ESPN Drops AI Tells Tool for 2026 WSOP Final Table as $10 Million Prize Looms

Q: Why did ESPN abruptly pull its controversial AI tell-reading tool right before the $10 million WSOP final table?

ESPN and its production partner pulled the AI tell-reading graphic before the WSOP Main Event final table because the tool had become too controversial and too unreliable for poker’s biggest stage. The system, built by engineer Luke Geel, analyzed broadcast video for eye movement, blinking, posture, chip handling and similar behaviors, then estimated whether a player likely had a strong hand, draw or bluff. But poker professionals and analysts said that was a weak foundation for real tell detection. Only a small number of tables were filmed, most of the 9,000-plus entrants were not shown often enough for meaningful pattern-building, and many important live tells, such as breathing, pulse and speech, are not captured well on a TV broadcast. Players also argued that body language on camera does not cleanly map to hand strength. A player can look nervous or confident because of the pressure of the moment rather than the cards they hold. Geel himself acknowledged the model’s limitations and said prior testing at other poker events produced mixed results. In that context, using the graphic at the final table, where $10 million was at stake and scrutiny is highest, risked misleading viewers and undermining confidence in the broadcast. Omaha Productions, which works with ESPN on the WSOP, said the feature would not be used for the final table. The move suggests ESPN treated the tool as an early-stage experiment or novelty, not a trusted competitive or editorial instrument.

2 articles · Updated · WIRED · Aug 4

Omaha Productions said ESPN’s AI “tells detection” feature will not appear during the 2026 World Series of Poker Main Event final table after being used in early-July broadcasts.
The tool analyzed camera feeds from 3 recorded tables, tracking eye movement, blinking, posture and chip handling to estimate whether a player held a strong hand, draw or bluff.
Poker pros questioned its reliability because the model was trained on limited footage from a field of more than 9,000 entries, leaving too little repeat data on most players.
Michael Gagliano, one of the final-table players chasing the $10 million top prize, said even exhaustive review of broadcast footage yielded little actionable tell information.
The decision underscores broader skepticism that camera-based AI can match live tell reading in poker, even as players expect similar tools to keep improving and face tighter rules.