Bryson DeChambeau Bought an AI Golf App, Then Asked Gemini Anyway

Bryson DeChambeau shot 65 on Day 1 at Asiad Country Club in Busan, dropped to 71 on Day 3, then spent Saturday night asking Google Gemini about the physics of why his club wouldn’t turn over at impact. He came back Sunday, shot another 65, finished third at the LIV Golf Korea final leaderboard one stroke out of a playoff, and told reporters he still doesn’t have the answer.

He has made a career habit of mining technology for an edge. Before winning the 2024 U.S. Open at Pinehurst, he used an AI swing-analysis app to find and correct a slight rightward drift in his shots, then named the company at his winner’s press conference. In April, he bought it.

A Night on the Range in Busan

DeChambeau entered the tournament at Asiad Country Club sharing the first-round lead at 5-under 65, tied with Crushers GC teammate Charles Howell III and HyFlyers GC’s Scott Vincent. His 68 on Friday left him within range of the front, and with two rounds remaining his position looked solid.

Saturday’s round took his swing. From the opening holes of Round 3, his hands felt like they were getting ahead of his body, blocking the clubface from releasing through the hitting zone. He finished the day at one over par, a 71, then went to the practice range and hit balls for several hours. Nothing resolved it.

“I was slamming the club in the ground trying to figure out what to do,” he said after Sunday’s round. “I was frustrated. Been trying everything in my body. I didn’t actually figure it out on the range.”

Back in his room, he started a conversation with Gemini. The exchange covered the physics governing club rotation: alpha torque, gamma torque, what mechanical forces block a clubhead from turning over at impact, and how grip pressure and tension factor in. What the chat produced was simpler than the vocabulary required to get there. He needed to relax his grip and let the club fold naturally.

On Sunday he came out with what he called “a little bit more freer hands,” felt the club better, and shot the joint-best round of the day, a 65. He finished at 11-under total, third place, one stroke behind the playoff that Joaquin Niemann won, collecting the $4 million first-place share of the event’s $20 million purse. Travis Smyth, filling in for the injured Paul Casey on the Crushers GC roster, helped the squad win the team trophy for a record 10th time in LIV Golf history.

The swing wasn’t fully fixed. He still missed wedges to the right during Sunday’s round, what he described as “holding on a little rather than just letting it go.”

Eight Figures, a Bellevue Startup, and a Google Deal

On April 7, seven weeks before Busan, the investor group he leads closed an eight-figure deal to acquire Sportsbox AI, a Bellevue, Washington startup that analyzes golf swings using AI and 3D motion capture from smartphone video. The terms were confirmed to Bloomberg by DeChambeau. Co-founders Jeehae Lee, a former LPGA Tour player, and Samuel Menaker, a veteran AI engineer with over two decades of experience, stayed on along with the company’s roughly 30 employees.

Alongside the acquisition, the group announced SAMI, short for Sportsbox AI Motion Intelligence, a conversational coaching assistant built on Google Cloud’s Gemini models. The system converts a smartphone swing video into a fully rendered 3D model, analyzes thousands of biomechanical data points per swing, including measures of turn, bend, side bend, sway, and lift, and delivers plain-language coaching advice within seconds. Granville Valentine, Google Cloud’s managing director for North America Generative AI, called the collaboration “an example of how agentic AI can help democratize elite coaching.”

• 150,000+ users on the Sportsbox 3DGolf app at the time of the deal
• $15.99/month consumer subscription rate ($110 per year)
• 30 employees retained under the Sportsbox brand

What separates the platform from a general chatbot is the input layer. Sportsbox’s 3DGolf technology compresses what previously required six-figure lab equipment into a smartphone video, extracting biomechanical measurements frame by frame without wearable sensors. A chatbot starts from the golfer’s description of what the swing feels like. The Sportsbox platform starts from what the body is doing. As part of the Google Cloud partnership, the Google Cloud logo began appearing on his bag at Augusta and future events, reportedly the first time the brand has been on a professional golfer’s equipment. The coaching assistant was still in beta and beginning its Q2 rollout when Korea took place.

How a Late-Night Torque Discussion Got Mocked

After Sunday’s round, he described the overnight session in enough detail that the golf world took note. Andy Johnson, the founder of Fried Egg Golf and host of the Shotgun Podcast, framed the reaction plainly: a late-night deep-dive into torque physics had produced a tip any caddie would hand out for free.

