Can ChatGPT Write Your Workout Plan? An Honest 2026 Test

Yes, ChatGPT can write your workout plan — and it is better at it than most people expect. The large language model behind ChatGPT is the same class of engine behind the LLM coaching apps you would compare it against, so this was never a story about a toy versus “real” software.

Here is the catch. When researchers graded an AI chatbot’s exercise recommendations across 26 clinical populations in 2023, the advice was 90.7% accurate but only 41.2% comprehensive.¹ Accurate means it rarely tells you something flatly wrong. Comprehensive means it covers everything that actually matters for you.

Two of the standard objections to ChatGPT as a fitness tool have since expired. It now remembers you between chats, and in the United States it can read your Apple Health data.²³ Most articles on this question — including an earlier version of this one — still say it can do neither. This is the corrected version.

Quick answers

Question	Short answer
Can ChatGPT write a workout plan?	Yes. Exercise scientists rated an AI chatbot’s exercise recommendations 90.7% accurate across 26 clinical populations in a 2023 evaluation¹
Does ChatGPT give good workout advice?	Good as first drafts, thin as programs — 90.7% accurate but only 41.2% comprehensive in the same study¹
Are ChatGPT workout plans good?	Good enough to start from, not to follow. Prescription detail was the weak point — training volume fully specified in 0% of responses¹
Does ChatGPT remember my training between chats?	Yes, since April 2025 — saved memories plus reference to your chat history³. It is a summary, not a set-by-set log
Can ChatGPT see my Apple Watch or Oura data?	Yes, in the US, since 23 July 2026 — Health in ChatGPT connects Apple Health on web and iOS, all plans, ages 18+²⁴
Does it track progressive overload week to week?	No. Nothing in ChatGPT prescribes next week’s load from last week’s performed sets
Does it make up studies?	Yes — 47% fabricated references in a 2023 audit of ChatGPT medical content⁵, falling to 28.6% for GPT-4 in a 2024 analysis⁶
What does it cost?	Free tier; Go $8, Plus $19.99, Pro $100 or $200 per month as listed on the US App Store on 17 September 2026⁷
Is it a substitute for a trainer?	No. OpenAI states Health in ChatGPT is “not intended for use in the diagnosis or treatment of any health condition”²
Best prompt approach?	Front-load equipment, schedule, goal, experience and injuries in one message — accuracy was high, comprehensiveness depended on inputs¹
When do you need more?	When you want next week’s sets decided by last week’s performance rather than by a fresh conversation

The short answer: yes, but with a measured comprehensiveness gap

ChatGPT writes a competent workout plan in about thirty seconds, and the movements it picks are usually sound. In a 2023 evaluation, exercise scientists rated an AI chatbot’s recommendations 90.7% accurate but only 41.2% comprehensive across 26 clinical populations.¹ That gap — not the accuracy — is the whole story.

Think of a chef who cooks a technically perfect dish but never asks whether you are allergic to shellfish. The food is “correct.” It is just not built for the person eating it.

What the study actually counted matters. The shortfall was concentrated in exercise prescription detail — frequency and intensity were fully specified in 8% of responses, time in 4%, and training volume in 0%.¹ That is a programming gap, not a data-access gap.

Two caveats before you lean on those numbers. They come from the February 2023 version of ChatGPT, tested in March 2023 — two model generations back — and no equivalent re-test of current models has been published. Treat them as the best available measurement, not a current benchmark.

Amanda Zaleski and colleagues are explicit that AI-generated exercise advice should be used alongside clinical expertise and oversight.¹ Their senior author, Linda Pescatello, is Board of Trustees Distinguished Professor of kinesiology at the University of Connecticut and senior editor of the ninth edition of ACSM’s Guidelines for Exercise Testing and Prescription — this is a careful read of the technology, not a hostile one.

Closing that gap means handing the model your injuries, your recovery and your last six weeks of performed sets. Some of that is now possible inside ChatGPT. The rest is what a purpose-built coach like SensAI exists to do.

