AI competitor analysis for gyms: the 1-hour-a-month system
Keeping a systematic eye on your competition costs one hour a month if you use AI for the heavy lifting, and the most valuable output isn't their prices: it's their negative reviews. Every complaint about the gym across the street is an opportunity for you, backed by proven demand. "Machines always taken at 7pm" in their reviews means your ad for "train without waiting" already has an audience that's paying to avoid exactly that problem. This article builds that system: what to track, how to process it with AI, the monthly report prompt, and the discipline that makes it useful rather than obsessive.
This fits into the analysis role within the map of AI agents for gym marketing: here AI decides nothing, it prepares the decision for you.
What to track for each competitor
Start by narrowing the field. Your real competitors are the 3–5 facilities a prospective member would compare you against: same radius (2–3 km in the city, more in spread-out areas) and the same value proposition range. The €19.90 low-cost chain and the €120 boutique studio don't compete with each other even if they share the same street. More than five competitors on your radar is noise.
For each one, five things:
- Public prices and offers. What they advertise on their website, Google, and window display. Not what they actually charge after negotiating, you won't know that, and it matters less than you think.
- Active ads in the Meta Ad Library. The ad library is public and free: you can see every active ad for any page, how long it has been running, and what the creatives look like. An ad that has been live for three months is probably working; that's market research your competitor paid for. How to exploit it fully is covered in the Meta Ad Library guide.
- New Google reviews. The positive ones tell you what their customers value; the negative ones are your opportunity list. This is the most underrated source of the five.
- Social content that's working for them. Not their entire feed, only what stands out in engagement. One of their reels with 10x their usual likes signals a format or topic that resonates in your area.
- Schedule and service changes. New 6:30am class, pool closure, opening of a cross-training zone. Operational changes signal their strategy better than their advertising does.
The sources, all free
| Source | What it gives you | Review frequency |
|---|---|---|
| Meta Ad Library | Active ads, creatives, age of each ad | Monthly |
| Google Business Profile (Maps) | New reviews, average rating, customer photos, Q&A | Monthly |
| Competitor website | Prices, offers, services, hours | Monthly |
| Competitor Instagram | Content with traction, offers in pinned stories | Monthly |
| Google Alerts for their name | Local press mentions, openings, events | Automatic |
| Your own front desk | "I'm coming from gym X" and why they left | Ongoing |
The last row isn't a joke. Every sign-up who comes from a competitor knows something no tool will tell you. One question on the sign-up form ("were you at another gym? what made you switch?") is first-hand competitive intelligence, and it's free.
How to use AI on each piece
AI doesn't find the data for you (not yet, reliably scraping these sources is still the domain of dedicated tools). What it does is compress the reading, which is where the time actually goes.
Summarise 50 reviews into 5 themes. Copy the competitor's reviews from the last 3–6 months (Google Maps shows them all; copying and pasting 50 takes five minutes) and ask a language model to group them: praised topics, criticised topics, number of mentions for each, and two verbatim quotes per topic. What used to be an hour of reading per competitor is now ten minutes in total. Always ask for verbatim quotes: summaries without them tend to soften the edges, and it's the raw complaint that contains the ad angle.
Compare propositions from the websites. Paste your website text and theirs and ask for a comparison table: what each one promises, who it seems aimed at, what one offers that the other doesn't mention. The result is usually uncomfortable, which is good: most gym websites promise the same thing in the same words, and seeing it in a table pushes you to differentiate.
Generate the monthly report. This is the prompt that turns the one hour of collection into a decision document. Save it and reuse it every month:
You are a competitive intelligence analyst for a local gym.
My gym: [name, proposition, price, area].
Competitors being tracked: [list].
This month's data (pasted below):
1. New Google reviews for each competitor
2. Active ads in the Meta Ad Library (description of each and start date)
3. Changes detected on websites/prices/schedules
4. Posts with unusual traction on their social channels
Return:
- 5-sentence summary of what changed this month
- The 3 most repeated complaints about competitors that I can address
(for each: verbatim quote + suggested ad angle)
- Competitor ads that have been active for 60+ days (probably profitable)
- ONE recommended decision for this month: an offer, a creative
angle, or a service change. Just one, with the reasoning.
