Strava has completed one of the largest data cleanups in cycling app history.

Using three advanced machine learning models, the platform scanned every major segment leaderboard to remove fake times that frustrated competitive cyclists for years. The results: 2.3 million e-bike activities deleted, 1.6 million vehicle rides removed, and 293,000 athletes restored to the top 10.

James, a Strava engineer, explained on Reddit that the system now detects rides recorded on e-bikes but uploaded as regular cycling. The cleanup focused on the top 100 rides for every segment worldwide, targeting vehicles, misclassified sports, and e-bikes disguised as normal bike rides.

Tourist-heavy areas were hit hardest, where rental e-bikes often dominated leaderboards. One Reddit user from Sedona said: "Every segment here was taken by a tourist on an e-bike. I LOVE that Strava is fixing this."

The tech behind it is impressive. E-bikes have distinctive patterns: faster uphill, slower on flats, and unnaturally smooth speed graphs. Heart rate data also helps, as genuine KOM attempts almost always show high HR, unlike e-bike rides.

This isn’t against e-bike riders they have their own category. The goal was simply to protect traditional cycling leaderboards.

The community response has been overwhelmingly positive. As James noted, "Data integrity work is never really done," but cyclists can now finally trust what they see. For those chasing legitimate PRs, this cleanup validates years of hard work.