more facts and sources in my database, clicking backward through generations wasn't very appealing. So I decided to see whether ChatGPT could help me find the answer.

Searching My GEDCOM with ChatGPT

I exported my entire RootsMagic database as a GEDCOM file and uploaded it to ChatGPT. I initially asked whether there was a way to search an online tree or GEDCOM to determine my oldest confirmed record.

There was.

Since I suspected the Hammond and White families, I asked ChatGPT to start there. It searched the GEDCOM for those surnames, examined their dated events, and looked at the sources attached to those events.

Almost immediately, it found a very early candidate:

2 September 1603 — the christening of Thomas Hammond in Melford, Suffolk, England.

Wow! Had we found my answer?

Not quite.

A Date Is Not Necessarily a Document

When we looked more closely at the citation, we discovered an important distinction.

My source for Thomas Hammond's 1603 christening is Frederick Stam Hammond's History and Genealogies of the Hammond Families in America, published in 1902–1904.

Thus, I have a source for the date, but I don't have a copy of the 1603 parish record. Nor can I say that I personally found that original record.

The search of my WHITE ancestry produced a similar issue.

My ancestor Elder John White has a birth date recorded as between 1595 and 1605. He also has some wonderfully early events in my database, including his departure from London aboard the Lyon on 22 June 1632 and his arrival at Boston on 16 September 1632.

However, one of my sources for those events is Lucius M. Boltwood's Genealogies of Hadley Families, published in 1862.

Again, that's evidence supporting the event—but it isn't the 1632 document that I personally found.

Changing the Question

At that point, ChatGPT and I were really dealing with three different questions:

What is the earliest date in my family tree?

What is the earliest event in my tree supported by a source?

What is the oldest historical document that I have personally found for an ancestor?

Those aren't necessarily the same thing.

And the third question is what this week's Saturday Night Genealogy Fun challenge actually asks.

So, I asked ChatGPT to change its approach. Instead of simply looking for early dates with citations, I asked it to search the GEDCOM for pre-1700 events whose citations looked like original or official record sources.

That search uncovered several possibilities.

One was the 22 December 1652 marriage of Joshua and Joan Coggeshal, which I have sourced to Rhode Island vital records.

Another was a record in the Hammond family:

17 December 1662 — the marriage of Thomas Hammond and Elizabeth Hammond in Newton, Massachusetts.

My RootsMagic database contains multiple sources for that marriage, including Newton vital records. Beginning in the 1660s, I also have a series of Hammond births documented in Newton vital records.

Thus, my answer isn't the enticing 1603 date I initially thought it might be. That date is in my database, but I haven't personally found the underlying parish record.

Instead, based on the citations in my RootsMagic database, one of my oldest documented records appears to be the 1652 marriage of Joshua and Joan Coggeshal in Rhode Island, with my early Hammond records following close behind.

But Perhaps the Search Was the Real Genealogy Fun

What I found particularly interesting about this challenge wasn't just the answer. It was figuring out how to find the answer.

Without ChatGPT, I probably would have opened RootsMagic, navigated to some of my earliest ancestors, and started examining facts and sources one person at a time.

Instead, I exported my tree to GEDCOM and asked ChatGPT to analyze it.

ChatGPT searched the GEDCOM, identified early events, compared the Hammond and White families, examined the types of sources attached to those facts, and helped me eliminate dates that were supported only by later compiled genealogies. We then refined the search to look specifically for pre-1700 events associated with sources that appeared to be official records.

In other words, AI didn't prove the genealogy.

It helped me find where to look.

And perhaps that is one of the most useful roles AI can play in genealogy. My RootsMagic database contains years of research. The information was already there. ChatGPT didn't discover a new 1652 marriage record for me—it helped sift through the information I had accumulated and identify which records deserved another look.

There's also an important lesson here about our genealogy databases.

The oldest date in our tree isn't necessarily our oldest record.

And the oldest sourced date isn't necessarily the oldest document we've personally found.

Sometimes answering a seemingly simple genealogy question requires us to stop and ask exactly what the question means.

Thanks to this week's Saturday Night Genealogy Fun challenge, I not only went looking for my oldest record—I learned something about the records and citations already hiding in my own database.