The Health Data Sharing Audit Before AI

AI can help you organize wellness notes, sleep trends, wearable summaries, and questions for a clinician. But health data is sensitive. The first move is not a prompt. The first move is an access audit.

Before pasting any health-related summary into an AI tool, take ten minutes to see who and what already has access to your data. Then reduce the amount of personal information you share.

Step 1: Review iPhone Health sharing

On iPhone, open the Health app and tap Sharing. Review every person, provider, app, and device with access. Remove anything stale.

Look for:

• People who no longer need access.
• Apps you no longer use.
• Devices that are no longer current.
• Provider connections you do not recognize.
• Categories that are broader than necessary.

The goal is not paranoia. The goal is basic hygiene. Health data tends to accumulate permissions over time, and most people do not remember every connection they approved.

Step 2: Review ChatGPT data settings

Open ChatGPT settings and review Data Controls. If you are handling sensitive wellness information, consider using Temporary Chat and turn off model improvement if that is your preference.

Temporary Chat is useful because it keeps the conversation out of regular chat history and avoids memory creation. That does not mean you should paste raw medical records. It means you should share less, not more.

Step 3: Convert raw data into a safe summary

Share summaries instead of raw exports whenever possible.

Good:

• "Average sleep was 6 hours 20 minutes for four weeks."
• "Travel weeks were worse."
• "Resting heart rate increased after three late dinners."
• "Energy felt lower on days with late caffeine."

Avoid:

• Names.
• Exact addresses.
• Provider portal screenshots.
• Insurance or account numbers.
• Raw medical records.
• Full device exports.
• Exact dates when relative timing is enough.

Build a privacy-safe summary

The safest useful summary is boring. It should describe patterns without exposing identity. Write it like a note you would be comfortable handing to a clinician, not like a data export.

Use this structure:

• Time period in broad terms, such as "the last four weeks."
• Metrics in averages or ranges, not raw logs.
• Context that may matter, such as travel, sleep schedule, work stress, meals, caffeine, or exercise.
• What changed from your usual baseline.
• Questions you want help organizing.

Example:

Over the last four weeks, my average sleep was about 6 hours 20 minutes. The worst nights were travel nights and late-work nights. Resting heart rate was higher during the same stretch. I am not asking for diagnosis or treatment. I want help organizing patterns and questions to discuss with a qualified professional.

That is enough for AI to help with organization without needing names, exact dates, records, screenshots, or account details.

Where AI can help and where it should stop

AI can help you summarize patterns, prepare questions, notice obvious correlations, and make a visit with a professional more organized. It should not diagnose, prescribe, tell you to stop medication, reinterpret lab results as a medical conclusion, or replace professional judgment.

If the answer starts sounding like a diagnosis, redirect it. If the answer suggests treatment, redirect it. If the answer gives you a scary conclusion from thin data, do not keep prompting until it sounds certain. The right output is a clearer list of questions, not a false sense of medical certainty.

Use this prompt

I am pasting a privacy-safe wellness summary. Do not diagnose. Do not suggest treatment. Help me organize the data into patterns, questions, and one low-risk habit to discuss with a qualified professional. If the information is not enough, ask clarifying questions instead of guessing.

This prompt gives the AI a safe job: organize, clarify, and prepare questions. It explicitly blocks diagnosis and treatment suggestions.

What the output should be

A useful answer should give you:

• Patterns to discuss.
• Questions for a qualified professional.
• Possible low-risk habits to consider.
• Missing information that would make the summary more useful.
• A reminder not to treat the response as medical advice.

If the model gives a diagnosis, medication advice, or a treatment plan, stop and redirect it.

Use this:

Do not diagnose or recommend treatment. Rewrite this as a neutral summary of patterns and questions for a qualified healthcare professional.

The visit-prep output

The best final artifact is a short visit-prep note:

• Three patterns you noticed.
• Three questions to ask.
• One habit you want to discuss.
• What information you may need to bring.
• What you should not assume from the AI response.

That keeps the tool in its lane. AI helps you prepare. The professional helps you interpret.

Action steps

• Audit Health app sharing before you use AI with wellness information.
• Use Temporary Chat and review ChatGPT Data Controls for sensitive summaries.
• Paste only a privacy-safe summary and require the AI to organize questions instead of giving medical advice.

This information is for educational purposes only and is not medical advice. Please consult a qualified healthcare professional before making changes to your health routine.

Sources: https://support.apple.com/guide/iphone/share-your-health-data-iph5ede58c3d/ios and https://help.openai.com/en/articles/7730893-data-controls-faq