Mateo is nine, and this fall he decided he loves college football.
I did not see this coming. I've watched college football my whole life—weekends in the fall are basically built around it in our house—but I never pushed it on him. He just walked into the living room one Saturday, sat down next to me, and asked who was winning.
That was in September. By October he was asking questions I couldn't answer fast enough. Why do they stop the clock? What's a redshirt? Why is that team ranked higher if they lost? Why does everyone care about that number next to the name?
Some of those I could answer. Some I answered badly. And a few I realized I'd never actually understood—I'd just been nodding along for thirty years.
So one Saturday morning, before the games started, I sat down with him and ChatGPT and we worked through his questions together. It was genuinely one of the better hours I've spent with him.
It was also a good reminder that ChatGPT will confidently explain things it doesn't quite have right.
Here's the honest breakdown—what it nailed, what it fumbled, and what I had to correct myself.
Why I Did This Instead of Just Explaining
Two reasons.
First, Mateo asks questions in rapid fire, and I don't have a good way to pause the conversation and look things up. ChatGPT doesn't mind being interrupted. I do.
Second—and this is the more honest reason—I wanted to see how it would handle a topic I actually know well.
I've written before about how ChatGPT is good at explaining things in plain English. But most of those examples were about documents I didn't understand. This was different. This was a topic where I could catch it being wrong.
That turned out to be useful. Not because it was wrong a lot, but because of how it was wrong.
What It Got Right
Let me start with the good, because there was a lot of it.
The Basics of the Game
Mateo wanted to understand downs and distance. I gave ChatGPT a simple framing:
"Explain how downs work in American football to a 9-year-old who has watched a few games but doesn't understand the rules yet. Use plain language. Keep it under 150 words."
What came back was genuinely good. Four downs to gain ten yards. If you get ten, you get four more. If you don't, the other team gets the ball. It used the idea of a "fresh set" of downs, which is exactly the right metaphor for a kid.
Mateo got it immediately. He then spent the next ten minutes applying it to everything he saw on screen, which was mildly exhausting but also great.
Why Rankings Exist
He wanted to know why some teams have a number next to their name. I asked for an explanation aimed at a kid, and it gave a solid answer: the number is a ranking, rankings are opinions voted on by people and computers, and a high ranking doesn't mean a team is guaranteed to win.
That last point was the one Mateo needed. He'd assumed the number meant "this team is better." Understanding it as an opinion changed how he watched the games.
Why the Clock Stops
I asked for a simple explanation of why the clock stops so much in college football compared to the NFL. It explained the first-down rule and the out-of-bounds rule reasonably well, and it noted that college football's clock rules are different from the pros.
That's correct and useful, and I wouldn't have thought to phrase it that way.
What It Got Wrong
Now the part that matters.
The Redshirt Explanation Was Half Right
Mateo asked what a redshirt was. It's a real question, and it's genuinely confusing.
ChatGPT's answer was that a redshirt player is one who sits out a year to preserve eligibility. That's true as far as it goes.
What it left out is the part that makes it confusing in practice—that a player can appear in a limited number of games and still keep the redshirt, that the rule has changed in recent years, and that "redshirt freshman" doesn't mean what it sounds like.
Mateo asked a follow-up: "So if he doesn't play at all, does he get an extra year?" The answer was a clean yes. The real answer is "yes, with conditions, and the conditions have changed."
I caught this because I follow the sport. If I hadn't, Mateo would have walked away with a simplified answer that was wrong in exactly the way that would embarrass him later.
The Transfer Portal Answer Was Outdated
This one was more serious.
Mateo asked why players switch teams so much. I asked ChatGPT about the transfer portal, and it gave an explanation that described the rules as they were a few years ago.
The basic idea was right. The specifics were not current. Rules around transfers, eligibility, and immediate playing time have changed substantially in recent years, and ChatGPT's answer reflected a version of the sport that no longer exists.
This is the big one. Rule changes in college sports happen fast, and a model trained on older information will confidently describe the old rules as if they're still in effect.