[DATA PASTED HERE]
If you're already generating ads with AI, the angles that come out of this report feed directly into your ad creation prompts: you pass the competitor's verbatim complaint as context and ask for variants that address that friction without naming anyone.
The analysis that matters: other people's negative reviews
This deserves its own section because this is where the money is. A negative review of your competitor is a customer in your area, with proven budget (they were paying), describing in their own words what they need and aren't getting. There is no cheaper market research.
The mechanics: from the complaints the AI has grouped, look for the ones that meet two conditions. They repeat (one mention is an anecdote; five is a pattern) and you can genuinely solve them. The second condition is the one everyone skips. If their complaints about crowding at 7pm overlap with your own peak-hour crunch, you don't have an angle, you have the same problem.
When both conditions are met, the path is direct: the complaint becomes a creative angle without ever naming the competitor. "Machines taken" → "reserve your rack, train without queuing". "Nobody corrected my form in six months" → "our floor trainer knows your name by week one". "Impossible to cancel" → "no commitment, leave whenever you want". The ad works because it doesn't invent a need: it responds to a frustration that already exists 500 metres from your door.
And the uncomfortable mirror: what AI does with their reviews, anyone can do with yours. Your negative reviews are your competitor's opportunity list, which is one more reason to manage your online reputation seriously and respond to every review, especially the bad ones.
What not to do
Obsess over their prices. Classic mistake. Your competitor's price is contextual data, not an instruction. If you drop to their price without their cost structure, the only thing you've copied is their margin, not their model. Price wars between neighbourhood gyms are won by the low-cost chain that wasn't participating.
Copy offers without understanding the model. Their "first month free" might work because they have a 12-month minimum that recovers the cost, or because their lead volume lets them burn margin on acquisition. Copied without those components, the same offer fills your gym with free-month tourists. Before copying anything, answer one question: what makes this offer profitable for them, and do I have that piece?
Mention competitors in your ads. Not gracefully, not with "honest comparisons", not with recognisable hints. Legally it's murky ground, and commercially you're giving away awareness: some of the people who see the ad will go check out the competitor you just named for them. The right angle attacks the problem, never the rival.
Spend more than one hour a month on it. Competitive intelligence has very fast diminishing returns. The second hour of the month scrolling a competitor's Instagram doesn't produce information, it produces anxiety. Your edge won't come from knowing them better; it will come from executing your own thing better. They also have your negative reviews to read.
One decision per month, just one
The discipline that separates this system from a hobby: each monthly report ends in a decision, and just one. A new offer, a new creative angle, or a service change. Not three. One.
The reason is practical. If you change three things at once you won't know which one worked, and a gym doesn't have the statistical volume for multiple simultaneous tests. One decision a month is twelve experiments a year with clean read-outs, more than 95% of the gyms in your area will ever run.
A complete cycle example: in the October report, AI detects that competitor A has accumulated eight new complaints about overcrowded group classes, and that their oldest running ad (4 months active) promotes exactly those classes. Decision for the month: launch a "limited spots per class, guaranteed reservation" angle with a creative showing your class with visible space. In November you measure: if the cost per lead for that angle beats your average, it stays and scales; if not, it gets filed and the November report proposes the next one.
An agent platform can close part of this loop on its own: in Pilotium, for example, the chosen angle goes to a campaign with its own creatives without going through the ads manager. But the human piece doesn't disappear: deciding which competitor complaint to attack this month is still your judgement call, and that's exactly what the hour you invest is worth.
Start today with the fifteen-minute version: open the Meta Ad Library, search for your three main competitors, and note how many ads each has and how long they've been running. If any of them has been advertising for months and you haven't, you already have the first conclusion of your first report.