TL;DR
A general chatbot (ChatGPT, Claude, or Gemini) will write you a solid 3-4 day training plan for free — the catch is they default to the “influencer split”: 5-6 days, 90 minutes, perfect sleep assumed. The fix is the prompt below, which forces real days/week, a low-energy version of every session, and a mandatory deload week. Put the plan in a persistent thread (a ChatGPT Project or saved memory) so progression carries over week to week instead of resetting every chat. AI builds the structure; only a human coach catches form drift.
The task
You want to work out 3-4 days a week. You have either a commercial gym, a corner of your apartment with a kettlebell and a pull-up bar, or nothing but a yoga mat. You’ve started a 6-day Instagram split twice, made it to week 3 both times, and quit when one bad sleep week killed your streak. You want a plan that survives a bad week, accounts for the fact that Tuesdays you’re exhausted and Saturdays you have energy, and progresses you toward strength or fat loss without making you choose between “go hard” and “do nothing” on any given day.
Which AI to use (and is free enough?)
For workout planning, the free tier of any major chatbot is genuinely enough to get a good plan. The reason to pay is continuity — keeping one training thread alive so the model remembers last week’s loads and adjusts this week.
| Tool | Free plan | Paid entry | Best for workout planning | As of June 2026 |
|---|---|---|---|---|
| ChatGPT (GPT-5.5) | Yes — ~10 msgs / 5 hrs on GPT-5.3, then mini | Plus $20/mo | Persistent Projects + cross-chat memory to hold your plan and history | Memory now reaches free users too |
| Claude (Sonnet 4.6 / Opus 4.7) | Yes — limited Sonnet 4.6 | Pro $20/mo ($17 annual) | Long, careful single-shot programs; honors constraints tightly | 1M-token context on paid |
| Gemini 3.1 Pro | Yes — via Google account | Google AI Pro $19.99/mo | Plugs into Google Calendar/Workspace if you block sessions there | Renamed from “Gemini Advanced” early 2026 |
Verdict: start free in ChatGPT and create a Project named “Training” so every session log lives in one place — see our ChatGPT Projects guide and ChatGPT memory guide. Claude Pro is the better pick if you want one long, constraint-heavy program written in a single pass. None of these track sets for you — for auto-logging and auto-progression you’d add a dedicated app like Hevy, Dr. Muscle, or Alpha Progression on top.
Where AI helps — and where it does not
AI is excellent at assembling a plan from established progressions (linear, RPE-based, double progression), balancing the split for recovery, and proposing substitutions when an exercise isn’t available or aggravates an injury. It can also pre-bake a low-energy version of each session — the thing that actually keeps habits alive through bad weeks. What AI cannot do: coach your form, rehab an injury, prep you for a competition, or watch your bar path. For any of those, see a real coach. AI gives you a structured plan; only a human (or a mirror with attention) catches form drift.
The named failure mode: the influencer split — the model gives you a 5-6 day program that assumes 90 minutes per session, perfect sleep, optimal nutrition. You start it Monday, miss Thursday, feel guilty Friday, quit Saturday. Force the prompt: realistic days per week, realistic minutes per session, and a low-energy version baked in for every session.
What to feed the AI
- Your primary goal — pick one: strength, general fitness, fat loss, endurance, mobility
- Equipment available — commercial gym, home gym with specifics (rack, barbell, dumbbells, bands), bodyweight only
- Realistic days per week and minutes per session — actual, not aspirational
- Injury history and current niggles — explicit (“knees on heavy squats”, “left shoulder limits overhead press”)
- Training age — first 6 months, 1-2 years, 3+ years (progression speed differs)
- Sleep / stress baseline — honest, because it changes the plan
- 1-2 movements you love (more adherence) and 1-2 you hate (substitute, don’t fight)
- The single behavior you want to be doing automatically by week 8
Copy-ready prompt
Use placeholders in square brackets and fill each one before sending. Works as-is in ChatGPT (GPT-5.5), Claude (Sonnet 4.6), or Gemini 3.1 Pro.
Build my weekly workout plan.
Primary goal (pick one): [strength / general fitness / fat loss / endurance / mobility]
Equipment: [commercial gym / home setup with specifics / bodyweight]
Realistic days/week: [3 or 4 max]
Realistic minutes/session: [45-75, honest]
Injury history and current niggles: [list — be explicit about what triggers what]
Training age: [first 6 months / 1-2 years / 3+ years]
Sleep + stress baseline: [good / mediocre / chronically poor]
2 movements I love + 2 I hate: [paste]
Behavior I want automatic by week 8: [paste]
Return:
1) Week split (3-4 days only), balanced for recovery. Show which muscle groups recover before the next session hits them.
2) Per session: warmup, main lifts/movements (sets x reps + intensity), accessories, cooldown. Total within stated minutes.
3) Progressive overload rule — what specifically I increase week to week (load, reps, RPE, density).
4) Low-energy day version of EACH session — half the volume, same movement pattern, counts as a workout for streak purposes.
5) Deload week schedule (every 4-6 weeks) — what changes that week.
6) Two movement substitutions per main lift — for injury, equipment unavailability, or "just hate today."
7) Adherence safety net: the minimum version of this week that still counts as 'on plan' even if I miss 2 sessions.
Rules:
- Honor my real days/week. Do not give me 5 days if I said 3.
- Account for the niggles I listed. Do not put loaded barbell squats on someone with knees-on-heavy-squats unless cued as optional.
- Every main lift must have a low-energy alternative.
- Deload is mandatory — strength stalls and injury risk climb without it.
