ESPN's New AI Snitch Will Watch Poker Players Blink

Peyton Manning Hired a Robot to Stare at Poker Players

Here's the setup: Peyton Manning's production company, Omaha Productions, is handling the WSOP broadcast for ESPN. And instead of just putting cameras on the felt and calling it a day, they brought in an independent AI engineer named Luke Geel to build a system that analyzes player behavior, including posture and blink rate, to flag moments when someone might be bluffing or holding the best possible hand. According to ZME Science, the tool was developed specifically for this broadcast.

Geel's system doesn't just watch the game live. It compares what it sees against historical footage, looking for repeated physical patterns across previous hands. A blink. A shift. A change in stillness. The AI cross-references those patterns with outcomes to build what is, essentially, a surveillance profile of each player's nervous system.

This is either the coolest thing to happen to poker on television since Chris Moneymaker turned a $40 satellite into a $2.5 million championship in 2003, or it is the opening act of something significantly more unsettling. Maybe both.

Before You Panic: They're Not Doing This in Real Time

To their credit, Omaha Productions is being cautious about how they deploy this thing. As ZME Science reports, the AI-assisted analysis will only focus on players after they've been eliminated from the tournament, avoiding any real-time exposure that could actually affect the outcome of a hand.

So the system isn't going to flash a graphic that says "BLUFFING" over a player's head while a pot is still live. It's going to be used as a post-hoc storytelling tool, a way to explain to viewers at home what was happening physically when a player made a big call or ran a massive bluff. Think of it like the Madden telestrator, except instead of drawing circles around receivers, it's charting the involuntary muscle movements of someone trying desperately not to look scared.

That is a meaningful distinction. For now. The word "for now" is doing a lot of heavy lifting in this story.

The Engineer Who Built It Isn't Even Sure It Works on Good Players

Give Luke Geel points for honesty. He told Sportico, as reported by ZME Science, that building the system was "significantly more difficult than I had initially hoped." He can't just upload a YouTube link and tell the AI to find someone's tells. The data has to be structured, labeled, and matched to outcomes across multiple hands and multiple sessions.

And here's where it gets genuinely interesting: Geel himself is skeptical the system can do much against the elite tier of the game. When people asked him why he doesn't just run the AI on Daniel Negreanu, one of the most decorated players in the history of the sport, Geel's answer was basically, "because it won't find anything." Negreanu, he told Sportico, has almost certainly worked hard to strip out any detectable physical tells. Geel said there's no point running the AI on a top pro yet.

So the technology is being showcased on players who are, by definition, not good enough to survive the Main Event. The people the AI can actually read are the ones who already lost. That is either a reasonable starting point for a developing technology or a very polite way of admitting the system has significant limitations, and ZME Science doesn't entirely resolve which.

This Is Bigger Than a Poker Broadcast and Everyone Knows It

Geel has already told Sportico that future versions of this kind of system could track signals like heart rate and facial flushness, things that go well beyond what any camera can currently detect in a broadcast studio. And he floated some of the other use cases himself: car dealerships studying what excites shoppers. Coaches profiling athletes. Anyone who wants to know what you want before you say it.

ZME Science draws a parallel to AI systems built for soccer that analyze a penalty taker's body movement before the kick is even struck, trying to give goalkeepers predictive data on direction. Researchers working on that problem are already cautioning that even with AI, penalty shootouts will never become fully predictable, because once players know what's being looked for, they start faking those signals.

That's the trap at the center of all of this. Surveillance changes behavior. In a game built entirely on deception, the logical response to an AI that spots your tells is to train yourself to generate false tells. Which means the AI has to get better to keep up. Which means the players have to get even better to beat the AI. At some point you're not watching poker anymore, you're watching an arms race between human consciousness and machine pattern recognition, and the poker table is just where it started.

Now think about what this looks like pointed at a debate stage. Presidential candidates, senators, anyone who has ever stood at a podium and said something they didn't entirely mean, already get analyzed by body language experts on cable news. But that's one pundit making guesses. A system like Geel's, trained on hours of a politician's past speeches and public appearances, cross-referencing posture shifts and blink rates against moments of known deception or documented reversals, would be something else entirely. Whether that sounds like a fantastic accountability tool or a nightmare depends entirely on who's running it and what they plan to do with the results.

The Political Angle Nobody in Washington Wants to Talk About

Here's a thought experiment worth sitting with. The 2024 presidential debate between Joe Biden and Donald Trump drew roughly 51 million television viewers, according to Nielsen. Every one of those viewers watched two men stand at podiums for 90 minutes and try to project certainty, confidence, and command. Cable news pundits spent days afterward dissecting body language, facial expressions, and verbal stumbles. All of that analysis was human, impressionistic, and largely unverifiable.

Now replace the pundits with a system like Geel's. Feed it every public appearance a candidate has made over the past decade. Let it build a behavioral baseline. Then run the debate footage through it and flag every moment where the candidate's blink rate, posture, or micro-expressions deviate from their established patterns. You'd have a machine-generated map of which answers made them nervous and which ones they'd clearly rehearsed until they felt nothing at all.

The implications don't stop at presidential debates either. Congressional hearings, press briefings, public testimony under oath, campaign rallies where a politician is telling three different audiences three slightly different versions of the same promise. All of it is recorded. All of it is analyzable. The question isn't whether someone will eventually build this. The question is whether voters will trust the output, whether campaigns will use it to train their candidates against detection, and whether the whole exercise ends up producing politicians who are better liars rather than more honest ones. Spoiler: history suggests the latter.

The Broadcast Itself Is Still Real Poker

Pulling back from the existential implications for a second: the WSOP Main Event has already begun on the ESPN app, with the final table scheduled for full ESPN coverage in early August. The buy-in is $10,000. This is real money, real cards, real players grinding through one of the most grueling tournaments in sports.

Omaha's head of content Dan Gati told Sportico that the goal is to recapture some of what made poker appointment television during the Moneymaker era, while reflecting how the game has evolved in the years since. Live poker, Gati says, is "as robust as it's ever been," despite largely disappearing from mainstream television. Bringing it back to ESPN is the bet, and the AI component is the hook designed to give casual viewers something to latch onto beyond just watching people stare at each other across a green felt table.

That's a fair enough pitch. Poker is genuinely fascinating once you understand what's happening beneath the surface. An AI that can surface some of that subtext for a general audience could make for great television, assuming it actually works and assuming Omaha uses it responsibly.