What changed in 2025 and 2026

Two of the four classic “ChatGPT can’t do this” objections are out of date. ChatGPT gained cross-chat memory in April 2025 and US Apple Health access on 23 July 2026.²³ Two objections survive: it runs no progression loop, and it still fabricates sources at measurable rates.⁵⁶

Objection, as usually written	Status, 17 September 2026	What actually happened
“It forgets you between chats”	Out of date	Memory shipped February 2024; referencing your broader chat history followed in April 2025, initially for Plus and Pro and at first excluding the EEA, UK, Switzerland, Norway, Iceland and Liechtenstein, since extended to those regions³⁸
“It can’t see your wearable or sleep data”	Out of date in the US	Health in ChatGPT, first tested through a dedicated hub in January 2026, became available to all US users 18+ on web and iOS across Free, Go, Plus and Pro on 23 July 2026, connecting Apple Health, medical records, Function and MyFitnessPal²⁴
“It doesn’t track progressive overload”	Still true	Reading a workout is not prescribing the next one. No ChatGPT feature decides next week’s load from last week’s performed sets
“It invents studies”	Still true, but improving	47% of references in a 2023 audit of ChatGPT medical content were fabricated⁵; GPT-4 hallucinated 28.6% in a 2024 analysis⁶

Note what these changes do not address. The 41.2% comprehensiveness shortfall was a missing-prescription-detail problem — sets, intensity, volume — and neither memory nor a health connector writes a prescription.¹ They fix different limitations than the ones that study measured, and nobody has re-run it.

Two limits on the health connector matter for training specifically. It is United States only, and connecting Apple Health is done in the ChatGPT iOS app.⁴⁹ And OpenAI states plainly that the service is “not intended for use in the diagnosis or treatment of any health condition.”² It is a surface for reading and discussing your health data, not a training-prescription engine.

If you want the detail on what your ring or watch actually writes into Apple Health in the first place, our guide to what syncs from Oura, WHOOP and Garmin into Apple Health covers the gaps.

What ChatGPT is genuinely good at

ChatGPT is excellent at four fitness jobs: generating a structured first-draft plan in seconds, explaining exercises in plain English, answering discrete one-off questions, and providing judgment-free accountability at 11pm. Researchers studying it for resistance-training prescription found it applied progressive overload and periodization correctly when asked to design a program.¹⁰

1. Fast first-draft plans. Ask for “a 4-day upper/lower split with dumbbells only,” and you get a structured, sensible program in seconds. Washif and colleagues found ChatGPT generated contextually appropriate resistance-training programs; GPT-4.0 went further, spontaneously adding active recovery, nutrition and hydration guidance nobody asked for.¹⁰

The same paper is candid about the limits: GPT-3.5 prescribed uniform 90–120 second rest intervals regardless of training phase — a departure from the specificity principle — and omitted methods such as low-load and blood-flow-restriction training.¹⁰

2. Plain-English explanations. “What is the difference between a Romanian deadlift and a conventional one?” You get a clear answer without the gatekeeping of a forum thread.

3. Motivation and accountability nudges. It never judges, never tires of your questions, and can reframe a missed week without the guilt spiral.

4. One-off questions. “Is it normal for my legs to shake on the last squat rep?” For discrete, self-contained questions, a single good answer is all you need.

Our breakdown of how AI workout generators actually build a plan goes deeper on the mechanics. The short version: generation is the easy part, and ChatGPT does it well.

Where a bare chatbot still breaks down

Two real gaps remain. ChatGPT has no progression loop — nothing in it reads last week’s performed sets and prescribes this week’s load — and it fabricates citations at rates measured from 47% in 2023 down to 28.6% for GPT-4 in 2024.⁵⁶ A third, softer gap: its memory is a curated summary, not a training log.

It has no progression loop. Adding load or adding reps both drive muscle growth, and either route works — but something has to record what you did and decide what comes next.¹¹ ChatGPT can read a workout you completed. It has no mechanism that turns that reading into next week’s prescription.

