A good strength coach does two things. They watch you train and they think about your training when you are not there. They notice that your squat has not moved in three weeks. They notice that you added two extra sets to arms on Tuesday and then could not recover for legs on Thursday. They notice that your overhead press always stalls when your bench press volume is high. They hold these patterns in their head and use them to adjust your programme before problems become serious.
This is what coaching is. Not motivation. Not yelling louder. Pattern recognition across time, applied to programming decisions.
The problem is that most people cannot afford a coach. A decent one charges £50 to £150 per month. A good one charges more. And even the ones you can afford are limited by their own memory and attention span. They have twenty clients. They review your data once a week. They miss things.
AI coaching is not replacing human coaches. It is making the analytical side of coaching available to every lifter with a phone.
What AI Actually Does Well
Trend Analysis Across Time
Humans are bad at spotting slow trends. A 1 percent decline per week in your squat e1RM is invisible session to session but amounts to a 12 percent decline over three months. Your brain does not naturally compute compound trends. It compares today to last week and calls it flat. An AI system compares today to last month and last quarter and catches the drift before it becomes a crater.
Volume and Balance Tracking
How many sets of push versus pull did you do this week? What about this month? Is your push-to-pull ratio 2:1 when it should be closer to 1:1.5? Most lifters have no idea. They train what they enjoy, neglect what they do not, and wonder why their shoulders hurt. An AI system tracks volume by muscle group and movement pattern automatically, and flags imbalances before they cause injury.
Fatigue Modelling
The acute-to-chronic workload ratio is a concept borrowed from sports science. Your acute load is what you have done recently. Your chronic load is your longer-term average. When acute load spikes well above chronic load, injury risk increases. When it drops well below, you are detraining. Keeping the ratio in a productive range is one of the most important but least visible aspects of programming.
Calculating this by hand requires tracking total volume across all exercises, all sessions, for weeks at a time. An AI system does it automatically and warns you when you are pushing too hard or backing off too much.
Context-Aware Suggestions
The difference between a generic chatbot and a coaching AI is context. A chatbot can tell you what a deload is. A coaching AI can tell you that you specifically need to deload your bench press by 15 percent for two sessions because your e1RM has been flat for five weeks while your overhead press continued to progress, which suggests the issue is local fatigue in your chest and front delts rather than systemic fatigue.
That level of specificity requires knowing your training history, your current programme, your exercise selection, your rep ranges, your recent performance trends, and your injury history. Without that context, any suggestion is generic. With it, the suggestion becomes actionable.
What AI Does Not Do Well
AI cannot watch your form. It cannot see that your squat depth has shortened by two inches over the last month. It cannot see that you are compensating with your lower back on deadlifts. It cannot see that your right elbow flares on bench press. These are visual cues that a present coach catches and an app cannot.
AI cannot read your emotional state. Some people need to be pushed. Some people need to be held back. Some people respond to data and logic. Some people respond to encouragement and empathy. A human coach adjusts their communication style to the individual. An AI system gives you the same tone regardless.
AI cannot account for everything in your life. It can ask about sleep and stress and energy, but it cannot know that you had a terrible argument with your partner, that your kid was up all night, or that you skipped breakfast because you were running late. These things affect your training. A human coach in the room with you sees it in your face. An AI system does not.
This is why AI coaching supplements human coaching rather than replacing it. The ideal setup, for anyone who can afford it, is a human coach for the subjective side and an AI tool for the data side. The human watches your form and manages your psychology. The AI watches your numbers and manages your periodisation.
How It Works In Practice
You finish a training session and log your sets. The AI system processes the data. Behind the scenes it updates your e1RM estimates, recalculates your training volume by muscle group, checks for plateau signals, and adjusts your fatigue model.
If everything is on track, nothing happens. You train the next session as planned. The system is working but you do not see it because there is nothing to flag.
If something is off, the system surfaces it. Your squat e1RM has been flat for four sessions. Your push volume is 40 percent higher than your pull volume. Your training load has spiked above your chronic baseline. Each flag comes with a specific suggestion and a one-tap action. Accept the deload. Rebalance your push and pull. Reduce your accessory volume for a week.
Between sessions, you can ask the coach questions. "Why is my overhead press stalling?" The AI looks at your data and responds with specifics. Your overhead press volume is fine but your bench press volume increased three weeks ago, and the two exercises share anterior delt and tricep demand. Reducing bench volume or swapping to a variation with less front delt involvement might free up recovery for overhead press. Here is the swap, tap Apply.
This interaction takes 30 seconds. Without AI, this analysis would require you to manually review weeks of training data, calculate volume by muscle group, cross-reference exercise overlap, and arrive at the same conclusion. Most lifters simply do not do this. They change things by feel, or they do not change at all.
The Economics of AI Coaching
A human coach costs £50 to £150 per month. An AI coaching tier in a gym app costs £5 to £15 per month. The value proposition is not "AI is as good as a human coach." It is "AI gives you 60 percent of the analytical value at 10 percent of the cost."
For the vast majority of lifters who train alone, that is a good deal. They were not going to hire a human coach anyway. The alternative to AI coaching is no coaching. And no coaching means plateaus last months instead of weeks, volume imbalances go undetected until something hurts, and programme changes are based on feel rather than data.
The cost to run AI coaching is also dropping. Prompt caching, smaller specialised models, and efficient context management mean that each coaching interaction costs fractions of a penny to operate. This makes the subscription model sustainable for the developer and affordable for the user.
Where This Goes Next
The near-term future of AI coaching in strength training is better context and better integration. Wearable data, sleep tracking from smartwatches, heart rate variability, all feeding into the readiness model. Camera-based form analysis using phone cameras, identifying technique drift before it causes injury. Integration with nutrition tracking so the AI knows you have been in a 500-calorie deficit for three weeks and adjusts its expectations accordingly.
The longer-term future is genuine personalisation. Not just reacting to your data, but learning your individual response patterns. How quickly you adapt to volume increases. How many sessions you can push before you need a deload. Which exercise variations produce the best transfer for you specifically. This requires years of accumulated data per user, but the systems that start collecting that data now will be the ones that can offer truly personalised coaching in the future.
For now, the bar is simpler. Does the AI read your training data? Does it spot problems? Does it tell you what to change? If yes, it is worth using. If no, it is a chatbot with a gym theme.