À retenir
A scoping review found only 11 studies (of 598 papers screened) that used large language models such as ChatGPT to give exercise advice, and they report tailored plans, time savings and better engagement. Only 5 of the 11 used experimental designs, so the evidence is early and mostly rests on expert opinion and user feedback.
Study design
Revue des preuves
Reading basis
Published abstract
Reading focus
Design & applicability
Cette explication éditoriale existe en anglais, chinois simplifié et chinois traditionnel. Cette interface affiche la lecture en anglais.
Question de recherche
Good exercise plans take time and expertise from medical and sports professionals. Chatbots built on large language models (LLMs) might make personalised plans cheaper and easier to get. The authors set out to map and classify how LLMs have been used for exercise recommendations and physical activity, and to find the gaps.
Population et méthode
Scoping review of English-language papers up to March 21, 2024, searched in Web of Science, PubMed, IEEE and arXiv with terms such as LLMs, chatbots, fitness plan and physical activity. Two independent reviewers screened papers against predefined criteria. Findings were grouped by theme and summarised in narrative form, with no pooled statistics.
Résultats
The authors’ reported results. Read each estimate with its population and uncertainty.
- 1Of 598 papers found, 11 (1.8%) met the inclusion criteria.
- 2ChatGPT-based models were used in 6 of 11 studies (55%).
- 3Usability was judged by expert evaluation and user feedback in 8 of 11 studies (73%); only 5 of 11 (45%) used experimental designs to test an LLM intervention.
- 4Across studies, LLMs could produce tailored exercise recommendations, save clinicians time, use patient-specific data to improve safety, raise engagement and support behaviour change, and widen access in remote or underserved communities.
Limites et incertitude
- 1The 11 studies are few and varied, and most judged usefulness by expert ratings and feedback rather than by measured health outcomes.
- 2The abstract does not give effect sizes, so how much a chatbot plan improves activity or health is unknown.
- 3The search ended in March 2024 and covered English-language papers only, in a field that moves quickly.
- 4This reading is based on the abstract only.
Mise en perspective
Chatbots may help draft or personalise exercise ideas, but the authors stress they supplement, not replace, human expertise, and expert validation is needed for safety. If you have a medical condition, check any AI-generated plan with your clinician or a qualified trainer before following it.
Vos données
If you ask an AI tool for a plan, compare it against what your own data shows: weekly active minutes, resting heart rate and how you recover (HRV, sleep). Judge it over several weeks on the same device rather than by single sessions.
Comprendre ce type de preuve
Check which designs were included, how searches and bias assessment were conducted, and the last search date. Publication date alone does not make the underlying evidence current.
Résumé structuré des auteurs English
Le résumé original en anglais est reproduit avec attribution sous la licence ouverte indiquée. Il s’agit d’une source, pas d’une interprétation MoveVitals. L’analyse préparée figure ci-dessus ; vérifiez les conclusions importantes dans la source.
Background
Regular exercise recommendations (ERs) and physical activity (PA) are crucial for the prevention and management of chronic diseases. However, creating effective exercise programs demand substantial time and specialized expertise from both medical and sports professionals. Large language models (LLMs), such as ChatGPT, offer a promising solution by helping create personalized ERs. While LLMs show potential, their use in exercise planning remains in its early stages and requires further exploration.
objectives
This study aims to systematically review and classify the applications of LLMs in ERs and PA. It also seeks to identify existing gaps and provide insights into future research directions for optimizing LLM integration in personalized health interventions.
Methods
A scoping review methodology was used to identify studies related to LLM applications in ERs and PA. Literature searches were conducted in Web of Science, PubMed, IEEE, and arXiv for English language papers published up to March 21, 2024. Keywords included LLMs, chatbots, ERs, PA, fitness plan, and related terms. Two independent reviewers (XL and CH) screened and selected studies based on predefined inclusion criteria. Thematic analysis was used to synthesize findings, which were presented narratively.