Worth knowing: how progression should be quantified is genuinely contested. Hammert and colleagues argue that tracking increases in total training volume “tells very little about muscle growth adaptations,” since the connection between accumulating volume and adding muscle is non-causal.¹² The loop matters more than the arithmetic you run inside it.

Its memory is a summary, not a log. The memory system stores curated facts and impressions from your chats.³ That is genuinely useful — it will remember the shoulder you tweaked in March. It is not a set-by-set record of 5×5 at 82.5kg on 3 September with the last rep grinding.

A scoping review of LLMs for exercise concluded they work best as a supplementary tool rather than a replacement for human expertise, because their outputs fall short of professional-level expertise and need expert validation.¹³ Our own reading adds a second reason: a chat window keeps no validated record of what you actually trained.

SensAI holds that record directly — performed sets, loads, reps and constraints, carried forward into the next week’s plan rather than re-summarised.

It still fabricates citations. In one audit of 115 references across ChatGPT-generated medical content, 47% were entirely fabricated and another 46% were real citations with wrong details — only 7% were both authentic and accurate.⁵ A separate comparative analysis put GPT-4’s reference-hallucination rate at 28.6%.⁶ The trend is downward across model generations. You still cannot tell which citation is the wrong one.

And recovery-aware adjustment is not automatic. ChatGPT can now read your sleep and activity if you connect Apple Health in the US.² It will not, on its own, lower today’s prescription because your HRV cratered last night. A meta-analysis of heart-rate-variability-guided training found it improved vagal-mediated HRV more than predefined training, with small and non-significant differences in aerobic fitness and endurance performance.¹⁴ Reading the data and acting on it are different capabilities.

The hallucination problem nobody screenshots

People screenshot ChatGPT’s slick workout plans constantly. Nobody screenshots the part where it invents the study backing them up — audits put that at 47% fabricated references in 2023-era medical content⁵ and 28.6% for GPT-4 in systematic-review references.⁶ The chatbot delivers a fabricated source in exactly the same confident tone as a real one.

A fabricated citation is mostly harmless when you are reading for curiosity. It is a real problem when it is wrapped around a claim like “studies show you should train this injured joint through pain.”

Use it as a first-draft writer and explainer. Do not hand it the role of final medical or programming authority — especially with an injury or chronic condition, where expert review of AI-generated exercise prescriptions is treated as essential for catching safety gaps.¹⁵

ChatGPT vs a coach that knows you

Side by side, the bare chatbot and a coaching app share the same LLM brain. The difference in September 2026 is narrower than it was in 2024 — memory and Apple Health access have closed two rows — and it now concentrates almost entirely on whether anything turns your performed training into the next prescription.

Capability	ChatGPT (bare, September 2026)	SensAI
Generate a first-draft plan	Yes, in seconds	Yes
Plain-English explanations	Yes	Yes
Remembers injuries and preferences across sessions	Yes — memory plus chat-history reference since April 2025³	Yes — held as structured user context
Reads wearable sleep, activity and HRV	Yes, US only, connected in the iOS app, since 23 July 2026²⁹	Yes — HealthKit: sleep and activity from Apple Watch, Garmin, Oura, WHOOP; HRV from Apple Watch
Set-by-set training log	No — memory stores summaries, not performed sets	Yes — guided tracking, load and reps per set
Decides next week’s load from last week’s sets	No	Yes — weekly regeneration from performed training
Adjusts today’s session to today’s recovery	No — reads the data, does not re-prescribe	Yes
Risk of fabricated citations	Measured at 28.6%–47% depending on model generation⁵⁶	Reduced by grounding prescriptions in your own logged data
Intended for health conditions	No — “not intended for use in the diagnosis or treatment of any health condition”²	No — fitness coaching, not medical care

The top four rows are now a tie, and that is a real change from the 2024 version of this comparison. The bottom rows are where the difference lives, and they all describe the same thing: a closed loop between what you did and what you are told to do next.

