ESPN Used an AI Tool to Detect Poker Tells at the 2026 World Series of Poker. Pros Are Not Impressed.

Key points

• ESPN aired an AI tells-detection tool during the 2026 World Series of Poker Main Event broadcast in early July 2026.
• The tool was built by Luke Geel, an AI engineer for the US Air Force, and tracked eye movements, blink rate, posture and chip-handling to estimate hand strength.
• Over 9,000 players entered the 2026 Main Event, but only those at three filmed tables generated training data, leaving the AI with a very thin sample.
• Professional player Michael Gagliano, a 17-year veteran who reached the final table, says even he struggled to act on the footage, let alone an algorithm.
• ESPN's production partner confirmed the tool will not appear during the final table broadcast, without giving a reason.

For a few days in early July, viewers watching ESPN's live coverage of the 2026 World Series of Poker Main Event noticed something new on screen. A text overlay appeared on certain players showing live movement measurements, plus a chart labelled a "hand strength model" that guessed whether the player held a strong made hand, a drawing hand (one that needs a single card to become strong), or a bluff.

The tool was built by Luke Geel, an AI engineer who works for the US Air Force. His system watched every filmed hand at the three camera-covered tables and built what he calls a tells database: a record of each player's eye movements, blink rate, posture, chip handling and finger fidgeting, cross-referenced with the actual outcome of each hand.

Does it actually work?

Pros say the data is far too thin for the system to be reliable. More than 9,000 players entered the 2026 Main Event, but the vast majority never sat at a filmed table. Even those who did rarely stayed long enough for the AI to build a meaningful picture.

Michael Gagliano, a 17-year poker professional who made the final table and is competing for the $10 million top prize, spent the two-and-a-half-week break between the final table and the final broadcast reviewing every second of ESPN's streams, hunting for tells on his remaining opponents. He came away with little he could use.

"I don't know how much actual information I'm going to be able to act on from what I saw," Gagliano says. An AI working from the same footage faces the exact same problem.

Shaun Deeb, a two-time WSOP Player of the Year who finished 15th in the 2026 Main Event, points to a deeper issue. A camera can track visible signals. It cannot read intention.

"Physical tells are so much more expansive than I think the public realizes," Deeb says. "There are leg tells, checking tells, verbal tells, breathing tells, pulse tells. Most of those can't be picked up by a camera."

Body language also shifts with context. A player who looks nervous might be reacting to the stakes, not the strength of their cards. As Gagliano puts it: "Maybe my body language is referencing the situation rather than the hand strength."

That gap between visible signal and actual intention is exactly what our 27 July piece on why AI vision systems miss what's right in front of them explored: pattern recognition breaks down when context is stripped away.

What does this mean for anyone watching at home?

For casual viewers, the overlay is broadcast entertainment, not a reliable guide to what is happening at the table. Geel himself told Wired he has run blind tests on other competitions and got mixed results, and he acknowledges a larger sample of hands would improve the tool.

Deeb is blunt about the broadcast version: "I think they randomly found something to try to make it like another sport, and I just think it was swing-and-a-miss."

ESPN's production partner, Omaha Productions, confirmed the tool will not be used for the final table. No reason was given.

Longer term, skilled players already study hours of recorded footage on regular opponents. As AI image analysis improves, that process could get faster and more detailed, particularly at high-roller events where the same small pool of professionals play each other repeatedly on camera. For now, Deeb still trusts a human observer watching in person over any camera-based system.

"We always had a spot for the person spotting the tells to be watching the player in person," Deeb says, adding that a live watcher "was much better than what you get on TV."

The honest read on this experiment: ESPN got a slick-looking overlay and not much else. The tool's real limitation isn't technical, it's that poker tells only mean something when you already know a player deeply. A database built on a handful of filmed hands can't get you there, and the pros know it.

Common questions

Could this kind of AI tool be used to cheat at live poker?

No. Tools like this are not permitted at any live poker table, and the system only runs on broadcast footage after the fact. It has no way to feed information to a player during a hand.

Is the AI reading minds?

Not even close. It tracks visible patterns like blinking and posture. Whether those patterns mean anything useful depends on far more context than a camera can provide, and professional players say human experts watching in person are still far more accurate.