Skip to main content

AI & Health

What Happens to Your Health Data When You Paste It Into ChatGPT?

6 min read
Smartphone showing a chat interface next to a smartwatch on a desk, cool blue and purple lighting

What Happens to Your Health Data When You Paste It Into ChatGPT?

It depends on which product and which setting you used. A general ChatGPT conversation where you type in your sleep or HRV numbers can be stored and, unless you opted out, used to improve future models. A conversation inside the dedicated ChatGPT Health feature is treated differently by OpenAI's own policy and is excluded from model training by default.

That distinction gets lost constantly, because from the outside both look like the same chat window.

Is Pasting Health Data Into Regular ChatGPT Different From Using ChatGPT Health?

Yes, and the difference is the setting, not the content. OpenAI's general privacy policy says ChatGPT conversations may be used to train its models unless a user turns that off, through "Do not train on my content" in the privacy portal or the "Improve the model for everyone" toggle in Settings (OpenAI, How your data is used to improve model performance). That default applies to anything typed into a standard chat, including a wearable's sleep score or a symptom you're asking about.

ChatGPT Health, OpenAI's dedicated health feature launched to general availability on July 23, 2026, carries a separate notice. By default, medical records and health data connected through Health Features, along with conversations inside ChatGPT Health itself, are not used to improve OpenAI's foundation models or to target ads, regardless of a user's general model-training setting (OpenAI, Health Privacy Notice). So the same sleep data gets different treatment depending on whether it was typed into the general chat box or shared through the Health feature's connected-records flow.

The practical takeaway: if you're describing wearable numbers in a plain ChatGPT conversation rather than through the Health feature, assume standard training and retention rules apply unless you've explicitly opted out.

Does ChatGPT Delete My Health Data If I Delete the Chat?

Deleting a chat starts the clock, but it is not instant. OpenAI's retention documentation states that a deleted conversation is scheduled for permanent deletion from its systems within 30 days, unless it was already de-identified and disassociated from the account, or OpenAI must retain it longer for a legal or security obligation (OpenAI Help Center, Chat and file retention in ChatGPT). A deleted chat also cannot be restored once that process begins.

This is a general consumer-product retention window, not a health-specific one. If you want a shorter retention standard for sensitive conversations, that currently means opting out of model training and relying on the 30-day deletion schedule, since there is no separate "health mode" retention timer inside the general ChatGPT product.

What About Other General AI Chatbots?

The pattern holds across the category: training and retention are governed by product-wide settings, not by whether the content you typed happens to be health data. Anthropic's Claude, for example, now retains conversations for up to five years if a user allows that data to be used to improve the model, an extension from a 30-day default that took effect with a 2025 consumer terms update (Anthropic, Updates to Consumer Terms and Privacy Policy). Anthropic's privacy center confirms this applies specifically to chat and coding session content, which would include a block of pasted health metrics (Anthropic Privacy Center, Is my data used for model training?).

None of this means these companies are careless. It means a general-purpose chatbot is built around one consent toggle for all content types, and health information doesn't automatically get narrower handling just because it's more sensitive than, say, a request to draft an email.

Why Does This Matter More for Health Data Specifically?

Because health information carries regulatory weight that other chatbot inputs don't, and that weight depends on who is holding the data, not just what the data says. The FTC has said directly that AI companies offering model-as-a-service products can face liability under the FTC Act if they use customer data for purposes, including training, that contradict their stated privacy promises (FTC, AI Companies: Uphold Your Privacy and Confidentiality Commitments). The FTC has also stated plainly that if a company makes privacy promises about health-related information, expressly or by implication, the FTC Act requires it to live up to those claims, and that the obligation to maintain appropriate security exists even without an explicit promise (FTC, Health Privacy).

What this does not mean is that HIPAA covers your conversation with a chatbot. A general AI assistant is not a doctor, insurer, or clearinghouse, so typing your HRV trend into it does not create HIPAA-protected data. That gap, and which law actually applies instead, is covered in more depth in our explainer on whether HIPAA covers your fitness tracker.

What Should You Actually Check Before Pasting Health Numbers Into an AI Chat?

Three things, in order of how much they change your exposure:

Which product and mode you're in. A dedicated health feature with its own privacy notice, like ChatGPT Health, is not the same product as the general chat window, even inside the same app. Read the notice for the specific feature you're using, not the company's general privacy policy.

Whether model training is on by default, and how to turn it off. Most consumer AI chat products default to allowing training use unless you opt out. Check your account's data controls before a conversation starts, not after you've already typed something you'd rather not have retained.

Whether the conversation is tied to your identity. A chat tied to a logged-in account with your name and email is a different exposure than a one-off anonymous session. The fewer identifiers attached to a health conversation, the smaller the blast radius if that data is ever exposed or subpoenaed.

How Is This Different From Using a Purpose-Built Health App?

A purpose-built wearable health app is designed around one job: turning your metrics into insights, not improving a general-purpose model that also drafts emails and writes code for millions of unrelated users. That narrower scope doesn't automatically make every health app safer, the questions in our guide to connecting wearables to one health app still apply, but it does mean the data isn't competing for the same training pipeline as everything else a general chatbot is asked to do.

MotionSync's position here: the companies that make the AI models never receive your data, your health data is never used to train them, and the AI never receives your name, email, or other account identifiers alongside your health metrics. We never sell your data, never use it for advertising, and you can delete your account and everything in it at any time. We're also US-only for now, by design, while we build out the legal foundation to expand responsibly.

If you're deciding between a general AI chatbot and a purpose-built tool for your wearable data specifically, see our full comparison of MotionSync and ChatGPT Health for how the two actually differ in daily use, not just in data policy.

Frequently Asked Questions

Does OpenAI use my ChatGPT conversations to train its models?

By default, yes, for standard ChatGPT conversations, unless you opt out through "Do not train on my content" in the privacy portal or the model-training toggle in Settings (OpenAI Help Center). Conversations inside the dedicated ChatGPT Health feature are excluded from training by default regardless of that setting (OpenAI, Health Privacy Notice).

If I delete a ChatGPT conversation, is my health data gone immediately?

No. OpenAI schedules deleted chats for permanent removal within 30 days, unless the chat was already de-identified or OpenAI must retain it for a legal or security reason (OpenAI Help Center, Chat and file retention). The 30-day window is the standard for the general product, not a health-specific timer.

Is a conversation with an AI chatbot protected the way my doctor's records are?

No. HIPAA applies to health care providers, insurers, and their business associates, not to a general AI chat product. A chatbot conversation is instead covered, if at all, by general FTC consumer-protection authority and the FTC's Health Breach Notification Rule for connected health apps and devices. See our explainer on HIPAA and fitness trackers for the full breakdown of which law actually applies.


Related Articles:

  • privacy
  • chatgpt
  • ai-chatbots
  • data-rights
  • health-apps
Share

About the author

MotionSync logo

Written by MotionSync Team

Editorial Team

The MotionSync editorial team writes about wearable data, recovery, sleep, and heart health in plain English. Articles draw on peer-reviewed research and the metrics your wearable already tracks. Content is for general education and is not a substitute for medical advice.

Stay in the loop

Health insights, product updates, and wellness tips in your inbox. No spam, just useful content.

Unsubscribe any time. We respect your privacy.