For a fuller treatment of where automated and human coaching each win, see our guide on AI vs human personal trainers.

Which LLM is best for fitness advice?

For the plan itself, no frontier model is meaningfully better. Generation is the easy part — ChatGPT applies progressive overload and periodization correctly when asked¹⁰ — and in our testing Claude and Gemini produce comparable drafts. The separator is what each can see: only ChatGPT publishes an Apple Health connection.

Assistant	Cross-chat memory	Reads Apple Health	Set-by-set training log
ChatGPT	Yes — saved memories plus chat-history reference, April 2025³	Yes — US only, connected in the iOS app, since 23 July 2026²⁹	No
Claude	Yes — Free, Pro and Max, on web, desktop and mobile; health is a sensitive topic excluded from memory by default and must be switched on¹⁶	Not published	No
Gemini	Yes — “Past Chats” personal context, rolled out to free users worldwide except Europe on 26 February 2026¹⁷	Not published	No
SensAI	Yes — structured user context, not a chat summary	Yes — HealthKit: Apple Watch, Garmin, Oura, WHOOP	Yes

Claude’s default exclusion of health topics is the detail most people miss. If you want it to remember an injury across conversations, you have to turn sensitive-topic memory on deliberately.¹⁶

What each option costs, as of 17 September 2026

ChatGPT’s paid tiers run from $8 to $200 per month on the US App Store; dedicated coaching apps run $6.99 to $15.99 per month.⁷¹⁸¹⁹ A ChatGPT Plus subscription costs roughly three times a SensAI subscription, and the two are not substitutes — one is a general assistant, the other a training system.

Product	What it is	US price, 17 September 2026	Free access
ChatGPT Free	General assistant; includes Health in ChatGPT in the US²	$0⁷	Yes, permanent
ChatGPT Go	Higher limits on the same assistant	$8/mo⁷	—
ChatGPT Plus	The tier most people mean by “ChatGPT”	$19.99/mo⁷	—
ChatGPT Pro	5× or 20× Plus usage limits	$100/mo or $200/mo⁷	—
SensAI	LLM coach with HealthKit recovery data and week-over-week programming	$6.99/mo or $69.99/yr¹⁸	7-day full trial, no card
Fitbod	Proprietary strength algorithm, generated gym sessions	$12.99–$15.99/mo, $79.99–$95.99/yr¹⁹	7-day trial, card required
Freeletics	Algorithmic coach for bodyweight, HIIT and running	Training Coach tiers listed at $34.99–$79.99²⁰	Demo-grade only

These are the in-app purchase tiers listed on each product’s US App Store page on 17 September 2026. Listings frequently show several tiers at once while pricing tests run, and prices change without notice.

When to use ChatGPT — and when you need more

Use ChatGPT when the task needs general knowledge; use a training system when it needs your performance history. First drafts, exercise explanations and one-off questions are general-knowledge tasks. Progression, recovery-aware adjustment and injury management are history tasks, and history is the thing a chat window still does not keep.

Use ChatGPT when:

You want a quick first-draft plan and you will adjust it yourself.
You are learning what an exercise is, why it exists, or how to do it in principle.
You have a one-off question with a self-contained answer.
You want a no-judgment sounding board for motivation or reframing a bad week.
You are an experienced trainee who already tracks your own progression and reads your own recovery.
You want to discuss your Apple Health data — in the US, via the iOS app — rather than have something act on it.²⁹

You need a coach with your training history when:

You want next week’s sets chosen by last week’s performance, not re-derived in a fresh conversation.
You wear a tracker and want today’s workout to actually change because of last night’s sleep and HRV.
You cannot reliably tell whether your program is working — a question worth its own honest look in how to know if your workouts are actually working.
You are outside the United States, or on Android, where the Apple Health connector is not available.²⁹
You are managing an injury or condition where missing context is a safety risk, not an inconvenience.

