ChatGPT feedback improves sports technique measurably faster than feedback from a training partner – that is the finding of a randomised study from March 2026. What sounds like a university context at first glance is full of transferable ideas for anyone who wants to try ChatGPT technique-feedback training for themselves.
ChatGPT as a Technique Coach: What a Study on AI Feedback in Sport Shows
Christopher Klenk3 min readAt a glance
A study (n=56, randomised) shows: structured ChatGPT feedback improves sports technique significantly more than feedback from a training partner. The principle – describe the error, the AI delivers an explanation and a corrective drill – works without a subscription and without a video upload. And with today's analysis apps, you can get even more out of it.
How the ChatGPT feedback actually worked
Researchers at Qassim University (Saudi Arabia) split 56 handball students into two groups. One group worked with classic peer-to-peer feedback among fellow students. The other supplemented this process with ChatGPT – free of charge, via smartphone, no subscription.
The mechanism was deliberately kept simple: an observer described the error they saw in 1–2 sentences and fed it into a predefined prompt. The rest was fixed:
Prompt
You are a physical education instructor.
Skill: [Übungsname]
Correct Technical Criteria: [Kriterium 1], [Kriterium 2], [Kriterium 3]
Observed error: [Fehlerbeschreibung in 1-2 Sätzen]
Required: Identify the error precisely. Explain its mechanical effect
on performance. Provide a corrective instruction. Suggest one simple
remedial drill. Respond concisely and professionally.No video upload, no camera-based AI. Just text – and a clear frame for what the response should contain.
What the numbers say
Across all seven measured basic skills, the AI-feedback group showed significantly better results (η² = .33–.87). These are medium to very large effects – and they remained fully stable after Bonferroni correction. In addition, motivation, perceived usefulness and the intention to keep using AI feedback all improved.
It is not the AI alone that makes the difference, but the combination: structured human observation plus AI-generated explanatory depth.
Feedback from a training partner often fails because the observer sees something but cannot explain why it is wrong and how it plays out biomechanically. ChatGPT fills exactly this gap.
From text prompt to video analysis
I have been looking for exactly this approach for strength and endurance training for some time – and the study shows that the principle holds. The study's limit: deliberately no multimedia, in order to test the method cleanly in isolation.
What already goes beyond that today: apps like Ochy (running analysis) or Gymscore (strength exercises) deliver an initial movement diagnosis via video. You can feed the identified error straight into ChatGPT using the prompt template from the study. The app tells you what is wrong, ChatGPT explains why and what helps against it.
One step further: Gemini and ChatGPT now support direct image and video uploads. If you film an exercise and upload the clip with a structured prompt, you get the same feedback loop – without a specialised app. Whether this works reliably in practice is another question. But the technical basis is there, and I will test it specifically.
The principle from the study – a clear prompt frame, human observation, AI explanatory depth – is already usable today. If you know the limits of AI feedback, you can use it more deliberately. And if you prefer to work with Gemini: the TheFitFuturist Gemini Gem follows the same principle – structured input, deeper output.
Rakha AH. Front Sports Act Living. 2026 Mar 25;8:1772502. → Full text (Open Access)
