Agent Booking Index
Which hotel booking engines can AI agents actually use? We test whether an agent can get from an independent hotel's homepage to a priced room on its own website, and compare the booking engines underneath.
Pilot round: 20 October 2026, 28 independent UK hotels.
A scripted browser that reads pages as text tried to reach a priced room on each hotel's own website and stopped before any booking. It is not ChatGPT agent or another named product, and samples per engine are small, so treat these results as directional. A re-run with consumer agents is next. Method and limits.
Pilot results by booking engine
| Booking engine | Hotels tested | Passed | Partial or failed | Average /12 |
|---|---|---|---|---|
| Guestline | 7 | 7 | 0 | 10.7 |
| SynXis | 5 | 5 | 0 | 10.2 |
| SiteMinder | 2 | 2 | 0 | 10.0 |
| D-EDGE | 2 | 0 | 2 | 8.0 |
| Mews | 3 | 1 | 2 | 7.3 |
Also in the pilot but not shown: engines with a single hotel each, and three hotels with no online booking engine at all (enquiry forms only), which all failed.
What the full index will measure
- Completion
- Did the agent reach a priced room and the payment page? Pass, partial or fail.
- Speed
- Time and number of steps from the hotel's home page to a priced room.
- Understanding
- Did the agent report the right total price, taxes, fees and cancellation policy?
- Blockers
- Bot walls, CAPTCHAs, widgets the agent can't operate, logins and pop-ups.
Next to be tested
- Beds24
- Bookassist
- Cloudbeds
- eviivo
- Freetobook
- Guesty
- Hostaway
- Little Hotelier
- Lodgify
- OwnerRez
- Profitroom
- ResNexus
- RMS Cloud
- ThinkReservations
- WebRezPro
Is your engine missing, or do you run one? Email [email protected].