Here’s how Bryson DeChambeau, golf’s mad scientist, is using AI as a swing coach

Artificial intelligence is infiltrating every corner of professional sports, from scouting and injury prevention to scheduling. Now, it looks like golf has its most sophisticated AI adoption yet, and it’s happening in the bag of Bryson DeChambeau, the sport’s most notorious tinkerer.

“We’re building an AI golf coach,” DeChambeau says. “Essentially, it will be a golf coach that, based on data, will be able to tell you exactly what you’re doing, how to practice, and how to improve your game. We can take a golf swing, compile the information, upload it, and within a minute, it will give me what’s different from my gold standard set of swings.”

The setup is deceptively simple: a smartphone on a tripod gathering data via video, paired with Google’s Gemini AI to interpret said data. Combined, they create a swing coach so intuitive that DeChambeau uses it even moments before teeing off in a tournament.

“The mental game is something I’ve always struggled with,” he says. “But whenever I become a little more confident and comfortable with my feel, my mental game goes extremely positive. And this assistant has helped me become a lot more confident with my golf swing.”

AI + AI = Coach

DeChambeau’s coaching system starts with SportsBox, an AI-powered 3D biomechanical analysis app that analyzes over 30 key points on the body, club, and ball per golf swing. It measures everything from rotational range of motion to kinematic sequencing—the precise order in which different body parts accelerate and decelerate through the swing. This data is then processed by Gemini AI to turn those measurements into actionable coaching insights.

Think of SportsBox as the measuring tool, Gemini as the AI coach agent, and Google Cloud as the platform hosting it all.

The system starts by building and maintaining a database of DeChambeau’s optimal swings from recent years to create his “gold standard” set. So, when he hits a poor shot, the AI immediately measures that shot against his gold standard set and ranks the factors most likely contributing to the miss.

“We can take a golf swing, then upload it, and within a minute, it will give me what’s different from my gold standard set of swings,” he says. “It will give me a rundown list of the top [deviations] that are correlating to whatever’s causing me to miss.”

According to Granville Valentine, managing director of AI go-to-market at Google Cloud, it’s Gemini’s multimodal capabilities that bring the SportsBox data to life, creating the interactive coaching agent.

“Gemini is very differentiated on multimodality—the ability to ingest the combination of video, audio, text, and voice, and even livestreaming some of those capabilities into the model,” he says. “The combination of really deep video understanding plus core reasoning comes out in differentiated coaching guidance.”

The devil’s in the details

The granular nature of DeChambeau’s AI coaching reveals just how sophisticated modern sports analytics has become. The system uses Z-scores—statistical measurements showing how many standard deviations a movement is from the mean of a data set—to identify exactly where problems occur. Previously, DeChambeau would capture swing data but wait hours or days for analysis. With this technology, he gets feedback within a minute, allowing for real-time adjustments before a round.

“We were going through [the data] by hand in an Excel spreadsheet,” he says. “It was a manual process, very difficult. So you’re talking about months and months of trying to study the golf swing, now done in minutes.”

The data is also surprisingly precise. “Let’s say it’s a radial deviation at P6,” DeChambeau says. “That’s too much, meaning I’ve got too much wrist hinge, which makes the club come more outside in. So it’s very specific.”

For us non-DeChambeaus who got lost at “radial deviation” and checked out at “P6,” that’s where Gemini comes to the rescue. The AI’s ability to adapt its communication style allows users to train it to explain complex biomechanical concepts in terms appropriate for any skill level. Like other large language models, you can ask it questions, such as what specific terms mean, and as your understanding grows, it will adapt to give you more granular, technical data, meeting each golfer where he or she is at.

Old dog, new tricks

When he began using this technology earlier this year, DeChambeau found one of his fundamental beliefs about his swing challenged. For years, he says, he thought he needed to stay more centered over the ball—more on top of it—when hitting his driver. The AI consistently told him otherwise, saying he was too on top of the ball.

“It told me to keep swaying my chest just a bit back on the backstroke to get my center mass more behind the golf ball so I can allow the club to release through the impact more,” he says. “So that just blew my mind at how precise this assistant is. It was kind of a kick-in-the-butt moment of, wait, you gotta start trusting this thing.”

Eventually, he realized the AI’s objectivity as its strength.

“It’s unbiased,” he says. “It doesn’t tell you what it thinks you should do. It’s literally based on what you do when you’re doing your best, and keeps you in check with that.”

Democratizing elite-level instruction

The rapid evolution of AI coaching technology suggests we’re witnessing the early stages of a broader transformation in sports training. Valentine points to each new release of Gemini, which shows consistent step-function improvements in spatial awareness and reasoning capabilities.

“With each subsequent release, breakthroughs are happening,” he says, comparing Gemini’s current moment to the early days of Waymo self-driving cars, which needed time to become trustworthy enough for widespread adoption. “That level of trust—that level of breakthrough in the model itself—is now kicking over to a place where humans have the confidence to rely on this as a coach relative to a human coach.”

Still, Valentine says, the ultimate goal is not to replace human coaches, but to democratize access to elite-level instruction.

“I don’t think the objective is to get rid of coaches,” he says. “I think it’s to deliver access to those folks who don’t have access to coaches. There are lots of folks in the world who would probably be very well served to have access to coaching, it just hasn’t been available to them.”

At the PGA Tour level, DeChambeau believes there are further use cases for the tool, and that widespread adoption is inevitable once other players experience the results he’s seen.

“When these [other golfers] see what the capabilities are, they’ll immediately latch onto it,” he says. “Because it’s not about some theoretical idea. It’s about what works best for them as an individual. I can’t wait for a day when it’s a full-on coach, club fitter, you name it. We’re just at the beginning.”

View the full article