When an AI Audio Overview Found an Error in My Family Tree

Sometimes the most valuable result from using AI isn't a beautifully written story. Sometimes it is discovering that something in my genealogy database is wrong.

That happened while I was working on an ABC Biography for Rachel Christy Ricketts.

As part of my process, I had gathered information about Rachel from my RootsMagic database and the records I had collected for her. I then used that material with Google NotebookLM to explore different ways of turning genealogical research into a story.

One of those experiments was an Audio Overview.

Listening to Rachel's Story

NotebookLM's Audio Overview takes the sources in a notebook and creates a podcast-like discussion of the material. Instead of simply reading my genealogy report back to me, the generated discussion attempted to interpret the records and explain Rachel's life.

As I listened, something caught my attention.

There was a problem with Rachel's birthplace.

The birthplace recorded in my genealogy database did not agree with the evidence being discussed.

That was exactly the kind of thing I hoped this ABC Biography process might uncover—but I wasn't necessarily expecting an AI-generated audio program to be the thing that called my attention to it.

A Reason to Go Back to the Sources

Of course, an AI-generated statement isn't evidence.

When NotebookLM raised the issue, the next step wasn't to change my family tree simply because AI said something was wrong. Instead, it gave me a reason to return to the underlying records.

I reviewed the information I had for Rachel and determined that the birthplace in my file was indeed incorrect.

I have now corrected it.

What interests me most is that I don't know when—or whether—I would have noticed the error otherwise.

The information had been sitting in my files. I had looked at Rachel's records before. I had created a narrative report about her. Yet the incorrect birthplace had survived.

It took presenting the same information in another form for the discrepancy to become obvious.

Why Did the Audio Overview Help?

I think part of the answer is that we become accustomed to the information in our genealogy databases.

When I see a RootsMagic fact repeatedly, it can begin to feel established simply because I've seen it so many times. A narrative report may contain the same information, but I'm still looking at my research in a familiar format.

The Audio Overview changed that.

Instead of reading facts from a database or report, I was listening to someone discuss Rachel's life and the evidence for it.

That different presentation caused me to hear a conflict that I had apparently overlooked when reading the same information.

In that sense, NotebookLM wasn't replacing genealogical research. It was acting almost like another researcher sitting across the table saying:

"Wait. Does that birthplace fit with the records?"

That is a useful question.

Experimenting with the Audio

My experience didn't end with the Audio Overview.

I became curious about whether I could create an outline from the audio program itself. That led to a little experimentation.

I downloaded the Audio Overview from NotebookLM and tried uploading the resulting audio file into a new notebook. NotebookLM produced a paragraph summary rather than the outline I had hoped for.

At first, I considered that a failure.

Eventually, however, I realized that the summary contained the information needed to create an outline. More importantly, I started experimenting with some of NotebookLM's other Studio tools.

I created a data table, a slide deck, and a briefing document.

Each handled the same source material differently.

The data table essentially turned Rachel's life into a chronology. Much of it resembled the information in my original RootsMagic narrative report, only reorganized into a timeline.

The slide deck provided a visual interpretation of the sources.

The briefing document was particularly interesting. It analyzed the sources, identified the discrepancy in Rachel's information, and produced a detailed outline of her life.

In other words, I had taken essentially the same collection of genealogical information and viewed it as a narrative report, an audio discussion, a summary, a timeline, a visual presentation, and an analytical briefing.

And the inconsistency kept surfacing.

AI as Another Set of Eyes

This experience reinforces something I've been discovering throughout my experiments with AI and genealogy.

AI is useful not just for creating things from our research. It can also help us question our research.

That doesn't mean AI is automatically correct. Quite the opposite. Anything it says about our ancestors still needs to be checked against the original sources.

But asking an AI tool to organize the same evidence in several different ways can expose things we have overlooked.

A timeline may reveal that two events couldn't have happened in the order we thought.

A summary may highlight a contradiction.

A narrative may reveal a missing piece of the story.

An audio discussion may cause us to hear something that we didn't notice when we read it.

And an analytical report may put conflicting evidence side by side.

The Unexpected Benefit of My ABC Biography Project

When I began working through my ABC Biography process, my primary goal was storytelling. I wanted to use the records I had accumulated to tell more complete stories about the people in my family tree.

Rachel Christy Ricketts has reminded me that there is another benefit.

The process is also a form of research review.

By gathering the records, creating a narrative, feeding those materials into tools such as NotebookLM, and examining the resulting stories and analyses, I'm forcing myself to reconsider information that may have been sitting unquestioned in my database for years.

In Rachel's case, that process uncovered an incorrect birthplace.

That's a small correction in a large genealogy database.

But it also makes me wonder:

What else is hiding in plain sight in my files?