Let's start with what much of the industry would rather not say out loud: AI is going to replace the work of many coaches, and that's an easy prediction to make.
Writing a program, adjusting it after an injury and answering a nutrition or recovery question at 11 p.m. are tasks software already handles reasonably well, and in a few years it will handle them better.
Pretending otherwise protects no one. It's like assuming, when the iPad came out, that newsstands wouldn't feel the shift to reading the news on a screen.
For gym operators, the more useful question is this: who is the real threat, and which technology can carry the club into what comes next?
As has happened before, the threat and the opportunity are starting to look a lot alike.
Who wants to coach your members
The first contenders are some of the wearable makers. In May, Google turned the Fitbit app into Google Health, with a $9.99-a-month Gemini-based coach that suggests workouts and pulls together activity, sleep and nutrition data. Whoop has offered a coach built on OpenAI's models since 2023, and Oura has Oura Advisor.
Each of them has its own device, its own infrastructure and a subscription to sell, and that's no small detail: recurring revenue is the line item everyone is chasing right now.
They collect the data firsthand, from the member's wrist or finger, and use it to sell the member the same thing a club sells: someone who tells them how to train.
That makes them valuable sensor suppliers and direct competitors for the member relationship.
A club can happily use their devices, but every member who gets used to asking the maker's app for advice is a member whose trust now sits somewhere else. In September we looked at what that costs gyms in retention.
General-purpose assistants run on the same models but see the data last
Then there are the general-purpose assistants. In July, OpenAI rolled out ChatGPT's Apple Health integration to U.S. users, including those on the free plan, and in January Anthropic announced similar integrations for Claude.
Their models are no smarter than the ones the wearable makers already use, since Google's coach runs on Gemini and Whoop's is built on OpenAI. What they lack is a device of their own, so they read whatever reaches the phone's health data hub, whenever it gets there.
In the real world, it often doesn't, and the general-purpose assistant ends up giving recovery advice based on three-day-old data.
Third-party wearables write to Apple Health through their companion apps, and because of the limits operating systems put on apps running in the background, some only do so when the app is opened. If the member doesn't open it for a couple of days, the assistant is reasoning about sleep from two nights ago without knowing it.
Some of the most useful metrics never arrive at all: Whoop, for example, does not export heart rate variability to Apple Health.
Integrating directly with the manufacturer is a different story. Garmin, for instance, offers businesses a developer program with highly detailed health and activity APIs, so professional partners receive data from Garmin's platform as soon as the device syncs, without going through Apple Health.
On paper the general-purpose coach knows everything about the member. In practice it often works from an outdated picture full of gaps, and it answers with the same confidence an AI assistant sometimes shows when it states something false (what people in the field call a hallucination).
The AI coach that gym chains put in their own apps is caught in the middle
Between these two camps sit the vendors that sell clubs an AI coach aimed directly at members. In April, for example, New York Sports Club launched MYCO, a $15-a-month coach that members use through the vendor's own app.
Some of these products read wearables only through Apple Health, with the same gaps as the general-purpose assistants, and they rely on the same kind of models the big AI companies improve every few months. Meanwhile the wearable maker offers the member its own coach for a few dollars a month, with better data.
The result is a product with worse data than the device maker's and models no better than everyone else's.
For a club, that means paying for something members may soon get elsewhere for little or nothing, and after years of free digital services paid for with their own data, nobody is surprised anymore by a coach that costs nothing.
The trump card no tech giant holds
The club holds one thing none of these players has: data collected on the floor, at the moment it happens. Who came in and when, from the access control system; their heart rate during Tuesday's class; where they rank on the club leaderboard and in challenges with friends from class; which trainer works with them; and when they first skipped their usual time slot.
That data does not depend on an app being opened or an overnight sync. The club records it directly, and completely, while the member is training.
Garmin adds value on both fronts. During a class its wearables can broadcast heart rate in real time to the club's screens, and outside the club the direct integrations Garmin offers professional partners bring in complete, up-to-date data.
The technology a club can apply to that data is the same kind the giants use, but it works on better data.
The difference is that the club puts it in the hands of a person and builds a process around it, not a chat. As recovery data gets cheaper, the scarce skill moves to the gym floor.
Every morning, the software ranks members by churn risk, explains why each one made the list and drafts a first message based on that member's history. The trainer reviews it, edits it and sends it under their own name.
That is the route we chose for UPTIVO, with DASH ranking members by churn risk each morning and NATE, an AI assistant designed to give trainers superpowers, drafting programs and messages based on the club's own methodology, always subject to a trainer's approval.
Why the AI-augmented human trainer wins today
A line attributed to the economist Richard Baldwin has been making the rounds for a few years: AI won't take your job, but someone using AI will. On the gym floor, that plays out in very concrete ways.
AI on its own doesn't see a member's face at the front desk, doesn't know that the drop in visits coincides with a new job, and can't talk anyone into coming back with a well-timed phone call.
A trainer on their own sees all of that, but only for the few dozen people they can keep track of.
The members who quit are almost always the ones nobody is keeping an eye on. A trainer who uses AI sees the ranked list every morning and can stay on top of 100 or more members, and the time no longer spent writing programs goes back to the floor, correcting form and noticing who is missing.
Nobody can say for sure how long that advantage will last. In the short term, though, a trainer's real competitor is not the software but the colleague who uses it better. It is also why we think ABC's bet on AI that multiplies coaches is the right one.
What a chain should (and shouldn't) do
The first decision is to use wearables as sensors without giving up the member relationship.
Ignoring them would be an expensive mistake. ACSM ranked wearable technology the top fitness trend for 2026: nearly half of U.S. adults own a fitness wearable, and more than 70% of users have already used their wearable's data to decide how to train or recover.
The most common mistake is to add an AI coach to the app and hope it's enough to keep members loyal, or to hold on to them when an aggressive competitor opens right next door.
That risks breeding competition under your own roof: it teaches members, especially those used to budget-gym prices, that coaching can come from an app, and the tech giants sell apps with a built-in coach too, probably better ones than the club's.
There's a better route. Members can train with the Garmin or Apple Watch they already own, but advice, progress updates and phone calls should come from the club and its trainers, not from someone else's app.
When a vendor pitches an AI coach, ask where its data comes from and how fresh it is. One simple question tells you a lot: is the sleep data the AI coach used this morning from last night, or from two days ago?
Also ask how much effort it takes for the member. If getting an analysis means opening one app, then another, then tapping two buttons, almost nobody will keep doing it for more than a few weeks, because the technology that works in a gym is the kind members never notice.
If the answers aren't convincing, the budget is better spent elsewhere: on tools that help the trainers already on the club's payroll do their jobs better. Members will find an AI coach on their wrist or their phone anyway, while a trainer who remembers them and calls at the right moment is something they can only find at their club.
Sources
TechCrunch, Google's $9.99-per-month AI health coach launches May 19
MacRumors, ChatGPT's Apple Health Integration Now Rolling Out to U.S. Users