Analyze Your Apple Health Sleep Data with ChatGPT
Apple Health quietly records every night of sleep your Apple Watch sees - stages, duration, bedtime drift, awakenings. ChatGPT can read all of it in one go and tell you what your eyes would miss scrolling through the Health app. Here is how to do it, plus ten prompts that work out of the box.
By Martin
Co-founder, vitalina
How it works
- 1
Export your sleep data
Open vitalina, select Sleep Analysis, pick a 30-day range or longer, and export as CSV. CSV is what ChatGPT reads most reliably.
- 2
Upload the file to ChatGPT
Open a new chat, click the attachment icon, and select your sleep CSV. Free and Plus tiers both support file uploads.
- 3
Pick a prompt and ask follow-ups
Start with one of the prompts below. The real value comes from the second and third question, so keep going.
Get the big picture
Start here. These prompts give you a clean read of your last month before you dig into anything specific.
30-day sleep summary
“Here is my Apple Health sleep data for the last 30 days. Tell me my average time asleep, average time in bed, my sleep efficiency (asleep / in bed), and how consistent my bedtime and wake time were. Flag any nights that look unusual.”
Why this works: Gives you the four numbers that matter most before you go deeper.
Sleep stage breakdown
“Look at my REM, Deep, and Core sleep across the last month. Calculate the percentage each stage takes up of my total sleep, and tell me whether those percentages are in a healthy range for an adult.”
Why this works: Total sleep is a blunt number. The stage mix is where quality lives.
Sleep debt check
“Assuming my sleep target is 8 hours, calculate my cumulative sleep debt over the last 30 days. On which day was my sleep debt highest? Is the trend getting better or worse?”
Why this works: Surfaces a problem that hides if you only look at single nights.
Find patterns you would miss
ChatGPT is good at the kind of comparison you would not do by hand - weekday vs weekend, outliers, drift over weeks.
Weekday vs weekend
“Compare my sleep on weekdays versus weekends. Look at bedtime, wake time, total sleep, and sleep efficiency. How big is the gap, and is it big enough to qualify as social jet lag?”
Why this works: Most people have a weekday/weekend split they have never measured.
Find your worst nights
“Identify the 5 worst nights of sleep in this dataset (lowest efficiency, shortest duration, or most awakenings). For each one, tell me what stood out - was it a late bedtime, fragmented sleep, very little Deep sleep?”
Why this works: Helps you remember what was different on those days (alcohol, travel, stress).
Bedtime drift
“Plot my bedtime over the last 60 days as a trend. Has my bedtime been getting earlier, later, or staying stable? Show me the standard deviation week by week so I can see if my schedule is becoming more or less consistent.”
Why this works: Bedtime consistency is one of the strongest predictors of sleep quality.
Sleep, recovery, and training
For the best results, upload your heart rate and workout CSVs alongside sleep. ChatGPT can correlate across all three.
Sleep vs workout days
“I am uploading my sleep data and my workout data for the same period. On days I worked out, did I fall asleep faster, sleep longer, or get more Deep sleep? Compare workout days to rest days.”
Why this works: The single most useful correlation for athletes - works only if you upload both files.
Sleep and resting heart rate
“Here is my sleep data and my resting heart rate data. After nights of poor sleep (under 6 hours or low efficiency), how does my resting heart rate behave the next day? Is there a measurable lag between bad sleep and elevated RHR?”
Why this works: Quantifies what athletes call 'feeling off' the day after a bad night.
Recovery score from scratch
“Based on my sleep duration, sleep efficiency, time in Deep sleep, and resting heart rate, build me a simple daily recovery score (0-100) for the last 30 days. Explain the formula. Flag the days I should have taken it easy.”
Why this works: Builds you a poor-man's Whoop or Oura readiness score from data you already have.
Turn analysis into action
Once you have a read on what your data says, ask for output you can use - changes to try, a doctor-ready summary, a comparison against last quarter.
Three things to change
“Based on everything you have seen in my sleep data, give me 3 specific changes I could make this week to improve my sleep. Base each suggestion on something concrete in the data, not generic sleep hygiene tips.”
Why this works: Forces ChatGPT to ground advice in your numbers, not the internet.
Make it doctor-ready
“Summarize my sleep data into a one-page report I could bring to a sleep specialist. Include average duration, stage breakdown, consistency, and any patterns that look unusual. Keep medical jargon minimal.”
Why this works: Turns raw data into a hand-off your GP or sleep clinic can actually read.
Compare to last quarter
“I am exporting two ranges of sleep data: the last 30 days and 90-60 days ago. Has my sleep improved, gotten worse, or stayed the same? Compare averages and consistency, and tell me what changed most.”
Why this works: Progress checks need a baseline. Two exports give you one.
Tips for sleep-specific prompts
- •Always ask for the formula - if ChatGPT calculates a sleep score or efficiency, ask it to show how. It is easy to spot mistakes that way.
- •Mind the time zones - if you have travelled, tell ChatGPT, otherwise bedtime drift will look much worse than it is.
- •Ask for a chart - “Show me a chart of my sleep duration with a 7-day moving average” works and reveals trends faster than tables.
- •Combine with heart rate or HRV - sleep alone tells you the story; sleep plus recovery metrics tells you why.
- •Avoid medical claims - ChatGPT is good at description, not diagnosis. Treat outputs as a starting point for a real specialist.
A note on privacy
Sleep data is sensitive - it reveals your schedule, your stress, sometimes your relationships. When you upload it to ChatGPT:
- Go to ChatGPT Settings > Data controls > turn off “Improve the model for everyone”
- Or use a Temporary Chat, which is not used for training
vitalina itself never sends your health data anywhere. The export stays on your device until you choose to share it.
FAQ
Can ChatGPT analyze my Apple Health sleep data?
Yes. Export your sleep data as CSV from vitalina and upload the file to ChatGPT. It can summarize your average sleep duration, calculate bedtime and wake-time consistency, break down REM, Deep, and Core sleep, and spot nights that fell outside your usual pattern.
How much sleep data should I export?
At least 30 days. Sleep varies a lot night to night, so shorter ranges produce noisy averages. 60 to 90 days is ideal if you want to see weekday vs weekend patterns or compare months.
Which sleep metrics does Apple Health track?
Sleep stages (REM, Deep, Core, Awake), time in bed, time asleep, bedtime, wake time, and sleep interruptions. vitalina exports all of these as columns in your CSV so ChatGPT can reason about each one.
Can ChatGPT correlate sleep with workouts or heart rate?
Yes. Export sleep, heart rate, and workouts in the same session and upload all three CSVs. Ask it to find correlations - whether you sleep better on days you exercise, or whether your resting heart rate drops after weeks of consistent sleep.
Does this work with Claude or Gemini?
Yes. The same prompts work with any AI assistant that accepts file uploads. The export from vitalina is identical regardless of which model you use.
Related guides
- → What to ask ChatGPT about your Apple Health data - the full prompt library, all metrics
- → Export Apple Health sleep data - the export step in detail
- → Export Apple Health to CSV - why CSV beats JSON for ChatGPT
Ready to analyze your sleep?
Export 30 days of Apple Health sleep data with vitalina, then use any prompt above.