Benchmark · scanned 1 September 2026
Can an AI assistant book a salon appointment?
Ten of the largest salon booking platforms were re-scanned on 1 September 2026. The median agent readiness score was 56 out of 100, the best was 66, and not one of the ten published a capability manifest — the file that tells an assistant a booking can be made and how.
Colour International LLC · 1 September 2026 · colourintl.com/benchmarks/salon-booking-platforms
Salon owners cannot fix this themselves. They do not own the checkout — the platform does. Instrument it once and every business on that platform becomes bookable by an agent.
The numbers
| Platform | Domain | Score /100 | Capability manifest |
|---|---|---|---|
| GlossGenius | glossgenius.com | 66 | none |
| Phorest | phorest.com | 64 | none |
| Boulevard | joinblvd.com | 62 | none |
| Booksy | booksy.com | 60 | none |
| Zenoti | zenoti.com | 58 | none |
| Vagaro | vagaro.com | 54 | none |
| Meevo | meevo.com | 52 | none |
| Fresha | fresha.com | 49 | none |
| Mindbody | mindbodyonline.com | 47 | none |
| Timely | gettimely.com | 31 | none |
| Median | 56 | 0 of 10 |
Method
Each site was scanned once on 1 September 2026 with the open scanner in this company's own toolchain: eight unauthenticated GET requests to the homepage and its well-known paths, an identifying User-Agent, no forms submitted and nothing changed. Sixteen weighted checks across six categories — discovery, policy, identity, machine state, rendering and access — are scored to 100 points. Two further checks, covering WebMCP and RFC 8288 Link headers, are reported and deliberately not scored: both are settled or proposed standards that no deployed agent reads for discovery yet, and nobody should be marked down for declining to adopt them. Scoring is earned over applicable: a check that cannot apply is left out of the denominator rather than counted as a failure. The full method and every weight is published at https://colourintl.com/llms.txt.
Corrections
Two figures changed because of a fault in our scanner, not because of anything these companies did. The August run checked /sitemap.xml directly instead of the sitemap URL declared in robots.txt, so Booksy and Vagaro were marked as having no valid sitemap when both served one at the path they declare. Corrected, Booksy's August score was 60 rather than 55 and Vagaro's was 54 rather than 53. One figure moved for a real reason: Boulevard scored 67 in August and 62 now, having since dropped the AI usage policy from its robots.txt.
What this does not measure
These checks measure whether an agent can find, read and understand a site. They do not measure whether it can act on it — call a tool, complete a booking, or pay. A scanner that scores the action surface as well will return a lower number for the same site, and both numbers are correct about different things. No claim is made here about anyone's traffic, revenue, or placement in any AI answer; agent-driven attribution is not measurable today and a figure for it would be a guess.
Check it yourself
Every domain in that table is public and the scanner is free. Scan any of them at colourintl.com — same sixteen checks, no account — and compare. If a number here is wrong we would rather hear it than defend it: hello@colourintl.com.
This page is also available as Markdown: curl -H "Accept: text/markdown" https://colourintl.com/benchmarks/salon-booking-platforms