Connecting your health data to an AI assistant means letting your assistant read the values your phone and watch record, such as steps, heart rate and sleep. That way, a question like “How did I sleep this week?” gets answered from your own data instead of from general knowledge. Before you connect, there are three things to know: where the data goes, who can read it, and how you remove the connection.
What's the point of connecting?
With your data in view, the assistant can answer concrete questions: it can sum up a period, compare two weeks or list the details of a workout.
Your phone and watch record a lot of measurements every day, but looking at them often means clicking from one chart to the next. When an assistant can reach this data, you ask your question in your own words: “How has my resting heart rate changed over the last month?”, “What was the total distance of my runs last week?”, “Which nights did I sleep less than seven hours?” The assistant reads the values it needs, does the math and tells you.
There's a limit to this: an assistant isn't a doctor, and your data isn't a medical test. To describe it honestly, the connection makes it easier to ask questions about your own records; it doesn't make decisions about your health.
How does health data reach an assistant?
Today there are three common ways: exporting your data as a file and uploading it, using the assistant's own integrations, or letting the assistant read the data when it needs it, through a connector.
- Uploading a file. Some health apps let you export your data as a file. You upload that file to the chat. It's simple, but the file is a copy of your data at that moment: new days aren't added, and it's often very large.
- The assistant's own integration. Some assistants offer direct connections to certain apps. What's shared and how it's stored differs from product to product; you'll need to look for that in the product's own documentation.
- A connector. When you ask a question, the assistant requests the data it needs from a server. More and more of these connections are built on an open standard called the Model Context Protocol (MCP). The data is always up to date, and you can remove the connection whenever you like.
Where does the data go?
Your data usually makes three stops: the health store on your phone, the server of the app in between, and the company that runs the assistant. Each stop is a separate party, with its own rules.
On iPhone, health data lives in Apple Health; on Android, in Health Connect. An app can read this data only with the permission you give it, type by type. With connector-based solutions, the data is usually uploaded to a server, because the assistant can't reach your phone directly: for you to ask questions even when your phone is off, the data has to sit somewhere the assistant can reach.
When the assistant asks for data to answer a question, the part it asks for goes to the company that runs the assistant. For example, Anthropic runs Claude and OpenAI runs ChatGPT. From then on, the data is handled under that company's terms and privacy policy. The app in between can't call back data that has reached the company.
What should you find out before you connect?
In short: find out what's read, where it's kept, who can see it and how to undo it. A good app spells these out in its privacy policy and on its own screens.
- Read-only? Does the app write anything to your health store? Can the assistant change your data?
- Which types? Can you choose what you share on the permission screen? You can turn off the types you don't need.
- When is consent asked for? If the data goes to a server, are you asked for separate, explicit consent first?
- Who can read it? Trust explanations that list the parties one by one, not absolute claims like “nobody can see it.”
- Ads and selling? Does it say in writing that your health data isn't used for advertising and isn't sold?
- What do analytics tools see? If the app uses usage measurement or crash reporting tools: do they run only with your permission, are they also used to measure ads (this can differ between iPhone and Android), can you turn them off, and can they access your health data?
- What are the assistant company's settings? Check your own assistant's settings for how chat history is stored and whether your chats are used to improve its models.
- How do you undo it? How many steps does it take to remove the connection, turn sharing off and delete your account, and when does deleted data leave the backups?
How far can you trust the assistant's answer?
An answer is only as reliable as the data and the assistant behind it. The values are measurements from consumer devices, and assistants can make mistakes when they calculate or interpret.
Smartwatches and phones measure heart rate, sleep and steps approximately; how the watch sits on your wrist, and how you move, both have an effect. Assistants also sometimes read the wrong period, try to fill in missing data, or state something they're unsure of in a confident tone. So ask the assistant to show the numbers and dates its answer is based on, and check any important value in the app's own report.
An assistant's answer is not medical advice. If you have a concern about your health, or before making any medical decision, talk to your doctor.
How do you ask a good question?
Name the period, the measurement and the format you want. A question like “List my sleep time for the last 7 nights with dates and give me the average” gets a clearer answer than “How's my sleep?”
- Give a date range: “this week,” “November 1–14.”
- Ask for a comparison: “compared with last week,” “compared with the month before.”
- Ask about the source: “Which device do these values come from?”
- Ask about patterns, not diagnoses: something like “Which days did I take fewer steps?”
How does Dave Connect do it?
Dave Connect reads only the types you allow from Apple Health or Health Connect, and never writes anything there. Your health data isn't sent to Dave until you agree on a separate screen; once you do, your data appears in your reports and, if you like, an assistant you connect can read it, read-only, through MCP. Your data can be read by you, Dave's systems and the assistant you connect. The app has no ads and we don't sell your data. If you allow it, we measure usage with Google Analytics (on Android, also how well our ads work); we track crashes with Firebase Crashlytics. Your health data is never sent to these tools. You manage sharing, assistant connections and your account in the app. The details are on the Dave Connect page and in the privacy policy.
Sources
- Model Context Protocol, What is the Model Context Protocol (MCP)?
- Apple, Consumer Health Personal Data Privacy Policy
- Google, Manage your Health Connect data
Claude and ChatGPT are trademarks of their owners. Dave is not affiliated with these companies and is not endorsed by them.
