When people talk about AI in sport, they go straight to the elite end — the Premier League clubs with data teams the size of a small company, the marginal-gains stuff. Fine. That world doesn't need my attention; it has plenty.
The part that actually interests me sits much further down. My cousin is a club-level cyclist — nothing professional, weekend regional events, rides because he loves it. He picked up an AI coaching app last year and improved more in six months than in the previous two. When I asked what changed, he didn't mention the model or the data once. He said: "I stopped feeling like I was making it up alone." That's the line I keep coming back to. The tools that used to belong only to elite environments are now in the hands of people who'd otherwise have no coach at all.
What it's already doing at the top
Worth knowing what's standard at the professional level, because it trickles down fast:
- Video analysis. Every top club has systems that carve match footage up by player, event, tactical situation. A coach asks for "every time we lost possession in the attacking third" and gets a clip reel in seconds. That used to be three people, a remote, and a clipboard.
- Load and recovery. Wearables plus AI give continuous visibility into training load, recovery, injury risk. The good setups catch overtraining before it becomes an actual injury — and the best ones show the athlete the data and let them argue with it.
- Officiating. VAR, Hawk-Eye, automated strike zones. Plenty of critics, but the direction is set: subjective calls becoming objective where they can be. The sports keep their controversies. Just different ones.
- Talent ID. This is the one that cuts both ways. Models scanning youth footage can spot future stars earlier, which can help a kid from an under-resourced club get noticed — or can narrow what "looks like an athlete" and write off the late bloomers. Depends entirely on who's holding it.
What it's doing for the rest of us
This is the bit I get excited about. Form correction through a phone camera — your running gait, your tennis swing, your yoga — was gimmicky in its first generation and is genuinely useful now. Adaptive plans like my cousin's, which weigh his recent rides, his recovery, the weather, his target event, and rewrite next week accordingly. No substitute for a great human coach. An enormous upgrade on no coach.
Matchmaking in competitive apps pairs players at their own level, which isn't flashy but keeps people in the game instead of quitting after three thrashings. And injury prevention has reached the club level: a simple model can flag a young player whose load is spiking in a pattern linked to stress fractures. That insight used to require an elite medical staff. Now a volunteer coach at a Saturday club can see it.
The bits that should worry us
I'm not going to pretend this is all upside.
Surveillance creeps. The same monitoring built for professional squads is now sold to amateur leagues, corporate wellness schemes, and secondary schools — and the moment you're talking about children's bodies and health data, the stakes jump.
Talent gets commodified. Once every teenager can be scored, the pressure to optimise them starts young. Kids who'd have happily played three sports get funnelled into single-sport specialisation earlier and earlier, which the research is fairly clear is bad for both their development and their enjoyment.
And then there's the trap of measuring everything. The sun on your back, the teammate who makes you laugh, the daft satisfaction of a session in the rain on a Tuesday — none of that is a number. AI can nudge people into treating sport purely as performance, when for most of us it's also, maybe mostly, meant to be play.
What I'd like to see more of
Tools built for small-club budgets rather than Premier League ones, because most sport isn't played at that level. Privacy-first design as the default, especially for anything involving young people — local processing, minimal data, clear rights. Systems that treat the athlete as a partner who gets to interpret their own data, not a subject being measured. And some sensible limits: the recovery days and the off-pitch hours don't all need a sensor on them.
My cousin is still riding for the joy of it. The app didn't touch that. It just handed him a more thoughtful training partner for the days his human ones were busy — and made the sport he loves a little less lonely to be good at.
If you're thinking about AI in a sport or fitness setting, the accessibility angle is the one I'll happily talk your ear off about. Find me here.
