Does Apple Watch detect migraines automatically?

No. Apple Watch does not have a built-in migraine detector. What it does capture — heart rate variability, sleep stages, resting heart rate, wrist temperature — feeds prediction models that estimate today's risk, but the user still logs the actual attack.

Is the Apple Watch better than the iPhone alone for migraine tracking?

It's an addition, not a requirement. Published studies (Stubberud 2023; Empatica 2025) show wearable physiology improves prediction by a few AUC points over diary-only models. iPhone-only Hermly users still get useful forecasts; Watch users get more precise ones.

Which Apple Watch model gives the best signal?

Series 8 and later add wrist-temperature anomaly detection, which is the single most useful Watch feature for migraine forecasting per the Kapustynska 2024 study. Earlier models contribute HRV, RHR, and sleep, which are all useful too.

Does the Watch run the prediction model?

No. The iPhone runs both the cohort and personalised prediction. The Watch shows the latest snapshot the iPhone broadcasts via WatchConnectivity, plus a one-tap attack tracker that syncs back.

Can Apple Watch detect a migraine?

Creator: Hermly
Published: 2026-05-16T00:00:00.000Z

Short answer: no, Apple Watch does not detect migraines on its own — Apple has shipped no such API and no peer-reviewed study has demonstrated it. Longer answer: the physiology the Watch does capture — heart-rate variability, sleep stages, resting heart rate, and (Series 8+) wrist-temperature anomalies — has been linked to migraine onset in multiple cohort studies. Apps like Hermly use those readings to estimate today’s risk, but the user still confirms the actual attack.

This page lays out what the research supports, what it doesn’t, and which Watch features earn their keep in a migraine forecast.

What “detection” would actually require

For Apple Watch to genuinely “detect” a migraine — flag one in real time as it’s happening — it would need both a reliable physiological fingerprint and a validation cohort comparable to the AFib detection studies. Neither exists yet. Migraine is a brain event with peripheral consequences (autonomic shifts, vasomotor changes) but no sensor-friendly signature like the irregular rhythm of atrial fibrillation. As of 2026, every published “Watch can predict migraine” finding is probabilistic forecasting, not detection.

What the research does support

The wearables literature breaks down into three groups of signals, each modestly predictive on its own and additively useful in combination:

Nocturnal HRV. Multiple studies (Kapustynska 2024; Empatica 2025) find that lower HRV the night before is statistically associated with next-day attacks in some users. The effect is modest — single-digit AUC improvement — but consistent.
Sleep quality, not just duration. Sleep efficiency (asleep time / in-bed time) and deep-sleep fraction track better with attacks than total hours.
Wrist temperature anomaly (Watch Series 8+). Kapustynska 2024 ranked this the single highest-information Watch feature. It works as a derivation from the user’s own 14-day baseline, not as an absolute value.

The best published prospective study (Empatica/Gottesman 2025) reports a personalised AUROC of 0.68 for next-day migraine using nocturnal wearable signals alone — meaningfully better than chance and meaningfully short of perfect. The same study also found that none of the chronic-migraine participants had above-random performance, only the episodic ones. That asymmetry is important: if you’re in chronic territory (more than 15 days/month), wearable signals alone are unlikely to forecast usefully.

Where Hermly’s iPhone-only path stands

If you don’t have an Apple Watch, the iPhone-only path uses:

Sleep duration from HealthKit (if you log it manually or via a paired sleep tracker).
Menstrual cycle position from Apple Health.
Barometric pressure from WeatherKit.
Your own attack and pain history stored locally.

Across cohort studies, these signals deliver roughly two-thirds of the predictive ceiling an Apple-Watch-paired user gets. Hermly is designed so the iPhone-only experience is a first-class one, not a degraded version of the Watch one.

What the Watch path adds, concretely

When the Watch is paired and HealthKit access is granted, Hermly reads (without storing on a server):

24-hour HRV (SDNN) average, plus a 30-day baseline ratio.
Resting heart rate plus baseline deviation.
Sleep stage breakdown — efficiency, deep-sleep fraction.
Wrist-temperature anomaly z-score (Watch Series 8+).

These five signals feed the same cohort model that processes weather and cycle data. They typically lift personalised AUC by 2–4 points after 30 days of personal labelled data. See the methodology page for the full feature list and model architecture.

The Watch is the surface, not the brain

A practical note: the Hermly Watch app does not run the prediction model. It displays the latest snapshot the iPhone broadcasts via WatchConnectivity. The reasons are deliberate:

Battery — running a Core ML inference loop on the Watch costs significantly more energy than on the iPhone.
Trust — one model, one prediction. Two devices running independent inferences would invite drift.
Privacy — the Watch never holds your raw HealthKit history; the iPhone does.

So when you raise your wrist and see today’s risk, that number was computed milliseconds ago on the iPhone in your pocket — and broadcast to your Watch the moment Today refreshed. The Watch is the surface. The iPhone is the brain.

What this is not

Not a recommendation to buy an Apple Watch for migraine forecasting specifically — the marginal AUC lift, while real, doesn’t justify the cost for most users on its own. If you already have one, Hermly uses it. If you don’t, the iPhone-only path is the one we built first.

Not medical advice. Hermly is a wellness app, not FDA-cleared for diagnosis or treatment. This page is informational and should not replace conversation with your doctor. If you're experiencing a sudden, severe, or unusual headache, contact a medical professional or emergency services.