Strava Integrates with Google Gemini for Personalized Training Insights

This week, Strava announced a new integration with Google Gemini, giving subscribers another way to understand their progress and get more out of their training. It builds on the company’s recent launch with Claude, and it lets subscribers connect Strava directly to Gemini to get more contextual, personalized answers grounded in their own activity history.

Once connected, you can ask questions like:

• “Am I getting faster at the same heart rate?”
• “Can you give me a detailed analysis of how I performed in my last race?”
• “How is my strength training impacting my running?”

Some important context here: Strava’s AI “insights” have sort of sucked, and annoyed me, in the past. Athlete Intelligence (conveniently abbreviated as AI) is the Strava feature that writes text responses to your workouts. Strava calls these “insights,” but as far as I can tell they mostly regurgitate the information from the workout description, sometimes with added errors. I’ve found AI fitness summaries largely useless, and Strava’s version checks every box: It doesn’t use enough context to understand what’s going on, it presents generic advice as if it were personal, and it’s overconfident in incomplete data. So, skeptical as I am, I gave Strava’s latest Google Gemini integration a whirl. Here’s what it does, how to set it up, and whether it’s worth it.

How to Connect Google Gemini and Strava

The Gemini integration is rolling out to U.S. subscribers beginning now. To get started, open Personal Intelligence within Gemini, select Strava under connected apps, sign in with your Strava account, and authorize Gemini to connect.

The connection is optional and read-only. You choose whether to connect, and you can disconnect at any time from Connected Apps in Gemini. Worth knowing: The consent screen says Gemini will access workout activities, performance metrics, training plans, gear, clubs, and profile info, and that parts of your conversation, plus your name and email, may be shared with Strava and used according to its Privacy Policy.

How Google Gemini Interprets Strava Data

The first question I asked was one of Strava’s examples, “Am I getting faster at the same heart rate?” Gemini said yes, I was seeing solid efficiency gains. Like many AI tools, it told me what I wanted to hear. I had to point out that heat and temperature are huge variables. When I steered the chat in that direction, Gemini, of course, agreed immediately (“you hit the nail on the head”) and explained that my mid-September long runs happened in 75–84°F weather with higher humidity, which inflates heart rate, while October’s cooler temperatures made the same effort look better. That’s an important caveat the first answer left out entirely.

I then asked it to compare my past marathons, and in many ways, I had fun here. For instance, the suggested follow-ups prompted me to think of questions I hadn’t put on my list, like comparing my weekly running volume and long-run frequency leading into different races in different years.

Unfortunately, the results weren’t reliable at first. When I asked for a detailed analysis of every marathon-distance effort since 2019, Gemini said I had exactly one, the Marine Corps Marathon on October 25, 2026. In reality, I’ve logged at least eight. When I asked why it couldn’t see them, it explained that its access is limited to my recent history, roughly summer 2026 onward, because of how pagination limits the number of past activities returned at once. It also offered a workaround: Give it specific dates and locations.

That workaround works, but it means you have to already know what you’re looking for. And still, it wasn’t consistent. In other answers, Gemini referenced my 2025 NYC Marathon and 2024 Philadelphia race, so the data’s reach seems uneven. You need to be able to query older date ranges as directly and specifically as possible, and even then, double-check any numbers that matter to you.

I then asked “What am I doing right and what am I doing wrong as a runner?” After learning that queries need to be highly specific, I intentionally wanted to test a vague, simplistic question. The answer was, true to form, overly flattering—though I didn’t disagree with the guidance itself. I need more speed workouts. I know this. I didn’t need a chatbot to tell me, though maybe some other Strava user does. Gemini’s suggestions (polarize training, add one dedicated threshold day) are sound, yes, but they’re also generic. Most recreational runners could glance at my Strava and spot the same strengths and weaknesses.

Similarly, Gemini then set itself up for a slam dunk by prompting me to ask whether my easy runs were too fast. The answer: My easy days were happening at nearly my marathon pace (about 9:43/mi), and I should slow down by 45–90 seconds per mile, to roughly 10:30–11:15/mi. I completely agree. That being said, my big takeaway here remains the same: Don’t confuse AI-generated advice with some newfound, personalized insight. It’s often the same conventional wisdom you already know, or could figure out with a little research from a real, human source.