Blog Benchmarks · Beyond the Four Walls · Part 10
We scanned 98 fitness and wellness websites. None was ready for agents.
Gyms, studios, med spas, salons and recovery chains, scored on whether an AI assistant can find them, understand them and book. The median was 46 out of 100. Not one site lets an agent book a class or appointment.
Over nine parts we have argued that customers are starting to send AI assistants to find, compare and book, and that most local business websites were built for a person who reads, not for software that acts. That is an easy thing to claim. So we measured it.
On October 10, 2026 we ran Agent Ready, our free scanner, over the corporate websites of 100 multi-location fitness and wellness brands in the US: gyms, boutique studios, med spas, salons, massage and stretch chains, recovery and IV chains, and wellness clinics. Ninety-eight scanned cleanly. Two big gym brands timed out twice and are left out rather than guessed at.
Source: Sneeze It Agent Ready scan of 98 US fitness and wellness brand websites, October 10, 2026
The scores
The average was 43.2 and the median 46. The best site scored 76. Most of the industry sits in the middle band: readable, reachable, and almost completely unusable by an agent that wants to get something done.
Source: Sneeze It Agent Ready scan, October 10, 2026. Number of sites in each score band.
Med spas and wellness clinics did best. Salons and gyms did worst. Each category is small, a dozen or so brands, so read the order as a direction, not a ranking.
Source: Sneeze It Agent Ready scan, October 10, 2026
Agents can get in. They just cannot do anything.
Agent Ready scores four things, the four questions from our checklist: can an agent reach the site, understand it, act on it, and complete a booking? On the sites the scanner could get into, the pattern is stark.
Source: Sneeze It Agent Ready scan, October 10, 2026. 82 sites the scanner could read; Complete is scored on the 57 with a booking or sign-up path.
Reach is fine. Almost everyone lets agents in. Comprehend is middling. Act, the part that decides whether an assistant can actually book you, has a median of 14 out of 100.
Here is what that looks like check by check, on the 82 sites the scanner could read:
| What an agent needs | Sites that had it |
|---|---|
| Phone and email as real links | 7.3% |
| Forms an agent can fill in (labels, names, autocomplete) | 7.3% |
| Structured data with the business facts | 6.1% |
| An llms.txt briefing | 29.3% |
| A tool list for agents (MCP manifest) | 0% |
| Agent tools for booking (WebMCP) | 0% |
Five sites did have WebMCP tools. All five expose the same twelve shopping tools for their online store, which points to their store platform switching them on, not to the brand building them. Not one of the 98 offers an agent a way to see class times or book an appointment.
Nobody is blocking AI. The walls are something else.
We expected to find brands blocking AI assistants in robots.txt. We did not. Of the 82 sites with a readable robots.txt, none blocked the live assistants that do tasks for a person: ChatGPT-User, Claude-User, Perplexity-User or OAI-SearchBot. Only three blocked anything AI-related at all, and those were training crawlers.
The real barrier is security. 18 of the 98 sites sat behind a bot wall, the "checking your browser" page that a person clicks through and an agent cannot. From the outside we cannot tell whether anyone chose to turn agents away. Most of these walls come from firewall and CDN security settings that were switched on long before anyone sent an assistant to book a massage.
Source: Sneeze It Agent Ready scan, October 10, 2026. 16 flagged by the scanner plus 2 confirmed by hand.
And at the finish line, on the 57 sites with a booking or sign-up path the scanner could reach, 30 had a captcha an agent cannot pass. That is 52.6%. On a login or checkout, a captcha is doing its job. On a "book your free class" form, it is turning away the customer's assistant.
What the best sites do differently
The top ten scored between 63 and 76. Compared with the bottom half, the gap is not budget or design. It is a handful of plain, cheap things.
Bottom half (49 sites)
4 of 40 readable sites publish an llms.txt. 1 of 40 has structured business data. 0 of 40 have forms an agent can fill in. 27 of 40 show their content without JavaScript.
Top ten
9 of 10 publish an llms.txt. 3 of 10 have structured business data. 3 of 10 have forms an agent can fill in. 10 of 10 show their content without JavaScript.
The gap between the top and the bottom of this industry is about an afternoon of work, not a redesign.
The single biggest difference is llms.txt, a short plain-text briefing at the root of a website that tells an AI what the business is, what it sells and which pages matter. Nine of the top ten have one. Four of the bottom forty do. It takes an hour to write.
How we did it
We picked 100 US multi-location brands across seven categories, using each brand's corporate homepage rather than a single location. We left out our own clients and a few brands we work with in other ways. We checked every domain by hand, replaced four that were parked, redirected abroad or pointed at a single location, and scanned each site once, one at a time, with one retry on failure.
What to do on Monday
You do not need to wait for anyone. Every item on this list is something your web team can do this month, and none of them changes how the site looks to a person.
That is the end of Beyond the Four Walls. If you missed the start, Part 1 makes the case, and Part 2 explains every piece of plumbing on this list in plain English. Or skip the reading and see where your own site stands.
Sources
- Sneeze It, Agent Ready scan of 100 US multi-location fitness and wellness brand websites, October 10, 2026. 98 completed. Method and scoring: agentready.sneeze.it/methodology
