AI has no favorite workout app

Run the same kind of questions at fitness apps and the leaderboard goes flat. The top app, Nike Training Club, is named in about a third of grounded answers. That is enough to lead but not to own the category. Compare language learning, where Duolingo is named in more than four in ten and leads its field clearly; fitness has no app that reaches even that. The rest of it splits across Strong, Hevy, Fitbod, JEFIT, Freeletics, Strava and a long tail, none of them close.

Some categories have a clear leader. Fitness is a scrum. How strong the "AI consensus" is depends on the category.

That is the useful part for anyone marketing an app. Where the assistants already converge on a leader, like Duolingo in language, breaking in is hard. Where they don't, like fitness, the shelf is still open. A new workout app has room to be named. It is not arguing with a settled answer.

The names work against them

Fitness has a visibility problem rooted in its names. Several of the apps are named after ordinary words. Strong, Sweat, Ladder, Caliber. An assistant, and my own parser, struggles to tell "Strong" the lifting tracker from "strong" the adjective. A name that is a dictionary word is a tax on being recognised as an entity, and a source of noise in the count even after the case-sensitive match: Strong's rank relative to the field has moved from month to month more than a same-named app should.

The other end of that scale: an app with no footprint at all. I track a brand-new indie workout app with no marketing behind it. It was named zero times, by any assistant, on any question. Not penalised, just absent. With nothing written about it across the web, there is nothing for the assistant to retrieve. Visibility starts with existing somewhere the model can read.

What this is

Thirty-five questions put to four AI assistants, plus Google's AI Overviews captured from the web. US English. A preview, like the others. See the fitness index for the full table.