Can ChatGPT read your sleep data?

Yes, ChatGPT can read your sleep data, and so can Claude. In the US, ChatGPT can connect to Apple Health on any personal plan, free included, and Claude can on its Pro and Max plans, still in beta. Apple Health holds sleep from an Apple Watch, or from a ring or other tracker that syncs to it. A ring shares its sleep times with Apple Health, but its own scores usually stay in its app. Outside the US too, you can export a file from most sleep trackers and upload it to either one.

What a chatbot does well

A chatbot is good at explaining your numbers, like whether 20 minutes to fall asleep is normal or why your deep sleep number jumps around. It can work out your averages and spot an obvious change, such as later bedtimes since you started a new job. Before a doctor's appointment, it can help you get your questions down. Neither company pitches it as a stand-in for a doctor.

Where it trips up

Patterns that are only chance

A chatbot asked what goes with your bad nights can sound sure about a pattern that's only noise. With 60 nights and 20 things to compare, odds are something lines up by chance. It usually won't allow for how many things it compared, or say how many nights the pattern comes from, unless you ask it to.

Which night goes with which day

Most trackers file a night under the date you woke up, so the night after Tuesday's coffee shows up under Wednesday. Match them on the same date and you're comparing Tuesday's coffee with Monday night's sleep. A chatbot can make that mistake unless you tell it which date your tracker files each night under.

Agreeing with you

Chatbots tend to lean toward whatever answer your question hints at, which AI researchers call sycophancy. "Is magnesium helping my sleep?" invites a yes. Both companies test for it and say their newest models do it much less.

Follow-through is up to you

Changing a habit takes weeks. ChatGPT can flag a new trend in your Apple Health data, and both can set reminders, but you decide what they say and when, and you write the plan yourself.

Privacy

OpenAI and Anthropic both say the health data you connect isn't used to train their models, and you choose what to share. A file you upload to an ordinary chat falls under your usual training setting, so check that before you upload. Either way your data leaves your phone and sits on their servers, as it does with this app too, so read the settings before you connect anything.

Where Goodmorrow is different

Goodmorrow is for when you already know roughly what to do and it isn't happening. We use an AI model to write the replies to your morning notes. If you wear a ring or watch, it also writes our read of the last 60 nights. The read adjusts its main comparisons for luck and says how sure it is about each finding. You take on one habit for two weeks. We check in each morning, send a nudge when the habit needs one, and after the two weeks we tell you whether we think it helped. Like ChatGPT and Claude, the replies follow safety rules, tested before each change, so a note that needs a doctor should get pointed to one.