Strava lets subscribers use Google Gemini to analyze training data

AI-powered training insights

Cyclists often accumulate years of training data on Strava, but finding specific trends in that information can be difficult. While the platform records metrics like speed, heart rate, and training volume, identifying meaningful patterns usually requires manual digging through many different activities. The new integration with Google Gemini aims to simplify this process by letting athletes ask questions about their history.

Instead of relying on one-time data exports, Gemini can access live Strava information whenever a user submits a query. This allows athletes to explore training trends and compare workouts across different periods without navigating through months of recorded logs. The goal is to provide a conversational way to interact with personal fitness data to see how an athlete is progressing over time.

For example, a cyclist might ask the assistant if they are riding faster at the same heart rate compared to an earlier season. The AI can then examine the available activity data to identify differences in training volume, consistency, and performance. It can also provide breakdowns of recent races or examine how specific environmental conditions might have influenced a particular workout.

How the MCP connection works

The integration functions through Strava's Model Context Protocol (MCP) connector, which the company first introduced with Anthropic's Claude in June 2026. This system allows compatible AI assistants to retrieve authorized information directly from the platform. It includes certain training analytics that are typically only available through Strava's paid subscription features.

To use the feature, subscribers can navigate to the Customize section in Gemini and select Strava under the Connectors menu. Users must then sign in to their Strava account and authorize the connection. This setup ensures that the AI has read-only permissions, meaning it can retrieve data to answer questions but cannot upload, modify, or delete any existing Strava activities.

The connection is also manageable through the Strava website and mobile app settings. On the website, users can find the MCP option in the settings menu, while the mobile app places it under Settings and AI Assistants. This allows athletes who are already using the Google AI ecosystem to add a new way to examine their training without switching between different services.

Limitations of AI analysis

While the AI can identify numerical patterns, it does not always understand the external factors that influence those numbers. For instance, riding faster at the same heart rate might suggest better aerobic efficiency, but it could also be caused by different terrain, wind, temperature, or bike setup. The AI identifies the correlation in the data but cannot always determine the specific cause behind the change.

Data quality also plays a significant role in how accurately the AI can provide insights. If a user recorded workouts without using a heart rate monitor or power meter, Gemini cannot reliably reconstruct those missing measurements. Inconsistent recording methods or inaccurate GPS data can also lead to misleading conclusions when the AI attempts to compare different training periods.

Strava notes that AI-generated answers may differ from the information shown in its own internal training tools, such as Performance Predictions. Because different systems process the same underlying data in different ways, users should not assume every detailed explanation is perfectly accurate. Furthermore, the tool is not a replacement for a qualified cycling coach or a medical assessment of injury risk and overtraining.

Availability and privacy

The integration began rolling out to subscribers in the United States on October 28, 2026. Strava has confirmed that they plan to expand the feature to additional countries, though they have not yet provided a specific international release schedule. For now, American cyclists can find the latest connection instructions in the official Strava guide.

Privacy is a central component of the integration, as training histories can contain sensitive information about an athlete's routines and frequently visited locations. The system relies on explicit authorization, and users can revoke the connection at any time through their Strava account settings. This allows users to maintain control over which AI services have access to their personal activity logs.

It is important to note that read-only access does not guarantee that retrieved information will never be processed by the AI provider. Athletes are encouraged to review Google's applicable privacy policies and data controls before authorizing the connection. By using the MCP connector, the system avoids the need for repeated manual exports, ensuring the assistant always works with the most recent workouts available.