- The behavior I want automatic by week 8 should be reinforced by the structure (e.g., consistent first-lift across sessions).Shorter variant — 20-minute habit-builder
20-minute habit-builder workout plan for the first 4 weeks.
Goal: [one]. Equipment: [paste]. Days/week: [3].
Each session: 1 main movement + 2 supersets + 2-minute cooldown. Total 20 min cap. Plus a 10-minute "really tired but want to keep streak" version.Keep the plan alive across weeks
The single biggest upgrade is not a better prompt — it’s a persistent thread. In ChatGPT, open a Project called “Training” and drop the plan plus a one-line log after each session (“Wk2 Day1: bench 4x5 @60kg, all reps, easy”). Next week, ask: “Apply my progression rule to this week’s numbers from the log above.” The model now progresses you off real data instead of guessing. Claude users can pin the program in a single long thread; Gemini users can have it write the week into Google Calendar.
Sample output
A useful 3-day split (commercial gym, strength, 60 min): “Day 1 (push-focus) — Bench 4x5, Overhead press 3x6, Dips 3x8, Triceps 3x12. Day 2 (lower) — Squat 4x5, Romanian deadlift 3x6, Walking lunges 3x10, Calf raises 3x15. Day 3 (pull-focus) — Deadlift 3x4, Barbell row 4x6, Pull-ups 3xAMRAP, Curls 3x12. Recovery: 48 hours between heavy lower sessions, push-pull separated by a day. If you only do 2 of 3 sessions in a week, that’s still ‘on plan’ — do not skip and catch up the third on a fourth day.”
A useful low-energy version: “Bench day, low-energy: skip accessories, do main lifts at 70% normal load with same reps. 25 minutes total. Counts as a workout for streak purposes — keeps the habit without overloading recovery. Better than skipping.”
A useful progressive overload rule: “Add 2.5 kg to the main lift the week after you completed all prescribed sets and reps with form intact. If you missed any rep on the heaviest set, repeat the same load next week. No ‘I think I can’ jumps — earn the increase.”
A useful deload week: “Every 4-6 weeks (or when bar speed slows on rep 3 of your top set): drop main lift loads to ~60% of working weight and halve accessory volume, keeping the same movements. Continue sessions; the goal is preserving the habit while reducing fatigue. Many people skip deload, plateau, and quit. Plan it before you need it.” (This matches the published research range: surveyed strength athletes deload for ~5-7 days every 4-6 weeks, cutting volume or intensity by roughly 30-50%.)
A useful substitution set: “Bench press substitutes: dumbbell bench (similar pattern, kinder to shoulders) or push-ups + dip (no equipment). Don’t switch to incline if shoulder is the issue — incline loads it more.”
How to refine
- Cut to my real days: “I have 3 sessions, not 5. Cut the plan to 3 days. The cut sessions don’t get redistributed — they’re cut.”
- Bake in the low-energy version: “Every session needs a low-energy alternate that takes half the time and same movement pattern. The low-energy version is what keeps the habit alive through bad weeks.”
- Slow the progression: “Make the progression conservative. Better to under-shoot for 12 weeks than burn out at week 4. If I’m new, linear progression every other week is fine; not every week.”
- Honor the niggles: “I have shoulder issues on overhead press. Substitute with landmine press or floor press. Do not list overhead press as ‘optional’ — substitute it.”
- Plan the deload: “Insert a deload week at week 5. Define what changes that week. Do not skip — I will plateau or get injured without it.”
Common mistakes
- Copying a 6-day Instagram split when you have 3 days available — sets you up to fail by week 4
- No deload week — strength gains stall, injury risk climbs, you quit and blame yourself instead of the plan
- Skipping the low-energy version — most people quit when one bad sleep week kills their streak; the low-energy version saves the habit
- Aggressive progression (adding load every session) — works for 4 weeks, fails for 4 months
- Ignoring listed niggles (“the plan has loaded squats but my knees…”) — pain is information, not a sign of weakness
- Treating cardio and lifting as opponents — they coexist if dosed correctly for your goal
- No movement substitutions — equipment unavailability or one off day shouldn’t blow up the plan
- Starting a fresh chat every week so the model forgets your loads — use a Project or memory instead
FAQ
- How long until I see results?: Roughly 4 weeks for the habit to feel automatic, 8 weeks for visible strength changes, 12 weeks for body composition change. If you see nothing meaningful at 12 weeks, the problem is usually nutrition or sleep, not the workout.
- Should I include cardio?: Goal-dependent. For strength: 2x light cardio per week, separated from leg day. For fat loss: 3x cardio, mix of low-intensity (zone 2) and one high-intensity session. For endurance: cardio is the plan; lifting is 2x supplementary.
- What if I miss a week?: Restart on the same week, not the next. Don’t double up — fatigue compounds and the next week you’ll quit. Streak protection beats linear progression.
- AI keeps suggesting movements I hate. How do I bias it?: Tell it explicitly in the prompt: “I hate
[movement]— substitute and never propose it again.” Adherence beats theoretical optimality. The ‘best’ exercise you skip is worse than the ‘good’ exercise you do consistently. - Is the free version of ChatGPT or Claude enough?: For building and refining a plan, yes — the free tiers handle this easily (as of June 2026, ChatGPT Free allows ~10 messages per 5 hours on GPT-5.3). The reason to go paid ($20/mo for either) is persistent memory and higher limits so one training thread can run for months.
- Should I track everything?: Track main lifts (load + reps). Accessories: only track if it helps adherence. Tracking warmups and rest periods is usually procrastination dressed as rigor. If you want auto-logging, pair the plan with a tracker app rather than the chatbot.
Further reading
- The Effects of Overtraining and Recovery (NSCA) — why deloads exist, in coach-grade detail