If your situation lives entirely in the first list, ChatGPT alone is genuinely fine. If you keep landing in the second, you do not need a different AI — you need the same LLM intelligence connected to your history, your wearable, and your actual results. That is what SensAI is.

How to get the most out of ChatGPT for fitness

Front-load every constraint in one prompt, verify every citation it offers, paste your training log at the start of each session, and make recovery calls yourself. Accuracy was already high at 90.7% in the 2023 evaluation; comprehensiveness was the part that depended almost entirely on what you feed it.¹

1. Front-load every constraint. Not “a workout plan.” Instead:

“Write a 3-day full-body plan. 45 minutes per session. Dumbbells and a pull-up bar only. Goal is muscle gain. Cranky left shoulder — avoid overhead pressing. Intermediate, two years of consistent training. Give sets, reps, and a progression rule for each lift.”

That last clause matters: under a single open-ended prompt, the 2023 study found training volume fully specified in 0% of responses.¹

2. Ask it to cite — then verify every source. When it references a study, ask for title, authors and journal, then look it up. Given fabrication rates measured between 28.6% and 47%,⁵⁶ treat every citation as guilty until confirmed on PubMed or the publisher’s site.

3. Paste your log at the start of each session. Memory stores a summary of you, not your sets.³ Keep a running log — last week’s exercises, weights, reps, how each felt — and paste it in. This is the manual version of the progression loop, and it is exactly the bookkeeping a purpose-built coach automates.

4. Make the recovery call yourself. Even with Apple Health connected, ChatGPT will not lower today’s prescription because you slept five hours.² That judgment stays with you.

5. Re-check anything that touches an injury. Expert review of AI-generated exercise prescriptions for at-risk populations is treated as essential.¹⁵ That has not changed.

The bottom line

ChatGPT in September 2026 is a genuinely better fitness tool than it was in 2024. It remembers you.³ In the United States, through the iOS app, it can read your Apple Health data.²⁹ Anyone still telling you it does neither is working from a 2024 map.

What has not changed is the loop. Nothing in ChatGPT reads the sets you performed last week and decides what you lift this week — and that loop, not raw intelligence, is what turns a plan into training. Brad Schoenfeld, professor of exercise science and graduate program director in human performance and fitness at CUNY Lehman College, was the senior author on the trial showing that adding reps and adding load both build muscle.¹¹ Either route works. Neither happens by accident.

So use ChatGPT for first drafts, explanations and one-off questions — it is excellent at all three, and free at the entry tier. When you want the same LLM intelligence holding your actual training log, reading your HealthKit recovery data, and deciding next week from last week, that is a different tool. It is what SensAI was built to be. When you are ready to compare the field, our roundup of the best AI personal trainer apps in 2026 is the place to start.

Frequently asked questions

Is ChatGPT good enough to replace a personal trainer?

No. It generates accurate first-draft plans — 90.7% accurate in the 2023 JMIR evaluation¹ — and explains exercises well. It cannot watch your form, and OpenAI states its health features are “not intended for use in the diagnosis or treatment of any health condition.”² It does not replace the parts of coaching that depend on knowing you over time.

Will ChatGPT remember my workout history between chats?

Yes, since April 2025. ChatGPT combines saved memories with reference to your chat history, so it can carry your injuries and preferences across sessions.³ But it stores a curated summary, not a set-by-set training log — it will remember that your shoulder is cranky, not that you pressed 22.5kg for 8 last Tuesday.

Can ChatGPT use my Apple Watch or Oura data?

Yes, with limits. Health in ChatGPT reached all US users 18 and over on 23 July 2026, across Free, Go, Plus and Pro, connecting Apple Health, medical records, Function and MyFitnessPal.²⁴ Apple Health is connected in the iOS app, and it is US-only. It reads and discusses that data; it does not re-prescribe from it.

Is ChatGPT a workout generator?

Functionally yes, but only for the first draft. Give it your equipment, schedule and goal and it generates a full ChatGPT exercise plan in seconds, with sets and reps.¹⁰ What it does not generate is the second week — nothing in it reads what you actually performed and updates the prescription accordingly.

