When someone tells me they've tried artificial intelligence and "it wasn't very good," my first question is simple:
"Did you customize it?"
Most people answer no.
That's a little like hiring a brand-new research assistant, handing them a box of family papers, and expecting them to instantly know your research methods, writing style, citation preferences, and standards of proof. Of course they won't.
Artificial intelligence works much the same way.
Think of It as Training a New Research Assistant
Every genealogist develops habits over time.
Perhaps you always cite your sources in Evidence Explained style. Maybe you write dates as "21 March 1850" instead of "March 21, 1850." You might insist that every statement be supported by evidence, or refuse to identify someone as a parent without a convincing proof argument.
None of that is obvious to an AI unless you tell it. The more it understands your expectations, the less time you'll spend correcting its work.
What Should You Tell It?
Each AI platform has a place where you can save instructions that will be used in future conversations. Although every platform calls it something slightly different, the idea is the same.
Consider including information such as:
• Your experience level in genealogy.
• Your preferred citation style.
• How you like dates written.
• Whether you want the AI to distinguish clearly between facts, assumptions, and hypotheses.
• Whether you prefer concise answers or detailed explanations.
• Your preferred writing tone.
• The kinds of records you most frequently research.
• Any formatting preferences you use consistently.
Think carefully before adding information. These instructions should describe how you want the AI to work with you, not every detail about your personal life.
My Own Instructions
Over time I've taught my AI assistant quite a bit about how I work.
It knows that I expect evidence-based genealogy rather than speculation. It knows I prefer military-style dates and Chicago citations because Evidence Explained is under copywrite. It understands that if a record doesn't support a conclusion, it shouldn't invent one. It also knows I write for an educated audience and prefer engaging narratives instead of dry reports.
As a result, much of what it produces already looks remarkably close to something I would write myself.
Is it perfect? No. I still verify every fact, examine every citation, and edit every piece of writing before anyone else sees it. That's exactly what responsible genealogists should do.
Your Instructions Will Change
Don't think of these settings as permanent.
As you discover new ways to work with AI, you'll probably update them.
I add new instructions whenever I notice I'm making the same correction repeatedly. If I find myself saying, "Please don't use em dashes," or "Always distinguish evidence from inference," that's usually a sign the instruction belongs in my saved settings.
Over time your AI becomes a better research partner because you've taken the time to teach it how you work.
One Important Reminder
Customizing your AI does not replace critical thinking.
AI can summarize records, organize timelines, suggest research strategies, draft narratives, and help identify conflicts in the evidence. It cannot determine whether a conclusion is genealogically sound.
That responsibility remains yours.
Think of AI as an exceptionally fast assistant, not an expert witness.
Teach it well, supervise it carefully, and you'll likely discover that it becomes one of the most useful tools in your genealogy toolbox.
In my next article, I'll show you exactly where to find these settings in ChatGPT, and I'll also compare the equivalent features in Gemini, Claude, and Perplexity so you can customize whichever AI assistant you use.
No comments:
Post a Comment