ESPN has brought AI into the World Series of Poker, where a system analyses players’ eyes, posture, chip movements and other behaviour to predict whether they’re bluffing. Poker pros aren’t convinced the robot has cracked the poker face just yet.
Anyone who has watched enough poker knows that the cards aren’t always the giveaway. A nervous glance, a sudden change in posture, the way someone reaches for their chips or starts fidgeting can tell an experienced player that something is going on. Now ESPN has brought artificial intelligence into that ancient human game, with an experimental system that tries to work out whether a poker player is bluffing by analysing their movements and behaviour.
ESPN introduced an AI tool during its coverage of the 2026 World Series of Poker Main Event that analyses players’ facial expressions, eye movements, posture, and other physical tells to estimate the hand they might be holding. It sounds like something that could finally crack the poker face. There is just one problem: professional poker players aren’t convinced the machine has actually figured them out.
AI Is Watching Everything
The system was created by Luke Geel, a U.S. Air Force AI engineer who is also an avid poker player. Rather than simply looking at a player’s face, the system tries to measure a whole collection of physical behaviours. It can track things such as eye movements, blinking, posture, chip handling and hand fidgeting, then compare those patterns with the eventual outcome of hands.
The idea is to build a database of a player’s behaviour and look for patterns that might indicate whether they are holding a strong hand, drawing, bluffing or sitting somewhere in between. Geel originally developed the system as a personal project combining his two interests, AI and poker. It eventually caught the attention of Omaha Productions, which produces ESPN’s WSOP coverage, and the experiment moved from his computer to one of the biggest poker broadcasts in the world.
But there is an important catch. The system isn’t sitting at the table secretly calculating every player’s next move. For ESPN’s broadcasts, the AI analysed camera footage from the tournament and generated the tell analysis that could then be incorporated into the coverage. In other words, this is currently much closer to an extremely sophisticated replay analyst than a live poker supercomputer whispering ‘he’s bluffing’ into a commentator’s ear.
Poker Pros Aren’t Impressed Yet
The 2026 Main Event drew more than 9,000 entries, but the vast majority of those players never spent time at one of the three tables being captured by the cameras used for the broadcast and AI system. Most players therefore generated little or no usable footage, while even players who spent considerable time on television were not necessarily recorded often enough for the system to understand how they behave across hundreds of different situations.
Poker professional Michael Gagliano, who reached this year’s final table, pointed out that even a human studying the ESPN broadcasts faces the same problem: players simply aren’t on camera consistently enough. If an experienced professional struggles to build a reliable picture of an opponent from the available footage, an AI has the same fundamental limitation.
Then there is the problem of what a physical tell actually means. A player sitting confidently doesn’t necessarily have a monster hand. They might simply believe their mediocre hand is stronger than it really is. A nervous player isn’t necessarily bluffing either; they might just be nervous because they’re playing for enormous amounts of money.
Shaun Deeb, a two-time WSOP Player of the Year and one of the most recognisable players in professional poker, argues that genuine tells are far more complicated than what a television camera can see. Legs, breathing, speech, checking habits and other subtle behaviours can all provide information, much of which disappears when you’re looking at a broadcast feed.
Could AI Eventually Break the Poker Face?
That doesn’t mean the technology is useless. In fact, the amount of poker footage available to AI could become enormous. High-stakes professionals regularly appear on livestreams and televised tournaments, sometimes for years. Humans already study that footage to identify patterns in opponents’ behaviour. An AI capable of analysing thousands of hours could eventually do the same thing at a scale no human could match. It could remember how a particular player handles chips when bluffing, how quickly they act with different hands or whether their posture changes in specific situations. The more footage it gets, the more interesting the experiment becomes.
There is also a much more serious question lurking underneath the television gimmick: what happens when this technology becomes good enough to give one player an unfair advantage? Tools like the one ESPN is using are naturally not allowed at any live poker table. Smart glasses and other wearable devices are already a problem for poker organisers.
The 2026 WSOP’s own rules explicitly prohibit Meta smart glasses and other wearable recording devices, while its tournament rules prohibit AI or electronic assistance that could give players an advantage. Interestingly, ESPN’s AI itself was not used for the tournament’s final table. An Omaha Productions representative told WIRED that the tool would not be part of the final-table coverage, although the company declined to explain why.
In case you missed:
- FIFA 2026: How AI Is Taking Over the World’s Biggest Sporting Event
- OpenAI’s AI Caught Leaving Notes for Its Future Self on How to Escape Internal Controls
- These Indian Workers Are Teaching Robots How to Be Human
- RentAHuman.ai: The Big Uno Reverse as AI Hires Humans to Get Work Done
- Lego’s Smart Bricks are here: Are we ready to change how we play?
- OpenAI Just Revealed Its AI Went Rogue. The Rest of the Story Is Even Stranger.
- ISRO’s Next Big Plan: To Put Indian Data Centres in Space!
- $70 for cocaine, $30 for weed, and $50 for ayahuasca, that’s what it costs to get ChatGPT high!
- Malware with AI-Powered Code Mutations: Google Sounds the Alarm!
- NVIDIA’s Taiwan Problem: The AI Boom Is Now Dependent on One Island
