Is This Poker Player Deceiving? The AI Believes So

Any AI analyzing the footage would encounter the same challenges, even in a lengthy tournament like the Main Event.

Beyond Just Cookies

Interpreting tells is a complex task, and experts remain skeptical that a camera-based AI can outperform a human in this regard.

Most non-players first learn about poker tells from the climactic scene in the 1998 film Rounders. In this scene, Matt Damon’s character, Mike McDermott, folds an impressive hand to John Malkovich’s Teddy KGB after recognizing that he’s been outplayed—thanks to a tell related to KGB’s Oreo-eating habit at the table. This scene is arguably the most iconic poker moment in cinema, illustrating the mental duel that defines every poker game, even if it simplifies the reality.

“Referring to the Rounders Oreo tell, it’s actually more nuanced than that,” explains Shaun Deeb, a two-time WSOP Player of the Year and one of the game’s most prominent figures. (Deeb also made an impressive run in the 2026 Main Event, finishing in 15th place.)

“Physical tells are much broader than the public assumes,” Deeb asserts. “There are tells related to legs, checking, verbal cues, breathing, and pulse. The variety of tells is astonishing, and many can’t be captured by a camera.”

While AI can analyze visual and audio signals, it cannot interpret intention; the former holds limited value without the latter. Even if the technology hypothetically excels at identifying when a player exudes confidence or weakness, that insight alone does not reveal their actual hand.

“How does two pair rank for one player compared to another?” questions Gagliano. “A player might display confidence with a hand that’s weak, genuinely believing they hold the best hand.”

“In a casual tournament, I might perceive my two pair as very strong. However, in the Main Event, I’d feel more anxious due to the stakes, so my body language might reflect that context rather than just hand strength.”

For a broadcast tool, these limitations aren’t necessarily prohibitive. No one expects a perfect oracle—certainly not Geel, the tool’s creator. He openly acknowledges that a broader dataset of hands would enhance his tool’s effectiveness, sharing with WIRED via email that he’s conducted blind tests in other poker competitions with varying outcomes.

The feature may add some value for certain ESPN viewers, though players like Deeb still express skepticism. “I believe they randomly discovered something to make it resemble another sport, and I think it fell short,” Deeb states.

While viewers observed the tool during segments of the tournament broadcast in July, a spokesperson from Omaha Productions, which has an ESPN license for WSOP and other sports, noted in a text that the tool would not be utilized for the final table. The spokesperson did not provide reasoning for this choice.

Observing the Detectives

As AI technology advances, even skeptics agree that tools like these could develop quickly and be harnessed for profit. In poker’s “high-roller” scene, where buy-ins often exceed six figures, a select group of well-known professionals competes in events commonly broadcast. It’s likely that there are hundreds or thousands of hours of footage of these elite players, many of whom participate in numerous events annually. Poker players often analyze streamed and broadcast footage to gather intelligence on their regular adversaries. Could enhanced AI streamline this inherently human process?

Deeb, for instance, isn’t concerned. As a leading pro, he’s often hired to coach players during deep runs in the Main Event. He mentions that this coaching frequently involves bringing in a specialized live tells expert to observe both the opponents and the client (to identify any notable tendencies that need adjustment). A close friend of Deeb’s watched the streams during his recent run, performing a similar analysis on his behalf.