Am I Chasing Rabbit Trails or Following the Records?

Asking ChatGPT to Critique My Research Style

I imagine every family historian knows the feeling. You sit down determined to finally focus on one ancestor…just one…and within twenty minutes you’re knee-deep in a different surname, chasing a witness on a deed, a neighbor in a census, or an in-law mentioned once in a probate file. Your browser has fifteen tabs open, your notes are branching in four directions, and you start to wonder whether you’re actually making progress or just indulging in well-documented procrastination.

Am I actually making progress…or am I just bad at staying on task?

That question has been nagging at me lately. Not because I lack ideas or records, but because I worry about discipline. About sticking with one line long enough to resolve it. About knowing when a tangent is necessary and when it’s just…interesting.

I suspect I’m not alone in this. Many researchers quietly worry that jumping between families means they aren’t serious enough, focused enough, or methodical enough. We hear phrases like “stay in your lane” or “finish one line before starting another,” and they sound right…until the records refuse to cooperate.

So I decided to do something a little unusual. I asked my AI research assistant to critique me. Not to tell me how great my work is. Not to smooth things over. But to be honest.

For context, I don’t treat AI as gospel. I correct it regularly. I reject conclusions that don’t hold up. I use it as a thinking partner, a sounding board, and sometimes a mirror. This time, I wanted the mirror.

I first asked my AI assistant to give me a one-sentence summary of the last 10 surnames I had been

researching in my family tree and to include the big idea/question I was researching. The AI response:

• Peterson (paternal): Determining whether later generations of Petersons are correctly attached to John and Ruth Peterson or whether DNA and documents point to a different relationship entirely.
• Morehead (maternal): Testing whether Alexander Morehead’s family connections, especially Thomas and Calvin Morehead, represent sibling relationships or part of a broader FAN (friends, associates, neighbors) network in Fairfield and surrounding counties.
• Thurman (paternal): Reconciling records that alternately name Daniel Thurman and John Thurman to determine whether they are the same man or closely related individuals across Tennessee, Alabama, and Arkansas.
• Alexander (maternal): Establishing the children and lineage of James Alexander and Jerusha through pension records and indirect evidence in a period with sparse civil documentation.
• Thompson (paternal): Using agricultural and census records to understand Rodden (Roden/Rodden) Thompson’s economic standing and confirm identity consistency across Arkansas records.
• Oglesby (paternal): Clarifying the identity, age, landholdings, and family structure of Constant Oglesby amid variant spellings and early tax and land records.
• Mezo (paternal): Tracing Moses Mezo and Molly (Mary) Oglesby to confirm marriage, migration patterns, and correct surname usage amid repeated misattributions.
• Hancock (maternal): Reconstructing family relationships and land associations through deeds and neighbor analysis to separate documented facts from later assumptions.
• Lawson (paternal): Testing land and tax records to confirm whether similarly named Lawsons represent one individual over time or multiple related men in the same locality.
• Gooding (paternal): Sorting out sibling and in-law relationships among the Gooding family to understand how marriage ties explain beneficiary and witness patterns in early probate documents.

I asked this because I am concerned that I veer off too much and don’t stick to one line long enough. 

Since I have been using AI for about a year now to really help me organize, analyze, and plan research questions, I wanted to know what the AI bot thought of my research and to be honest.

The following is its response: 

I’ll be straight with you, but fair. You do not come across as unfocused or scattershot. What you’re doing looks messy on the surface, but methodologically it actually makes sense for the time periods and problems you’re working in.

Here’s the honest tension I see.

On the risk side:
Yes, there are moments where you pivot quickly…from Peterson to Thurman to Alexander to Morehead…and if someone only skimmed your work, they might think you’re chasing too many threads at once. That can dilute momentum if you’re not careful, especially emotionally, because none of these lines resolve quickly.

On the strength side:
Your pivots are not random. They are triggered by evidence. A probate file points sideways, a census conflict forces a detour, a DNA match drags another surname into the room. That’s not veering…that’s following the record set where it actually leads. Early 19th-century research is inherently collateral.

The deeper truth:
You’re not researching surnames. You’re researching problems. Each surname is a different angle on a single question…who belonged to whom, and how can we prove it without modern records. When you frame it that way, your work is actually disciplined, not scattered.

Where I think a small correction would help:
You might benefit from declaring “primary” and “supporting” lines at any given time. For example, Peterson is the primary problem; Gooding, Thurman, or Alexander are supporting investigations that exist only to resolve that core question. That doesn’t change your research…it changes how you mentally file it.

One practical suggestion that fits your style:
Before starting a new deep dive, write a one-sentence gatekeeping question like: “If I solve this, will it directly confirm, weaken, or eliminate a parentage hypothesis I already have?”
If the answer is yes, you’re on track. If not, park it.

Bottom line, honestly:
You are not drifting. You are working at the edge of what the records allow, and that edge forces lateral movement. The danger for you isn’t lack of focus…it’s burnout from juggling too many unresolved but legitimate questions at once.

End of the AI response.

Reading that response didn’t magically make my research simpler, but it did clarify something important. The problem isn’t that I explore too much. It’s that early records demand exploration. The danger comes when I forget to name which question is central and which paths exist only to serve it.

For me, using AI this way isn’t about outsourcing thinking. It’s about sharpening it. When I argue with it, correct it, or ask it to critique me, I’m forced to slow down and articulate why I’m doing what I’m doing.

And that, in the end, might be the most valuable research habit of all.