Can I use ChatGPT in the gym?

Yes, and it is genuinely useful there for swaps and explanations — ask it for a substitute when a rack is taken, or what a movement is meant to train. It will not know what you lifted on the previous set unless you type it in, because its memory keeps summaries rather than logs.³

Does ChatGPT track progressive overload?

No. Progressive overload requires something that records performed sets and decides the next prescription from them.¹¹ ChatGPT has no such mechanism — a new conversation does not inherit last week’s loads. You can fake it by pasting your log each session, which is manual bookkeeping rather than a progression system.

Does ChatGPT make up studies?

Frequently, though less than it used to. One 2023 audit found 47% of references in ChatGPT-generated medical content were fabricated and only 7% were both authentic and accurate;⁵ a 2024 analysis put GPT-4’s rate at 28.6%.⁶ Verify any specific claim or citation before acting on it.

What is the best LLM for a workout plan?

For the plan itself, no frontier model is meaningfully better. Generation is the easy part — ChatGPT applies established training principles when asked¹⁰ — and Claude and Gemini produce comparable drafts. The real separator is what the tool can see: only ChatGPT publishes an Apple Health connection,² and none keeps a set-by-set training log.

What does ChatGPT cost compared with a fitness app?

On the US App Store on 17 September 2026, ChatGPT lists Go at $8, Plus at $19.99 and Pro at $100 or $200 per month, with a permanent free tier.⁷ Dedicated coaching apps are cheaper: SensAI $6.99/mo or $69.99/yr,¹⁸ Fitbod $12.99–$15.99/mo,¹⁹ Freeletics Training Coach $34.99–$79.99.²⁰

What is the best way to prompt ChatGPT for a workout plan?

Put equipment, goal, schedule, experience level, session length and injuries in a single opening message, since accuracy was high but comprehensiveness depended on inputs.¹ Then ask it to cite sources, verify those yourself, and paste your training log at the start of each session to compensate for the missing progression loop.