I spent some long hours on the range trying to figure some stuff out and I was talking to AI quite a bit last night trying to go through some different physics principles that makes the club turn over, having some alpha torque and gamma torque put in there.

That was DeChambeau, speaking to reporters Sunday at LIV Golf Korea. The route through gamma torque arrived at a conclusion Johnson characterized as “something your mom tells you”: loosen your grip. Grip pressure is among the oldest coaching cues in the game, handed to beginners and tour players alike.

Johnson acknowledged something more nuanced alongside the critique. Some players will find genuine value learning from AI, he noted, and younger players coming into the game might engage with an AI coaching session the way others use a YouTube tutorial. DeChambeau, who processes swing problems in the language of physics and mechanics, may simply need to arrive at a grip cue through a gamma-torque discussion before it carries enough technical weight to register.

He’s not alone in building an AI relationship with his swing. Justin Rose has invested in Mustard Golf, an app developed with Golf Digest’s top-ranked U.S. instructor Mark Blackburn and performance coach Jason Goldsmith, which uses computer vision to analyze biomechanics from video. Rose’s approach starts from actual swing footage. DeChambeau’s Saturday-night session started from a description of how it felt.

For the companies building AI coaching tools, the common pitch is access: tour-level swing analysis for any golfer with a phone. The golfer who just paid eight figures for the access platform was using the free chatbot instead.

The Limit AI Coaching Can’t Clear

Knowing the Physics vs. Correcting the Feel

A review of biomechanics and motor learning research in golf, published in the journal Sports Biomechanics, found that golfers improved more when they focused on swing outcome or clubhead movement than when they focused on specific body mechanics. For skilled players, thinking about what the hands should do at impact tends to produce worse results than thinking about what the club should do through the hitting zone.

His Gemini conversation, stripped of the torque vocabulary, produced exactly that kind of cue. “Relax your grip pressure and let the thing fold naturally” is outcome-focused and external, exactly what motor learning research identifies as most effective for skilled players. The conversation didn’t prescribe a hand position. It told him to get out of the way.

Where the Data Ends and the Body Begins

He didn’t fully close out the problem on Sunday. Wedge misses bled back in under competitive pressure. “I’m still working it out,” he said. “I don’t have the answer.”

Text-based AI has a ceiling in motor skill coaching. It works from the golfer’s description of what their body is doing, which diverges from what the body is actually doing more often than most players recognize. A player who describes “hands lagging” might be producing the exact opposite; sensation and mechanical reality are unreliable translations of each other at tournament speed.

Professional golfers make this problem harder, not easier. Decades of practice build proprioceptive habits that are internally consistent and highly calibrated, but may be calibrated to a flawed movement pattern. His hands drifting forward through impact at Asiad Country Club wasn’t a deliberate choice; it was a pattern his nervous system had settled into, likely under accumulated fatigue across three competitive rounds. A physics explanation doesn’t reach that level. Repetition on the range, with or without a model watching, resets the pattern.

SAMI’s 3D approach starts upstream of the description problem, reading the actual body position from video and flagging what changed between a good swing and a poor one. Even so, 3D data stops short of the proprioceptive loop, the real-time feedback in the hands and forearms during a downswing, that governs whether the club releases or stays blocked. That loop is what competitive pressure compresses, and it’s the part no model reads yet.

At Pinehurst, Sportsbox helped him find a repeatable ball-flight pattern. In Busan, he was hunting a sensation. Those are different problems, and the gap between articulating the physics and delivering the feel is where elite motor performance lives.

DeChambeau’s 2026 Season and the U.S. Open

Korea came against the backdrop of a season in two distinct phases. Back-to-back LIV wins in Singapore and South Africa in March put him second in the season-long standings behind Jon Rahm. Then came two consecutive major missed cuts: the Masters in April and the PGA Championship in May, where he shot 76 in the opening round at Aronimink Golf Club.

The Korea result restored some momentum heading toward the next major on his calendar. A two-time U.S. Open champion, having won at Winged Foot in 2020 and at Pinehurst in 2024, he arrives at the 126th U.S. Open Championship at Shinnecock Hills Golf Club, running June 18-21 in Southampton, New York. Shinnecock is a par-70 links-style course that last hosted the Open in 2018, one of its five previous championships, and has a long record of amplifying ball-striking errors under wind and firm conditions. Misses that disappear at a modern tour venue leave marks there. “I feel like I’m on the right path now,” he said Sunday. Shinnecock, starting in 13 days, will be the first real measure of whether that’s true.