Every fitness app is racing to add an AI coach feature. The pitch is obvious: personalised training, available 24/7, at a fraction of the cost of a real trainer.
But before building it, we wanted to know if people actually want it. We used our Digital Twin network to ask regular gym-goers aged 25-40 — people who work out at least three times a week and already use fitness apps — for their honest reaction.
The answer wasn’t no. But it wasn’t the enthusiastic yes that product teams are hoping for.
The gut reaction is curiosity, not excitement.
When we asked for their instinctive response to an AI personal trainer feature, the answers landed in the same lukewarm zone:
“My gut reaction is pretty curious, but also a bit hesitant.”
“It sounds interesting, but I’m a bit skeptical. I’d need to see how it actually works.”
“I’m not sure. I’d probably think it sounds cool, but I’d have to see it work before deciding.”
Nobody said “absolutely.” Nobody said “I’ve been waiting for this.” The dominant response was conditional interest — willing to try, but not willing to assume it’s good. That’s a much harder product to market than one people are already asking for.
The bar for “useful” is higher than most teams realise.
We asked what an AI coach would need to do to be genuinely useful. The answers revealed expectations that go far beyond workout programming:
“It would need to be super adaptable. Like, if I’m having a bad day or feeling tired, it should adjust the workout.”
“It needs to understand my specific goals and how my body feels, not just generic plans.”
“It should be able to analyse my form from videos, not just tell me what to do.”
People don’t want a smarter workout generator. They want something that understands them — their mood, their energy, their body on that specific day. That’s what a real trainer does. Anything less feels like a gimmick.
Safety is the number one concern. Not price. Not features.
When we asked about hesitations, one fear dominated everything else:
“My main concern is accuracy. Is the advice actually going to help me or potentially cause an injury?”
“I wouldn’t want to get injured because some AI told me to do something wrong.”
“I worry it might give advice that’s too generic and not understanding my specific body or limitations.”
This is a trust problem, not a feature problem. The AI coach could be technically impressive, but if users don’t trust it with their safety, they won’t follow its advice. And an AI coach that people don’t follow is just a feature nobody uses.
Willingness to pay is low. And conditional.
We asked how much they’d pay for an AI coach feature:
“Maybe $5 to $10 a month, but no more than that.”
“If it was really, really good, maybe $10-15. But it would have to prove its worth.”
“I’m not sure I’d pay extra unless it was significantly better than the free features.”
The ceiling is low — $10-15/month at best, and only after the feature has proven itself. This means the AI coach can’t be a premium upsell on day one. It needs a free tier that builds trust before it asks for money.
The real competition is the human trainer people can’t afford.
The most revealing answers came when we asked about real personal trainers:
“A real trainer offers that human connection, accountability, and can spot form issues instantly.”
“A trainer could see my form in person and push me in a way I don’t think an AI could replicate.”
“It’s the personal connection and immediate, nuanced feedback that would be different.”
The benchmark isn’t other apps. It’s the human trainer. And the gap between what people expect from an AI coach and what they get from a real one is enormous. Building an AI coach that doesn’t acknowledge this gap will disappoint users before it helps them.
Test the concept before you build the feature.
OriginalVoices lets product teams query real users about features before writing a line of code. Not surveys with leading questions — open, honest reactions from the people who would actually use it. The insight isn’t “should we build this.” It’s “what would this need to be for people to actually use it.”