Strava now works with two of the big three chatbots, and that tells you more than any single feature does. The company announced on October 28 that subscribers can connect their accounts to Google Gemini, following a recent launch with Anthropic’s Claude. The pitch is simple: ask plain-English questions about your own training history and get answers grounded in your real data.
The examples are the useful part. You can ask whether you’re getting faster at the same heart rate, request a mile-by-mile breakdown of your last race, or check how strength sessions are affecting your running. Strava also added newer queries with this release. You can compare your fastest 10K this year against all time, with the date and activity behind each result. You can ask how much the heat slowed down a long run, since temperature, humidity, wind, and conditions are now part of the data Gemini can see. And you can check your longest streak by sport, plus when the current one began.
None of this is magic. Strava already shows much of it if you know where to tap. But it buries trends under months of activity feeds and paywalled analysis screens, and a conversational layer is a better way to reach them. The weather piece is the most interesting addition. Separating “my legs were bad” from “it was 90 degrees” is exactly the sort of question runners argue about, and it needs context a plain pace chart doesn’t give you.
Setup is light. The connection is optional and read-only, and you can disconnect whenever you want. Inside Gemini, you go to Personal Intelligence, pick Strava under connected apps, sign in, and authorize access. It’s rolling out in the US starting now, and it works for Strava subscribers on all Gemini plans. So you need a paid Strava account, but not a paid Gemini tier.
Why does this matter? Look at the competition. Whoop, Oura, and Garmin all ship their own AI coaching features inside their apps, and Apple is expected to keep pushing Health-based guidance. Those are closed systems built around one company’s hardware. Strava is taking a different route. It has data from more than 200 million users across 185 countries and works with nearly every watch and bike computer, so it can act as the neutral data source that other assistants plug into. Let Claude and Gemini do the language work, and keep the activity history as the asset.
That’s a smart hedge. Building a coaching model good enough to compete with frontier labs is expensive and probably unwinnable for a company Strava’s size. Renting that intelligence while owning the dataset is cheaper, and it makes the subscription more valuable without much engineering overhead. It also reduces lock-in worries for users who already live in one assistant.
There are caveats. Chatbots can misread numbers, and a confident but wrong analysis of your heart rate trend is worse than no analysis. Strava hasn’t said how it handles that, and the announcement leans on “grounded” without describing safeguards. Availability is US-only for now, and it’s unclear when other markets get it. Also, if you already use Claude with Strava, Gemini adds reach but probably not new capability.
Still, this is worth paying attention to. Expect more connectors from Strava and its rivals, and expect OpenAI to be next. For subscribers, it’s a genuine upgrade to how you read your own training. For the industry, it’s another sign that specialized data, not the model itself, is where the leverage sits.
