To analyze Apple Health data with AI, export it as CSV or JSON with HealthSave, then hand the file to an assistant. Upload it to ChatGPT or Claude for instant pattern-finding, or run a local model for full privacy. The export is the same either way; the difference is who sees your data, and this guide is upfront about that.
What AI is actually good at here
An assistant won't replace a doctor, but it's genuinely useful for the boring parts: summarizing months of data in plain language, spotting outliers, drafting a weekly recap, or eyeballing whether your resting heart rate tracks your sleep. You bring the data; the model helps you read it.
First, the privacy fork
This is the part most guides skip. Once your data is a file, you choose where it goes:
| Option | Where your data goes | Best for |
|---|---|---|
| ChatGPT / Claude / Gemini | That provider's servers, under their policy | Quick, powerful analysis if you're comfortable sharing |
| Local model (e.g. Ollama) | Stays on your own computer | Full privacy, no upload |
| Local model + Observatory API | Stays on your own hardware | Self-hosters who already run the Observatory |
HealthSave's own promise is that the app collects nothing and keeps your data on-device; the moment you upload a file to a cloud assistant, that's your decision, made knowingly. If that matters to you, use the local path.
Step 1, Export the data
In HealthSave, pick the metrics you care about and export CSV (most assistants read it best) or JSON (keeps structure). Both are free. Keep the file scoped, one or a few metrics over a sensible date range beats dumping everything.
Step 2, Hand it to your assistant
- Cloud (ChatGPT / Claude): start a chat, attach the CSV, and ask your question. Modern assistants parse tabular files directly.
- Local (private): run a model with Ollama or similar, then feed it the file's contents in your prompt or via a small script. Nothing leaves your machine.
Step 3, Ask good questions
Specific prompts beat vague ones. Some that work well:
"Summarize the trend in this resting heart rate CSV by month, and flag any unusual weeks."
"From this sleep data, what's my average sleep duration on weekdays vs weekends?"
"Here are two files, steps and resting HR. Is there any visible relationship?"
"Turn this HRV export into 3 plain-English takeaways I could mention to my doctor."
What to watch for
- Verify the numbers. Assistants can miscount or hallucinate; check any figure against the source CSV.
- It's not medical advice. Use it to understand your data, not to diagnose or treat anything.
- Mind what you share. Cloud uploads leave your device; if that's not OK, go local.
What's free vs Pro
| Capability | Tier |
|---|---|
| CSV / JSON export to feed any AI | Free |
| Full multi-year history | Pro |
| Self-host + REST API for a local model | Pro |
Pro is a one-time $24.99 (US price, varies by region), Family Sharing included, no subscription.
Honest limits
- HealthSave provides the data, not the model. You choose and run the assistant.
- Cloud assistants are not private, your file goes to the provider. Use local for privacy.
- AI can be wrong. Treat output as a draft to verify, not as fact or diagnosis.
- iOS only; reads Apple Health, so data must already be in HealthKit.
Related
- How to export Apple Health data, the formats and steps.
- Prefer spreadsheets? Analyze in Excel.
- Apple Health data in your homelab, for a fully local AI pipeline.
- How HealthSave handles your data.
FAQ
Can I upload Apple Health data to ChatGPT?
Yes. Export the metric as CSV or JSON with HealthSave and attach the file to ChatGPT (or Claude). The assistant can summarize trends, find anomalies, and answer questions. Be aware the file is sent to that provider.
Is it private to analyze health data with AI?
It depends on the tool. ChatGPT, Claude, and Gemini are cloud services, so your file goes to their servers under their data policies. For full privacy, run a local model (for example with Ollama) over the exported file, the data never leaves your computer.
What format works best for AI analysis?
CSV is the most universally understood by assistants and is compact. JSON is good when you want structure preserved. HealthSave exports both for free.
Will AI diagnose my health?
No, and you shouldn't ask it to. AI assistants can summarize and surface patterns, but they're not clinicians, they can be confidently wrong, and they don't replace medical advice. Verify any number against your source data.
How do I keep it fully local?
Export to CSV or JSON, then run a local LLM (such as one served by Ollama) against the file on your own machine. Self-hosters can also point a local model at the Observatory's REST API.
HealthSave is not a medical device. It is for informational purposes only and does not diagnose, treat, cure, or prevent any disease or condition. AI assistants are not clinicians and can produce inaccurate results; always verify against source data and consult a qualified professional. Cloud AI services receive any file you upload; HealthSave's zero-data-collection promise applies to the iOS app, not to third-party tools you choose to use.
Get HealthSave
Download free, export your data as CSV or JSON, and analyze it with the assistant, cloud or local, that fits your privacy comfort.
Download HealthSave on the App Store