Strava removes millions of activities and cleans up its segments
Strava has performed a historical data cleanup using machine learning models to make its leaderboards fairer and has adapted its stance towards Garmin to preserve the synchronization of activities with external devices.
Strava intensifies segment cleaning with machine learning
Strava has implemented an advanced system of Artificial Intelligence to detect and remove misclassified activities that distort segment rankings on the platform.
This process includes, among others, cases such as e-bike activities registered as outings bike conventional, activity segments recorded on vehicles (for example, not turning off the GPS at the end) and mislabeled sports (for example, a bike ride turned into a race).
The latest figures and the true scope
In the most recent cleanup phase, Strava removed [the following] from the global leaderboard. 2,3 million activities of poorly rated e-bikes and 1,6 million of vehicle-related activities. As a result, around 293.000 athletes They regained legitimate positions in the top 10 of segments that previously occupied improper registers.
This adjustment is part of an ongoing effort by the platform to maintain the integrity of its leaderboards: throughout 2025 Strava removed millions of activities with automatically detected anomalies, which has reduced manual reports of irregularities such as e-bikes or cars altering the rankings by up to 85%.
How the models work: detecting “impossible patterns”
Strava has trained multiple machine learning models that analyze metrics such as speed, acceleration, and activity patterns to distinguish between human effort and activities that shouldn't influence leaderboards.
The goal is to ensure that times across segments are comparable and reflect actual performance, not artifacts or recording errors.
The battle with Garmin: from conflict to technical adaptation
Alongside the data cleanup, Strava was embroiled in a technical and legal dispute with Garmin. Own Matt SalazarStrava's Chief Product Officer, publicly explained the company's position in a message posted on Redditwhere he detailed the conflict surrounding data attribution and the use of the API.
In mid-2025, Garmin updated its API guidelines to require clear attribution when activity data originates from its devices, initially threatening to cut off API access if those standards were not adopted before the deadline. November 1th 2025.
After weeks of uncertainty and debate within the community, Strava adjusted its position and confirmed that will accept Garmin's terms and conditions To visibly display the origin attribution for activities without affecting the data flow between Garmin Connect and Strava. This has alleviated much of the concern that automatic workout synchronization might be disrupted.
The attribution requires that data from Garmin devices include a mention of the device model, which translates into a visible reference in the feed or activity details without necessarily requiring the logo to be displayed on all screens.
Why this matters to runners and cyclists
For most users, these changes mean two clear things: first, segment rankings are now more reliable and less prone to erroneous entries. Second, the connection between Garmin devices and Strava continues to function normally, and athletes can continue syncing their activities without interruption.
Tips to avoid problems with your activities
If you want to make sure your workouts are counted correctly on Strava, check the type of activity Before uploading, trim any automatic sections unrelated to exercise and correctly mark if you are using a e-bike.
This helps ensure your efforts are reflected in the appropriate leaderboards and prevents them from being excluded by automatic filters.
With these technical and data policy adjustments, Strava seeks to reinforce the credibility of its leaderboards and maintain interoperability with the devices that millions of athletes use every day.
