Fitness and health apps have been promising “smart coaches” and “personalised training plans” for years. But, to date, most programmes have been like online shopping recommendations, with exercises broadly matching your demographic profile and performance level.

Market analysts think the AI fitness market could be worth close to $35b USD by 2030. But how close are we to that future? Which company is training up the supertrainer? And how will that change the way we exercise, sweat and track our progress? And what do we need to know about this new world?

Confidence Udegbue has the perfect CV for designing an AI coach. The Vice President of Product at fitness app Freeletics studied electrical and computer engineering and teaches fitness classes in his free time. His broad shoulders, muscular biceps and infectious spirit are a dead giveaway: this guy knows what he’s talking about.

“In the gym, I can see immediately when someone I’m teaching is making a mistake,” says Udegbue. “But that expertise is hard to scale.” Freeletics is trying to solve that problem with AI. The app has been using a predictive algorithm since 2019 to suggest workouts based on demographic data and self-assessed fitness levels. This means that a 39-year-old man who has been training for two years and is at level 63 in the app won’t receive the same instructions as a 25-year-old beginner.

In 2024, Freeletics introduced the Coach+ feature – an AI-powered chatbot with Freeletics expertise and access to anonymised data from over 59m user journeys. Users can ask the virtual coach questions like, “How can I build muscle mass?” or “I feel weak – how can I motivate myself?”

Freeletics is currently testing a version that will allow the app to see you work out. As of April, users have been able to record themselves exercising on their smartphones. “AI counts the reps and gives direct feedback,” Udegbue says. That is particularly helpful because even experienced athletes do not always perform pistol squats or burpees correctly.

A personal coach was long the preserve of Hollywood actors, top models and CEOs – a highly competent service provider, always available whenever a slot opened up in their client’s busy schedule. They know their clients’ allergies, preferences and weak spots. They always know how to set the pace. Sometimes they’re pushy, sometimes they go easy. They are a mix of therapist, personal assistant and best friend – open 24/7, all major credit cards accepted.

In the soccer world, the manager is often called “boss” – a figure of respect who takes care of the players both on and off the field. A good coach can tell when something is off in a movement – when the person’s mind is elsewhere, or they’re lacking energy. Anyone who has had that person in their life knows that a good coach is worth their weight in gold, which is why there are coaches for everything – careers, relationships, nutrition – and why the idea of a personalised fitness coach is so appealing.

AI has no body or talent. It doesn’t know what it feels like for sweat to run down the skin or for muscles to cramp or for adrenaline to rush through the veins. But it does recognise patterns and make predictions that we humans can use increasingly often and, in the best-case scenario, find out more about ourselves in the process.

Mirrors show you how you see yourself. But the Magic AI Mirror promises that you will like what you see if you follow the exercises and tips on the reflective screen. Behind the glass surface is an AI coach who steers your workouts in real time.

Growl goes even deeper into movement detection. The start-up has developed an exercise boxing bag that captures every movement with 3D cameras and Lidar (light detection and ranging) technology. AI corrects your posture or encourages you when your energy decreases.

Freeletics is also banking on predictive AI. “Soon the system will recognise that user X has had an increased resting heart rate for days, so I won’t suggest high-intensity exercises,” says Udegbue.

The vision all companies are working on is a multimodal coach: AI that unlocks information – biometrics, genetics, video, training history – and conveys it intuitively to the user. But a perfect coach is more than just an algorithm. Researchers are working on reinforcement learning systems that set individual step goals that are challenging but achievable, and adapt whenever progress has been made.

“We will not be able to deliver on the promise of absolute personalisation for the mass market,” Eskofier says. But before you lose hope, you should know what he means by personalisation.

His laboratory supports, among other people, Sebastian Steudtner, the big wave surfer and world record holder. To do this, they measured his body in an MRI scanner, carried out psychological assessments, calculated strength curves and even fitted his surfboard and wetsuit with sensors.

Eskofier’s team created Steudtner’s digital twin. By the time the project concluded in May 2025, their AI system could already discuss with a real coach what angle Steudtner should surf a 100-foot wave at, and whether he’d be strong enough to do it.

“We can’t offer that service to millions of people,” Eskofier says. “But these systems can still create real added value.” He believes AI coaches are a good base: “AI can take over data processing and routine personalisation, while real coaches can focus on mentoring.”

AI coaches are getting smarter all the time, too, which is why it’s important to know what they can and can’t do. Limited data sets can lead to bias if too few women or people of below-average height are represented in the data.