AI fitness trackers and recovery scores: what they can and cannot tell you

You wake up, reach for your wrist and find out how you feel. Recovery sixty two percent. Strain moderate. Sleep score seventy one. Before your feet hit the floor, an algorithm has told you what kind of day you are having.

Wearables have become genuinely sophisticated, and the AI layered on top of them is good at spotting patterns. The trouble is that a confident number on a clean dashboard feels like a fact, and much of what these devices report is an estimate built on top of another estimate.

1. What your wearable measures and what it guesses

It helps to separate measurement from inference.

Measured reasonably well. Heart rate at rest and during steady activity, using optical sensors that are accurate enough for everyday purposes. Movement and steps. Time in bed. Skin temperature trends.

Estimated, with real error. Heart rate during high intensity intervals, where wrist sensors struggle with rapid changes and wrist movement. Sleep stages, which are inferred from movement and heart rate rather than brain activity.

Largely modelled. Calorie burn. Validation studies of consumer wearables consistently find heart rate acceptable and energy expenditure poor, with errors that can be substantial. If you are eating according to what your watch says you burned, you are eating according to a guess.

Entirely proprietary. Recovery, readiness and strain scores. These are the headline numbers and they are company specific formulas with no external standard.

2. Recovery and readiness scores explained

Most recovery scores combine three ingredients: resting heart rate, heart rate variability during sleep, and sleep duration or quality, compared against your own recent baseline.

That is a reasonable idea. Those signals do respond to training load, illness, alcohol, heat and stress. What the score cannot do is tell you why it moved.

A low score after a hard session means something quite different from a low score after a late dinner, a warm bedroom, a glass of wine, a stressful evening or the beginning of a cold. The device sees the same dip and reports the same number.

So use the score as a prompt to check in rather than an instruction. If it is low, ask yourself what happened yesterday. Then warm up and see how you actually feel, because how a warm up goes remains the single best readiness test available.

3. Heart rate variability without the hype

Heart rate variability is the variation in time between heartbeats, and it reflects autonomic nervous system activity. It has become the flagship wellness metric, and it is frequently misused.

Three things worth knowing.

• Absolute values mean little across people. Normal HRV varies hugely with age, genetics and measurement method. Comparing yours to a friend's is a waste of energy.
• Day to day noise is large. Reviews of HRV guided training find that weekly averages are far more informative than single readings.
• It responds to everything. Training, alcohol, illness, poor sleep, heat, dehydration and psychological stress all move it. That breadth is what makes it interesting and what makes it hard to interpret.

Watch the trend across weeks. A gradual decline alongside heavier training, worse sleep and rising resting heart rate is a genuine signal worth acting on. A single low morning is not.

4. Sleep tracking and orthosomnia

Consumer trackers estimate total sleep time fairly well against clinical measurement. Sleep staging is considerably weaker, since deep and REM sleep are being inferred from movement and heart rate rather than observed directly.

Then there is the effect of tracking itself. Clinicians described a pattern called orthosomnia, where people become preoccupied with achieving perfect sleep data, grow anxious about sleep, and sleep worse as a result. The tracker did not cause bad sleep. Worrying about the tracker did.

A practical rule: if the score disagrees with how you feel, believe how you feel. And if you find yourself checking the sleep tab before you have decided how you slept, turn it off for a few weeks. Our guide to stopping overthinking at night is often more useful than any sleep metric.

5. When tracking helps

Trackers are not the problem. Used well they do several genuinely useful things.

• They increase movement. Reviews of activity trackers show modest but real increases in daily activity, which is more than most interventions achieve.
• They make trends visible. Slowly rising resting heart rate, or a training load that crept up without you noticing, are hard to see without data.
• They catch illness early. A resting heart rate jump often precedes symptoms by a day.
• They keep training honest. A log of what you actually did beats what you remember doing.
• They show cause and effect. Seeing what alcohol or a very late night does to your night is more persuasive than being told.

6. When tracking starts to harm

There is a point where the data stops serving you.

• You check the app before you have registered how you feel.
• A low score changes your mood for the day.
• You skip movement you would have enjoyed because the device said recovery was low.
• You train hard while feeling unwell because the score was green.
• You feel anxious when the watch is not charged.
• Rest days feel like failure because the rings are empty.

If several of those are true, the fix is not a better device. Try a two week break from scores, keeping only step count or nothing at all, and notice whether you train less or simply worry less.

7. The five numbers worth watching

Most dashboards show thirty metrics. Five carry nearly all the signal.

• Resting heart rate, as a weekly average. Cheap, stable and genuinely informative.
• Total sleep duration, as a weekly average. Ignore the stages.
• Weekly minutes of moderate activity. Compare against roughly one hundred and fifty minutes plus two strength sessions.
• Training load progression. Whether the work is increasing gradually or in jumps.
• How you feel, written down. One word per day. It correlates with everything and costs nothing.

8. Your health data and privacy

Wearable data is intimate: where you were, when you slept, when your heart rate rose. Before enabling everything, check what the app shares with partners, whether data is used for product development, and whether you can export or delete it.

Turn off what you do not need. A device does not have to know your location continuously to count your steps.

When to ask a person instead of the app

Speak to a clinician if you have chest symptoms, unexplained breathlessness, fainting, persistent fatigue, a resting heart rate that keeps climbing over weeks with no clear cause, or sleep problems lasting more than a month. No consumer score is designed to diagnose anything.

For everything else, a coach who can look at your data alongside your training and your life will get further in twenty minutes than a dashboard will in six months. You can describe what you are seeing anonymously through Ask for help in the BYOU app. Your name and contact details are never shared with the experts who reply, and nothing is booked unless you want it.