Marathon and AI: When Artificial Intelligence Runs with You

From app to AI coach: a logical evolution

The paper training log is now a thing of the past. As technology finds its way into every aspect of daily life, running is no exception. GPS watches, pacing strategy, real-time data tracking... These tools have become essential, for amateurs and experienced athletes alike. The rise of coaching apps is evidence of this digital transformation. In 2024, the RunMotion Coach Running Barometer reported that at least half a million regular runners in France use a coaching app. The same app has seen downloads grow sharply since its creation in 2028. Today, RunMotion Coach has a community of 100,000 active users, a figure that increases by 50% each year. These platforms appeal because of their accessibility and their ability to offer support similar to that of a human coach, often at a much lower cost.

Personalized plans, availability taken into account, complementary activities built in, adaptation to the terrain... The benefits are numerous. But these solutions also have their limits: a relationship that can feel impersonal, standardized feedback and a lack of real-time adjustment.

To address these shortcomings, coaching apps are now incorporating artificial intelligence. At the cutting edge, tools such as RunMotion Coach use machine learning to offer evolving plans tailored to each runner's profile. By analyzing thousands of sessions using individual data such as age, goal, training frequency and terrain, AI continuously refines its recommendations. Better still, every run is analyzed in depth, including time, distance and heart rate, to provide contextualized, personalized feedback on performance.

Adaptive training plans: an evolution still in progress

While artificial intelligence is fueling plenty of debate, it is also making structured training more accessible than ever. An amateur runner without the budget for a coach or paid app can now generate a free training plan with a tool such as ChatGPT. In reality, though, AI integration in running is still limited. “Runners are not using AI directly“, explains Joseph Mestrallet, a data scientist. “Most of them can neither code nor manipulate algorithms. In practice, they use platforms such as Strava, Garmin or RunMotion Coach, which incorporate AI to provide users with intelligent feedback.”

These services use data generated by the user such as sleep quality, heart rate, stress levels and perceived fatigue, to adjust the training plan in real time. This approach is already well established among amateur runners, but it reaches an entirely different level among elite athletes.“At the highest level, we go much further. We collect data over several years: weight, body fat, as well as complex indicators such as changes in heart rate in response to different stimuli. The algorithm is fed all this information to model a unique profile and optimize the athlete's preparation.” For Joseph Mestrallet, it is only a matter of time before these tools become standard, even among recreational runners: “If AI works for elite athletes, it can be adapted for amateurs. In my view, we'll be there in two years.“

Performance prediction: how far can the analysis go?

Running a marathon often means targeting a very specific finish time. Think of athletes chasing qualifying standards for a major competition. With the rise of artificial intelligence, that quest for precision has reached a new level. Some apps can estimate with considerable accuracy what time a runner is capable of recording over the classic distance. To do so, the algorithm draws on data such as VO2 max, threshold pace, heart rate, training history and fatigue levels.

Platforms such as Strava, Garmin and Coros have already built this logic into their features. Fitness estimates, finish-time predictions, effort and endurance scores... These tools draw on thousands of data points to refine their forecasts and steer runners toward a realistic or ambitious goal.

As algorithms become more sophisticated, their predictions are still not infallible. Weather, stress, sleep, digestion, pain... These are all variables that even the most advanced AI cannot fully anticipate. When asked whether an athlete should trust AI 100%, Joseph's answer is clear: “Data is useful, but the human element is what matters. AI can make mistakes too! RPE, or Rate of Perceived Exertion, remains fundamental, and AI will never be able to anticipate a stronger headwind than expected or a poorly timed fueling stop. AI cannot model the human factor.”

“Data is useful, but the human element is what matters.

The limits of artificial intelligence: what machines still cannot do

Despite its remarkable advances, artificial intelligence remains an imperfect tool, far from being able to replace humans in every aspect of athletic preparation. Several obstacles still stand in the way of widespread adoption. The first that comes to mind is the failure to account for emotional and mental context. AI cannot yet accurately assess certain factors such as mental fatigue, stress or fluctuating motivation. Yet these factors also play an essential role in a long-term preparation such as marathon training. Remember that AI relies on the data it receives. In other words, if your sensor is inaccurate or your heart-rate monitor is reading poorly, the algorithm may offer faulty advice.

Data can also be the starting point for certain excesses, according to Joseph Mestrallet: “A lot of amateur runners do reckless things because of data. Some of them place too much importance on it. It becomes a race to keep pushing further, with runners putting themselves at risk of injury... We must not forget that running should remain enjoyable, and that data, like artificial intelligence, should help us optimize our preparation, not put us at risk of injury.”

Finally, personalization is still imperfect. Despite increasingly advanced tools, AI cannot account for emotional, personal or psychological factors. Conversely, excessive personalization can also trap runners in a need for constant control, at odds with the pleasure and spontaneity of running.

Toward a marathon run entirely by data?

What role will AI play in a runner's preparation 10 years from now? Could we imagine a runner guided entirely by AI on race day? What if, tomorrow, a runner no longer decided anything? Not their pace, their nutrition or even when to drink or pick up the pace. On race day, an ultra-powerful AI could theoretically manage every aspect of the race: pacing strategy down to the meter, fueling calculated to the second, micro-adjustments based on the weather, real-time heart-rate data or changes in the pack. The runner would only have to follow instructions through an earpiece connected to their watch. A marathon on autopilot. But would the experience risk losing its meaning? Running is also about instinct, feel and enjoyment. Running with your head, but also with your legs. For now, AI cannot read a runner's mind. And perhaps that is where the beauty of sport still lies.

The debate is also opening up on another front: technological doping. Does using highly advanced AI, capable of optimizing training, nutrition and recovery better than a human coach, amount to an unfair advantage? As is often the case, the answer depends on how doping is defined. For the World Anti-Doping Agency, a doping substance or method is one that is:

• Significantly enhances performance

• Potentially dangerous to health

• Contrary to the spirit of sport.

AI is not toxic. But it provides a performance gain, sometimes a considerable one, and raises the question of equal access. Not every runner has the means or the knowledge to use these advanced tools. For some, that may be enough to call it “ethical doping”. For others, as long as it is not banned, it is simply another tool for progress.

Artificial intelligence has clearly changed the way we train, recover and sometimes even run. It has also made training more accessible and more individualized. But AI will surely never replace the feeling of heavy legs, the emotions of standing on a start line or the pride of crossing the finish line.