A Cautionary AI Tale

If you're using AI to get your sports news, you should stop.

This new school year, many students and faculty are adjusting to policies about using LLMs. Some schools and universities are encouraging their use. Some have become big clients – wave of the future! – and are pinning substantial parts of pedagogy on it. No one wants to be the last tradesperson teaching buggy whip manufacturing, after all.

In my classes, there’s a no LLM policy. Up-marketed word generation has its place, I say, but that’s not in this class. You’re here to learn, expand your mind and your critical skills. You don’t do that by mastering the fine art of prompt writing. Or, as it has often been said on social media in the last year, you don’t take a forklift to the gym to move those weights around.

No one knows who wrote that first. I’m not asking ChatGPT. I’ve asked it to do a few other things. For instance:

“Write 200 words to begin a newsletter looking back on the first weekend of college football. Don't add a conclusion.”

The pattern making machine happily spit out these anodyne sentences, each boring enough to be rejected by any cigar chomping copy editor still bitterly hanging on.

“Week 1 of the college football season did not disappoint. From shocking upsets to dominant performances, the opening weekend served as a reminder of why fans count down the days every offseason. Programs with high expectations showed flashes of greatness—while others revealed early cracks that could reshape the playoff picture sooner than expected. Across the country, stadiums roared back to life, and with them came the chaos, passion, and unpredictability that define college football.”

It did not disappoint! For precisely half the fan bases that saw their teams attempt a team sport for the first time in front of a large crowd against hostile opposition. This is why fans count down the days! And, also, the Dr. Pepper commercials.

And some teams had cracks, a tectonic allusion to Alabama, and Tide fans who are, right now, hoping their athletic director can find another way into the playoffs.

“Several headline matchups lived up to the billing, delivering late-game drama and breakout stars. A few Heisman hopefuls began building their campaigns in style, while unheralded names emerged seemingly out of nowhere, rewriting depth charts and stealing the spotlight.”

Several headline matchups lived up to the billing, sayeth the device, which obviously did not stay up for Monday’s UNC-TCU game. Maybe the follow-up question should be why that contest was booked in primetime. Did ESPN think we’d tune in for Bill Belichick? And why were they right? Why, ChatGPT, do we do that to ourselves?

Some things are unanswerable.

“Meanwhile, ranked teams didn’t have it easy—more than one found themselves in unexpectedly tight battles deep into the fourth quarter.”

That’s a sentence any space-filling hack could peck out before halftime.

It only gets more generic.

“Perhaps most exciting was the atmosphere itself. Packed student sections, rowdy bands, and emotional pregame traditions reminded everyone what was missing during the offseason. Week 1 offered everything: statement wins, coaching debuts, overtime thrillers, and goalposts coming down. It's only one weekend, but already, the narratives of the 2025 season are starting to take shape—and a few surprises are bound to stick with us for weeks to come.”

I also asked ChatGPT to summarize the weekend, because surely it can do that. It, of course, replied in the first person.

“I thought ...”

No. You didn’t. Also, you aren’t a you.

It immediately tried to tell me that Ohio State and The Most Definitely Back Longhorns are archrivals. I don’t expect a distributor of ones and zeros to know this, but five games played across 20 years does not an arch rivalry make.

Incidentally, Google’s Gemini got that right. The preceding is a sentence seldom uttered or typed.

ChatGPT goes on, trying to summarize a key point from random games. I didn’t ask for specifics, so it is guessing that I’d care about Michigan’s big day rushing against New Mexico. Justice Haynes tallied 159 rushing yards and three scores against the Lobos, a team that was 126^th in rushing defense last year. It presumes I also had a peculiar interest on Nebraska’s Dylan Raiola. It tells me he threw two TD passes against UTEP. But…the Cornhuskers played Cincinnati, not UTEP. (UTEP fell to Utah State.) Dylan Raiola is a QB at Lincoln, and he did throw two TDs, though. ChatGPT mentions Purdue’s 31-0 “statement win.” That was a 31-0 trouncing of Ball State. If that’s a statement in West Lafayette this year, the Boilermakers are in for another horribly long season.

The point I’d like to make here is that I randomly picked three of the bullets ChatGPT offered me. One is wrong on the face of it; another is lacking any of the nuance your football hating relative could have brought to the conversation. Also, I spent three minutes Googling all of that to check its work.

Time well spent, technologists, time well spent.

And this? This is pure insight.

“Betting markets swung heavily in Week 1 – fans had clearly been thirsting for action after the offseason lull.”

In ChatGPT land, there are no other sports, books, or prop bets to be had. Finally, football is here to sate the needs of the inveterate.

This second prompt wrapped up thusly:

“It was a reminder that in college football, Saturday is always king, and all the bold preseason takes go straight out the window when the games begin.”

Freshmen will write better than ChatGPT. If you challenge them. They’ll enjoy it more, too. Readers absolutely would.

I asked ChatGPT, “What should I look forward to in Week 2 of college football?”

It gave me 216 words of pablum and boilerplate. It was throat clearing, neatly slotted under subheadings like “bounce-back opportunities,” “trap games,” and “freshmen and transfer watch.”

And finally, “Would you like a preview of specific matchups once schedules are out?”

No, ChatGPT, you’ve done nothing, and that’s enough.