Physiological
10 signals
- Sleep duration
- Sleep stages when supported by the selected device
- Sleep latency
- Wake-after-sleep-onset
- Heart rate
- HRV
- Respiration
- Movement
- Body or skin temperature
- Apnea-screening indicators
The AI Sleep Companion
This label appears wherever AI-generated analysis is shown.
Data inputs
Physiological
10 signals
Environmental
9 signals
Behavioral
9 signals
Clinical and emotional
5 signals
AI outputs
Confidence levels are shown, not hidden. Where the system cannot separate two explanations, it says so and routes the question to the reviewing clinician.
Consumer health AI
On July 23, 2026 OpenAI launched Health in ChatGPT for U.S. users. With permission, it connects Apple Health and supported medical records so lab results, visits, sleep and activity can be read in context rather than re-explained in every conversation.
OpenAI reports that health conversations are handled by GPT-5.5 Instant on free plans and GPT-5.6 Sol on paid plans, evaluated with hundreds of physicians on HealthBench Professional, and that connected records are not used to train foundation models or target ads.
Sleep Week is built to be compatible with that shift. The measurements taken during your four days, the wearable exports and the Sleep Blueprint are structured so they can be carried into a connected assistant you already use at home, extending the program past checkout.
OpenAI, Launching Health in ChatGPTMedical disclaimer and safeguards
Sleep Week is a proposed program. Descriptions of protocols, devices and partners are illustrative and are not a guarantee of clinical outcomes.
Dashboard demonstration
Night 1 is the untouched baseline. Night 2 tests the calibrated environment. Night 3 tests the complete personalized routine.
Night 1
Baseline
72°F · 1180 ppm CO₂ · 38 dB · 6 awakenings
Night 2
Calibrated environment
67°F · 780 ppm CO₂ · 30 dB · 4 awakenings
Night 3
Full protocol
66°F · 690 ppm CO₂ · 29 dB · 2 awakenings
Source: Illustrative Sleep Week sample dataset, 2026, not a medical record
Improvement, deterioration and uncertainty are all shown. A three-night comparison indicates response direction; it does not establish a diagnosis.
Overnight room temperature plotted against measured awakenings across the three nights.
Source: Illustrative Sleep Week sample dataset, 2026, not a medical record
A visible relationship in three nights is a hypothesis to test at home, not proof of cause.
Prototype output
The AI ranks candidate disruptors by the strength of the relationship it observed across your three nights, and states its confidence. Items needing clinical judgment are flagged for physician review rather than answered by the system.
Awakenings clustered in the first half of the night at 72°F and fell sharply at 66–67°F.
AI-assisted. Clinician reviewed. Screening indicators are never presented as a diagnosis.
Deliverable
Sleep window
Sleep opportunity 10:45 PM – 6:15 AM; fixed wake time 6:15 AM, including weekends
Sleep surface finding
Medium-firm hybrid with active cooling layer (Surface B), fewest awakenings and highest efficiency
Follow-up
90-day tracking with review at day 30 and day 90. Breathing indicators re-checked; formal evaluation if symptoms progress.
Ranked interventions
Physician reviewed before release. Subject to clinical review.
Hard limits
The system does
The system never
If you are experiencing a medical emergency, call 911 or your local emergency number. Do not use any AI interface for emergency symptoms.