The Operator Letter

Here is the conversation I have most weeks with a fitness CEO. They tell me about the chatbot they have just bought. They tell me about the lead-scoring model their CRM vendor is rolling out. They tell me about the dashboard refresh where everything is now AI-powered. They are pleased with themselves. They have moved on. The board update is filed. The slide is made. Next quarter is somebody else's problem.

I nod. I ask a different question. I ask them to open a private browser window, go to ChatGPT, and type the kind of question one of their members might type. Something like best gyms in Manchester for someone who travels for work. Something like is Third Space worth the money compared to Equinox in London. Something like what is the most family-friendly health club in Surrey.

Then we sit there together and watch the answer come back. And in most cases, their brand is not in it.

That is the thing no one is telling fitness CEOs about AI. The conversation they should be having is not about whether to buy a chatbot. It is about the fact that a meaningful and growing fraction of their next ten thousand prospective members will not find them by typing the brand name into Google. Those prospects are asking a machine for a recommendation. The machine is making one. And the recommendation is, increasingly, somebody else.

This is not an in-five-years thing. It is happening now. The fraction is in the single digits today and the rate of change is the only number that matters.

I am going to be more specific. There is a category of software people are calling AI Engine Optimisation, or AEO. It is roughly what SEO was twenty years ago, and it has the same air about it: vaguely embarrassing, easy to dismiss, sold by people who are not always serious, and almost certainly going to matter for the next decade. The thing AEO actually does is unfussy. It looks at how the language models view your business when somebody asks them to recommend one. It establishes whether a model has a structured idea of who you are, where you operate, and what you sell. It then asks the model the questions a prospect would ask, ranks your visibility against your competitors, and tells you what to fix.

That last sentence is the one I want you to read twice. The fix is rarely what a vendor will sell you on top of it. The fix is usually one or more of: machine-readable copies of your opening hours, your locations, your prices, your facilities. A llms.txt file at the root of your domain that tells AI crawlers in plain English what to do with you. JSON-LD schema in the head of your location pages. A coherent home page that does not assume the reader already knows what a HYROX class is, or that you have a swimming pool. None of this is glamorous. None of it has an awards category at the industry conferences. None of it photographs well.

It is also, I would say, the single highest-leverage thing most fitness operators could do this year.

Now. The reason no one is telling you this is straightforward and slightly bleak. The marketing-tech vendors who sell into your business have an installed base of customers who do not have any of these problems solved yet, and the path of least resistance is to keep selling net-new tools rather than admit that the foundations underneath your existing website are not picked up by the new generation of crawlers. The agencies are still optimising for Google. The conference circuit is still selling personalisation engines and computer-vision turnstiles. There is a generational lag of about eighteen months between when an industry shift becomes obvious and when the trade press starts writing about it. We are about six months into the lag. You have a window.

I will tell you what I would do, with the caveat that this is a letter and not a consultation.

I would start by asking the obvious questions in the AI engines yourself. Not your marketing lead. You. ChatGPT, Claude, Gemini, Perplexity. Spend half an hour. Ask the questions a member would ask. Pay attention to who shows up. Pay attention to whether you show up. Pay attention to whether what the engine says about you is accurate, and how accurate. If you are not there at all, that is your starting point in red. If you are there but the engine has a wrong fact about you, that is your starting point in amber. Either way, you now have something concrete to chase.

I would then look at your website the way an AI crawler looks at it. The single most useful exercise is to view-source on your top three location pages and search the HTML for the words schema.org. If you find them, good, somebody has done some of the work. If you do not find them, your problem is larger than you think. Your competitors who have it will keep getting recommended; the engines have a structured idea of them. You do not exist as a structured object on the web. You exist only as paragraphs of marketing copy, and the engines weight that less every quarter.

I would then make peace with the fact that this is going to mean an unfashionable few months. The work is technical. It is mostly invisible. It does not photograph well. It does not produce a quotable case study at the end of it. The board will not light up. What it produces is the ability to be recommended by a machine that, two years from now, will be doing a substantial share of the recommending. That is the trade. You are not paying for an output you can put in a quarterly review. You are paying for the right to keep existing in a category of search you currently do not exist in.

Now, the bad case. A lot of you will not do any of this. You will buy a chatbot. You will give it a tasteful name. You will put a small photo of it on your home page. Your marketing lead will produce a slide for the operating board called Our AI Journey. The slide will be impressive. The board will be pleased. The numbers will not move, and you will not be able to say exactly why. Twelve months later your cost per lead will be up, your share of the conversation in your city will be down, and a smaller competitor with one good engineer will be quietly winning the share you used to take for granted.

The good case is that you treat this the way you would treat any other piece of infrastructure debt. You do the work. You stop being recommendable only by Google and start being recommendable by the half-dozen things that will dominate discovery for the next decade. You quietly outrun the competitor who is spending the same money on a chatbot.

I will tell you the thing that finally got through to one CEO I work with. He said, with some heat, that this was all very interesting but he was running a business and his job was footfall and revenue, not foundational web work. I asked him how many leads his Meta budget brought him last month. He told me. I asked him how many of those went on a member tour. He told me. I then asked him what would happen to the spend per lead if a generation of his prospective members did not go through Meta at all, because they had already asked ChatGPT, got a competitor recommendation, walked into the competitor, and never appeared on his attribution dashboard.

He went quiet. Then he went and fixed his schema.

That is the conversation. It is unglamorous. It is happening whether or not you are paying attention. The vendors who profit from your inattention are betting on the fact that you will keep your attention elsewhere. I would not.

The next letter, in a fortnight, is on a related question: which AI investments fitness operators are quietly regretting, and the one investment almost no one has made that I would make first. If that sounds like the kind of thing you want in your inbox at seven in the morning on alternate Fridays, the subscribe form is below.