AI Life / Health / AI Coaching

Early evidence says AI coaching can work for some goals, and the strongest study is in diabetes prevention. More on the AI coaching hub .

The evidence

  • JAMA randomized trial. 368 adults with prediabetes. An AI-led diabetes prevention program was noninferior to human-led coaching at 12 months (JAMA ). That’s a real result from a solid design.
  • Smaller studies. A 2021 npj Digital Medicine study with 42 people is an example of the earlier, smaller work. Small samples mean weak conclusions.
  • Systematic review. A review in PubMed Central pulls together the broader literature (PMC ).

What it doesn’t show

  • Most studies test structured programs, not chat features in a consumer app.
  • Results for one goal (preventing diabetes) don’t carry over to sleep or fitness automatically.

Using one well

  • Pick one goal.
  • Track the behavior, not just the score.
  • Talk to a doctor for anything medical. An app isn’t one.

For consumer coaches, see Whoop vs Oura . AI coaching also shows up in planning; see Motion vs Reclaim .