901 Operators, 27 Countries: The Uncomfortable Truth About AI Search

It was at HFA in March at the CEO drinks event that Jon Brady, the president of Midtown Athletic Clubs, hosted, that he stood up and told the room that AI agents were already affecting his business. It had been a subject matter, an itch that I had been waiting to focus on, but at that time was neck deep in building Antares' conversational intelligence layer, which powers our agents. A conversation and exchange of messages with Jon and I found myself tipped into a rabbit hole I've only just climbed out of in the last week.

That rabbit hole became the largest piece of primary research we have done as a business. We audited 901 fitness operators across 27 countries against a framework that scores how ready a website is to be read, understood, and cited by AI agents. The headline finding is that the global fitness industry has barely started. The highest score recorded across all 901 sites was 68 out of 100. The global average was 21. Eighty-four per cent of operators sat at the level we call Passive: technically reachable by an AI agent, but offering no structured signal about what the business is, where it operates, or what it sells. Not one operator in the world has fully optimised for AEO yet.

The discipline at the centre of this is Answer Engine Optimisation. AEO is the practice of getting your brand to appear, accurately and favourably, in the answers that ChatGPT, Perplexity, Copilot, Claude, and Gemini give to your prospective members. It is not SEO with a new acronym. The underlying mechanics are different, and because of that the competitive map gets redrawn.

I will say plainly what I have spent 10 years thinking quietly, as both a marketing professional and a business owner. SEO became a market for charging operators a fortune to chase a position on a page they were never going to reach. There were genuine technical fundamentals underneath, but the agency layer above it ran on aspiration. AEO is different. The work is more technical, more structural, and the outcomes are measurable in a way SEO outcomes often were not.

The most uncomfortable part of the research was not the operators who had done nothing. It was the operators who had done something, and got it wrong. Forty-nine sites in the audit, including some of the most recognisable multisite groups in the industry, have explicit rules in place that forbid AI agents from reading their pages at all. In most of those cases the rules were not a considered policy decision. They were the default setting on a Cloudflare feature called "AI Scrapers and Crawlers," switched on by an agency or a developer who understood the words "block AI" but not the consequence. The result is that some of the largest, best-resourced operators in fitness are currently invisible to ChatGPT, Claude, and every other answer engine. They have paid, in effect, to be excluded from the conversation.

There is a wider context that makes this matter now rather than later. A February 2026 analysis of 858,457 sites recorded 68.9 million AI crawler visits in a single month, with 59% of all sites receiving at least one. Total LLM referral traffic to those sites grew 72.7% year over year. The behaviour is already operating at scale across the open web.

For fitness operators specifically, three things are true:

1. AEO resets the visibility race, so the operators who have sat comfortably on page one of Google for a decade do not get to bring that ranking with them.

2. The visitor profile is different, with research showing AI-allowed sites draw 3.2 times more human traffic and 2.7 times more form completions than sites that are blocked or invisible.

3. The same model that answers "best gym in London" today will, within the next two years, complete the trial booking, ask the qualifying questions, and hand a warm prospect to your front desk.

In my next article, I will take a deeper look at what can be done to give you the best chance of prospering from these changes. It is not rocket science, and if you know what you are doing it can make a near instant improvement and pay dividends for a long time to come.