It Was Vague on Rankings
When Mateo asked how the rankings are decided, the answer got fuzzy. It mentioned polls and computers but didn't explain that there are multiple different rankings, that they disagree with each other, and that the one that matters most for the championship is a specific committee's opinion.
That's not exactly wrong. It's just not the answer to the question he actually asked, which was "why do these two numbers on the screen disagree?"
I answered that one myself.
How I Handled It
Here's what I actually did, in the moment.
I corrected as we went. When the redshirt answer was incomplete, I jumped in and added the missing pieces. When the transfer portal answer was outdated, I said so directly: "That's how it used to work. It's different now."
I stopped asking about rules that change. Once I realized the model's knowledge had a cutoff, I stopped using it for anything current—eligibility rules, conference alignments, playoff format. If it's changed in the last couple of years, I don't trust it.
I used it for the parts that don't change. How downs work. Why the field is a hundred yards. What a touchdown is worth. The fundamentals of the sport have been stable for decades, and ChatGPT explains those very well.
I let Mateo see me check it. This is the part I'm proudest of. When I caught something wrong, I didn't quietly fix it. I said, out loud, "That doesn't sound right—let's check it." He watched me question the answer instead of accepting it. That lesson is worth more than any football rule.
The Test Card
Task: Explain college football concepts to a 9-year-old, in plain language, on demand.
Prompt or workflow: Simple "explain X to a 9-year-old" prompts for foundational concepts. Manual verification for anything involving current rules, eligibility, rankings, or recent changes.
Starting conditions: A curious kid with rapid-fire questions, a parent with general knowledge but gaps, and no desire to pause the conversation to look things up.
Time before ChatGPT: Answering from memory, which was fine for basics and unreliable for anything specific. Sometimes "I don't know, let me look it up later," which meant never.
Time after ChatGPT: About 5 seconds per answer. Plus occasional verification time when something sounded off.
Editing required: Significant for rules-adjacent topics. The basics needed no correction. The current-events material needed a lot.
What went wrong: Two categories of error. (1) Incomplete answers that were technically true but missing the part that makes them useful. (2) Outdated answers presented with full confidence, specifically around rules that have changed in recent years.
Privacy notes: No personal information involved. Mateo's name wasn't included in the prompts—I referred to him as "a 9-year-old." No school, location, or family details.
Who should not use this method: Anyone relying on it for current rules, eligibility requirements, recruiting information, or anything that changes season to season. Anyone who won't verify—the confidence of the output doesn't match its accuracy on fast-changing topics.
What This Actually Taught Me
I went into this thinking it would be a nice way to answer kid questions. I came out of it with a clearer picture of where ChatGPT's knowledge has edges.
It's excellent on stable concepts. The rules of the game, the shape of the field, the basic structure of the sport—all of that has been the same for a long time, and ChatGPT explains it well.
It's unreliable on anything that changes. Rules, eligibility, rankings systems, conference structures. It speaks about these with the same confidence as the basics, and that's the problem.
It doesn't know what it's missing. The redshirt answer wasn't wrong so much as incomplete, and incompleteness is harder to catch than error. You have to already know a topic to notice that something's been left out.
The confidence is the same either way. That's the sentence I want Mateo to remember, and honestly it's the sentence I want to remember too. ChatGPT sounds exactly as sure when it's right as when it's describing a rule that changed two years ago.
The Unexpected Part
Here's the thing I didn't expect.
Mateo now asks me, "Is that from the computer or do you know it?"
He picked up on the distinction without me explaining it. He watched me question an answer, and now he questions answers.
That's the actual outcome of this experiment. Not that he learned the rules of college football—he would have learned those anyway. It's that he learned that a confident answer isn't automatically a correct one.
That's a skill I'd like him to have long before he's old enough to use these tools himself.
The Bottom Line
ChatGPT is a genuinely good explainer for things that don't change. It's an unreliable one for anything current, and it won't tell you which is which.
Use it for the fundamentals. Verify anything that sounds like it might have changed recently. And when you catch it being wrong, say so out loud—especially if a kid is listening.
I tried it on my couch so you don't have to. Useful beats impressive, and knowing the difference between the two is the whole skill.
