I Tested the New Strava Integration With Gemini and I Have My Impressions.

This week, Strava announced a new integration with Google Gemini, giving subscribers another way to track their progress and get more out of their workouts. Building on the idea behind the recent Claude integration, this integration allows subscribers to directly connect Strava to Gemini to receive more contextual, personalized responses based on their own activity history.

Once connected, you can ask questions like:

• “Does my pulse speed up at the same heart rate?”

• “Could you give me a detailed analysis of my performance in the last race?”

• “How does strength training affect my running performance?”

Here’s some important context: Strava’s AI-powered “analytics” feature has been quite frustrating and unsatisfactory for me in the past . Athlete Intelligence (conveniently abbreviated as AI) is a Strava feature that records text responses to your workouts. Strava calls it “analytics,” but as far as I can tell, it mostly repeats information from the workout description, sometimes with errors added. I’ve found AI-powered fitness summaries to be largely useless, and Strava’s version ticks all the right boxes: it doesn’t use enough context to understand what’s going on, it presents general advice as if it were personalized, and it’s overly confident in incomplete data. So, being a skeptic, I decided to give Strava’s latest integration with Google Gemini a try. Here’s what it does, how to set it up, and whether it’s worth it.

How to connect Google Gemini and Strava

Gemini integration is now rolling out to subscribers in the US. To get started, open the Personal Features app in Gemini, select Strava from the list of connected apps, log in to your Strava account, and authorize Gemini to connect.

Connecting is optional and read-only. You decide whether to connect or not, and you can disconnect from connected apps in Gemini at any time. Important: the consent screen states that Gemini will have access to your training data, performance metrics, training plans, equipment, clubs, and profile information, and that some of your communications, including your name and email address, may be shared with Strava and used in accordance with its privacy policy.

How Google Gemini Interprets Strava Data

The first question I asked was one of Strava’s examples: “Am I speeding up at the same heart rate?” Gemini answered affirmatively; I see a significant performance improvement. Like many AI tools, it told me what I wanted to hear. I had to point out that heat and temperature are very important variables. When I steered the conversation in that direction, Gemini, of course, immediately agreed (“you’ve nailed it”) and explained that my long runs in mid-September were in 75-84°F (24-29°C) temperatures and high humidity, which inflates my heart rate, while the cooler October temperatures made the same efforts look better. This is an important point that the first answer completely missed.

I then asked the program to compare my past marathons, and I enjoyed it in many ways. For example, the additional questions it offered prompted me to consider questions I hadn’t included in my list, such as comparing my weekly running volume and the frequency of long runs before different races over the years.

Unfortunately, the results were initially unreliable. When I requested a detailed analysis of all marathons since 2019, Gemini reported that I only had one—the Marine Corps Marathon on October 25, 2026. In reality, I had logged at least eight. When I asked why the system couldn’t see them, it explained that access was limited to my recent history, roughly from the summer of 2026 onward, due to pagination limiting the number of past activities that can be returned at once. It also suggested a workaround: specifying specific dates and locations.

This workaround works, but it requires you to know what you’re looking for in advance. However, it turned out to be inconsistent. In other responses, Gemini referenced my 2025 New York City Marathon and 2024 Philadelphia Marathon, so the data coverage seems uneven. You need to be able to query data for older time periods as accurately and specifically as possible, and even then, double-check all the figures that are important to you.

Next, I asked, “What am I doing right and wrong as a runner?” Knowing how specific the questions needed to be, I deliberately chose to test a vague, simplistic question. The answer, as expected, was overly complimentary—though I didn’t argue with the recommendations themselves. I need more speed training! I know it! I didn’t need a chatbot to tell me this, though perhaps other Strava users will. Gemini’s suggestions (polarizing training, adding a threshold day) are reasonable, yes, but they’re also universal. Most recreational runners could look at my Strava and notice the same strengths and weaknesses. Similarly, Gemini then prepared for the crucial question by suggesting I ask if my easy runs are too fast. Answer: My easy runs were close to marathon pace (around 9:43/mile), and I should slow down by 45-90 seconds per mile, to around 10:30-11:15/mile. I completely agree. However, my main takeaway remains the same: don’t confuse AI-generated advice with some new, personalized knowledge. Often, it’s the same old wisdom you already know or could learn with a little research from a real, human source.

Result

If you ask me—a human!—this is a classic case of using AI for routine tasks, leaving the critical thinking to your own human brain. Gemini didn’t reveal anything truly insightful and missed important points, such as the effect of heat on heart rate, until I brought them up myself. But it processed the data faster than I could have done manually, and its subsequent suggestions prompted comparisons I might not have thought of.

For many users, the most sensible approach may be to forego AI summaries and focus on raw data. After all, your training load, sleep duration, and heart rate dynamics are already quite usable for practical purposes.

As with everything AI, it’s important to remember the gap between data and wisdom. Much wisdom is lost—even discarded—when you blindly trust an AI coach. Make sure you’re not ditching all the other skills you need as a runner and as a human being, such as listening to your body and perhaps running a little slower in the heat, even if that means your data is compromised.

If you use Google Gemini to analyze Strava data:

• When discussing events from previous years, please clarify dates and locations.

• Object. Before accepting flattering conclusions, ask about the weather, the terrain, and fatigue.

• Think of the results as a quick data collection, not as a personal trainer.

And if well-documented but still unclear privacy risks are worth taking advantage of, that’s your right.

More…