References

Zaleski AL, Berkowsky R, Thomas Craig KJ, Pescatello LS. “Comprehensiveness, Accuracy, and Readability of Exercise Recommendations Provided by an AI-Based Chatbot: Mixed Methods Study.” JMIR Medical Education, 2024;10:e51308. https://mededu.jmir.org/2024/1/e51308 ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹ ↩¹² ↩¹³
Mehta I. “OpenAI makes ChatGPT Health available to all U.S. users.” TechCrunch, 23 July 2026. https://techcrunch.com/2026/07/23/openai-makes-chatgpt-health-available-to-all-u-s-users/ ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹ ↩¹² ↩¹³ ↩¹⁴ ↩¹⁵ ↩¹⁶ ↩¹⁷ ↩¹⁸
OpenAI. “Memory FAQ.” OpenAI Help Center, accessed 17 September 2026. https://help.openai.com/en/articles/8590148-memory-faq ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹
Clover J. “ChatGPT’s Apple Health Integration Now Rolling Out to U.S. Users.” MacRumors, 23 July 2026. https://www.macrumors.com/2026/07/23/chatgpt-apple-health-integration/ ↩ ↩² ↩³ ↩⁴
Bhattacharyya M, Miller VM, Bhattacharyya D, Miller LE. “High Rates of Fabricated and Inaccurate References in ChatGPT-Generated Medical Content.” Cureus, 2023;15(5):e39238. https://pmc.ncbi.nlm.nih.gov/articles/PMC10277170/ ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹
Chelli M, Descamps J, Lavoué V, Trojani C, Azar M, Deckert M, Raynier JL, Clowez G, Boileau P, Ruetsch-Chelli C. “Hallucination Rates and Reference Accuracy of ChatGPT and Bard for Systematic Reviews: Comparative Analysis.” Journal of Medical Internet Research, 2024;26:e53164. https://www.jmir.org/2024/1/e53164 ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹
OpenAI. “ChatGPT.” US App Store listing (id6448311069), in-app purchase tiers, accessed 17 September 2026. https://apps.apple.com/us/app/chatgpt/id6448311069 ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷
OpenAI. “Memory and new controls for ChatGPT.” OpenAI, 13 February 2024. https://openai.com/index/memory-and-new-controls-for-chatgpt/ ↩
OpenAI. “Health in ChatGPT.” OpenAI Help Center, accessed 17 September 2026. https://help.openai.com/en/articles/20001036-health-in-chatgpt ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶
Washif JA, Pagaduan J, James C, Dergaa I, Beaven CM. “Artificial intelligence in sport: Exploring the potential of using ChatGPT in resistance training prescription.” Biology of Sport, 2024;41(2):209-220. https://pmc.ncbi.nlm.nih.gov/articles/PMC10955742/ ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶
Plotkin D, Coleman M, Van Every D, Maldonado J, Oberlin D, Israetel M, Feather J, Alto A, Vigotsky AD, Schoenfeld BJ. “Progressive overload without progressing load? The effects of load or repetition progression on muscular adaptations.” PeerJ, 2022;10:e14142. https://pmc.ncbi.nlm.nih.gov/articles/PMC9528903/ ↩ ↩² ↩³
Hammert WB, Kataoka R, Yamada Y, Song JS, Kang A, Spitz RW, Loenneke JP. “Progression of total training volume in resistance training studies and its application to skeletal muscle growth.” Physiological Measurement, 2024;45(8). https://pubmed.ncbi.nlm.nih.gov/39178897/ ↩
Lai X, Chen J, Lai Y, Huang S, Cai Y, Sun Z, Wang X, Pan K, Gao Q, Huang C. “Using Large Language Models to Enhance Exercise Recommendations and Physical Activity in Clinical and Healthy Populations: Scoping Review.” JMIR Medical Informatics, 2025;13:e59309. https://pmc.ncbi.nlm.nih.gov/articles/PMC12133071/ ↩
Manresa-Rocamora A, Sarabia JM, Javaloyes A, Flatt AA, Moya-Ramón M. “Heart Rate Variability-Guided Training for Enhancing Cardiac-Vagal Modulation, Aerobic Fitness, and Endurance Performance: A Methodological Systematic Review with Meta-Analysis.” International Journal of Environmental Research and Public Health, 2021;18(19):10299. https://pmc.ncbi.nlm.nih.gov/articles/PMC8507742/ ↩
Choi M, Park J, Lee M, Beom J, Jung SY, Lee K. “AI-Generated Exercise Prescriptions for At-Risk Populations: Safety and Feasibility of a Large Language Model Assessed by Expert Evaluation.” Journal of Clinical Medicine, 2026;15(6):2457. https://doi.org/10.3390/jcm15062457 ↩ ↩²
Anthropic. “Claude’s memory works everywhere, and you decide what’s in it.” Claude by Anthropic blog, 25 August 2026. https://claude.com/blog/claudes-memory-works-everywhere-and-you-decide-whats-in-it ↩ ↩²
Li A. “Gemini app rolling out ‘Past Chats’ personalization to free users.” 9to5Google, 26 February 2026. https://9to5google.com/2026/02/26/gemini-past-chats-free/ ↩
Origami, Inc. “SensAI: Fitness Sensei.” US App Store listing (id6738963099), accessed 17 September 2026. https://apps.apple.com/us/app/sensai-fitness-sensei/id6738963099 ↩ ↩² ↩³
Fitbod, Inc. “Fitbod: Gym & Fitness Planner.” US App Store listing (id1041517543), in-app purchase tiers, accessed 17 September 2026. https://apps.apple.com/us/app/fitbod-gym-fitness-planner/id1041517543 ↩ ↩² ↩³
Freeletics GmbH. “Freeletics: Workouts & Fitness.” US App Store listing (id654810212), in-app purchase tiers, accessed 17 September 2026. https://apps.apple.com/us/app/freeletics-workouts-fitness/id654810212 ↩ ↩²