Results
An initial search identified 598 papers, of which 1.8% (11/598) of studies were included after screening and applying selection criteria. Of these, ChatGPT-based models were used in 55% (6/11) of the studies. In addition, 73% (8/11) of the studies used expert evaluations and user feedback to assess model usability, and 45% (5/11) of the studies used experimental designs to evaluate LLM interventions in ERs and PA. Key findings indicated that LLMs can generate tailored ERs, save time in clinical practice, and enhance safety by incorporating patient-specific data. They also increased engagement and supported behavior change. This made PA guidance more accessible, especially in remote or underserved communities.
Conclusions
This review highlights the promising applications of LLMs in ERs and PA but emphasizes that they remain a supplement to human expertise. Expert validation is essential to ensure safety and mitigate risks. Future research should prioritize pilot testing, clinician training programs, and large-scale clinical trials to enhance feasibility, transparency, and ethical integration.
Sources et attribution
Using Large Language Models to Enhance Exercise Recommendations and Physical Activity in Clinical and Healthy Populations: Scoping Review.
Lai X, Chen J, Lai Y, Huang S, Cai Y, Sun Z, Wang X, Pan K, Gao Q, Huang C.
JMIR medical informatics · 2025-05-27
DOI: 10.2196/59309 ↗PubMed · 40424584 ↗Europe PMC · Licence indiquée: cc by ↗Abstract © the original authors, reproduced under the recorded open licence. Translations support reading; use the English record to check terminology and numerical results. Our methodological interpretation is separate from the authors’ conclusions.
Interprétation éditoriale MoveVitals · contenu éducatif · aucune évaluation clinique individuelle enregistrée
Sources vérifiées 2026-11-01
Guides pratiques
À lire ensemble
7,000 steps a day linked to lower risk of death, heart disease and dementia than 2,000
Compared with 2000 steps a day, 7000 steps was associated with a hazard ratio of 0.53 (95% CI 0.46 to 0.60) for all-cause mortality across 14 studies, 0.75 (0.67 to 0.85) for cardiovascular disease and 0.62 (0.53 to 0.73) for dementia; cancer incidence was not significantly different (0.94, 0.87 to 1.01). Several outcomes rest on only two to four studies, and the certainty of evidence ranged from moderate to very low.
The Lancet Public Health
Brief bursts of vigorous effort in daily life linked to lower mortality in non-exercisers
Among 25,241 UK Biobank adults who reported no leisure-time exercise, a median of 3 vigorous bouts of 1 or 2 minutes a day, picked up by a wrist accelerometer, was associated with a 38% to 40% lower risk of all-cause and cancer death and a 48% to 49% lower risk of cardiovascular death than no such bouts, over an average of 6.9 years and 852 deaths. This is an observational study using one week of wrist data and an algorithm to define vigorous effort, so it cannot show that adding such bursts causes the lower risk.
Nature Medicine
Steps and mortality: gains flatten near 6,000–8,000 (age 60+) and 8,000–10,000 (under 60)
Pooling 15 cohorts (47,471 adults, 3,013 deaths, median follow-up 7.1 years), the second, third and fourth quartiles of daily steps (medians 5,801, 7,842 and 10,901) had hazard ratios for death of 0.60, 0.55 and 0.47 against the lowest quartile (median 3,553). The curve flattened at about 6,000 to 8,000 steps a day for adults aged 60 and over and 8,000 to 10,000 for younger adults. The data are observational, from a single week of step counting, mostly with waist-worn research devices.
The Lancet Public Health
Accelerometers on 36,383 adults: any activity, even light, linked to lower mortality
In eight pooled cohorts (36,383 adults, mean age 62.6 years, 2,149 deaths over a median 5.8 years), the most active quarter by accelerometer-measured total activity had a mortality hazard ratio of 0.27 (95% CI 0.23 to 0.32) against the least active, and the most sedentary quarter had 2.63 (1.94 to 3.56). Light activity alone showed the same pattern, with 0.38 (0.28 to 0.51) in the top quarter. These are observational associations in adults aged 40 and over from the US and western Europe, so illness-driven inactivity and confounding cannot be excluded.
BMJ

