AI booking agents could quietly shave hotel room rates, industry experts warn

Agents that act, not just answer

A new class of AI assistants does more than answer questions. Tools such as Instinct and Meta Muse can write and send emails, shop and book travel on a user's behalf without step-by-step instruction. As Hospitality Investor reported, a Financial Times editorial pictured such an assistant spending the night haggling over bills and cancelling subscriptions before free trials turn into paid ones.

For hotels, the question is what happens when that kind of persistent bargain hunter points itself at room rates. So far, agents have tended to book through an online travel agency (OTA), a site that lists many hotels and takes a commission on each stay. The ability to re-check prices after booking and switch when they drop is only starting to appear.

Why constant re-shopping hurts

People already do this by hand. An Ipsos survey of British consumers found that a quarter had gone back and re-shopped a hotel booking, and half of the remaining respondents said they would in future. Louise Tawadrous of Ipsos said those who held back cited practical obstacles, such as time, hassle and fear of something going wrong, rather than emotional ones.

An agent removes those obstacles. It does not tire, and it can watch prices continuously in the background. That is good for the guest and much less good for the hotel.

Patrick Angwin, founder of the consultancy Clanalytix, expects the main effect to be lower rates and, through commissions, thinner margins. His example: a £200 booking made directly with the hotel becomes £164 once the OTA takes its cut, and £147.60 after the agent's re-shopping pushes the price down further.

He adds that the damage is hard to see. Commission creep is buried in a wider cost line, and a stay that was cancelled and rebooked usually just looks like a somewhat cheaper rate in standard reports.

Defences, and why blocking is hard

Angwin suggests partial protections. Hotels can offer sensibly priced non-refundable rates, shorten the window in which guests can cancel, and ease off on discounting as arrival nears. All three make it harder for an agent to profit from waiting and swapping.

A harsher option is to turn agents away entirely. Amazon and Delta Air Lines have already done so. But Angwin argues that scale matters: Amazon can refuse agents and still keep most sales, whereas a single hotel that blocks an agent is unlikely to convince that guest to book direct. The agent would probably go through an OTA, with commission attached, or the booking could be lost altogether.

Pulling agents toward direct bookings

Angwin does not think the outcome is all negative. He expects companies to develop ways to limit price chipping and to win more direct bookings through agents.

Some are already trying. Cloudbeds, a provider of property management systems (the core software hotels use to run reservations), launched a direct integration with Meta Muse in October 2026. Its CEO, Adam Harris, said that when an agent finds a Cloudbeds property, the booking lands in the hotel's own system as a direct booking, and the guest relationship stays with the hotel.

First, will the AI mention you at all?

Before any booking, an agent has to know a hotel exists. A study by Stiplo looked at general AI tools such as ChatGPT and 659 accommodation providers in Barcelona. It found 143 of them, or 22 percent, were invisible to AI. Independent hotels were eight times more likely than chain hotels to be missing. Even when a hotel was recommended, a direct booking link was rarely offered, and the AI pointed to OTAs more often than to hotel websites.

An industry is forming around this problem, called GEO, or Generative Engine Optimisation: the practice of making a business easy for AI systems to find and recommend. Stiplo, founded by former Ennismore chief digital officer Carlo Del Mistro, is one of several firms offering audits. The others named include UP Hotel Agency, INNsight, Arise Hotel Marketing and Publicus.

No one has fully cracked it. The big AI platforms do not publish how they rank results, so nobody can say exactly how content, reviews and structured data are weighed. Angwin put it bluntly: "It's still the wild west out there." Much of what is sold as GEO is, in his view, simply solid digital marketing: accurate, well-organized information that AI can trust.

How fast this could grow

Agents account for only a tiny share of hotel bookings today. A Mastercard report expects one in ten people to routinely use AI agents to shop and pay by 2030. Angwin says adoption will rise as people grow confident in the security and control they have over card and personal details.

His advice to owners for now is practical. Check how your hotel is described on different AI platforms, and measure what each sales channel really leaves you after costs, using net rate and contribution by channel rather than headline revenue. Otherwise the cost of agent bookings stays hidden.

These views came from a panel at the Annual Hospitality Conference 2026 in Manchester. The share of bookings involved is small, so much of this is forecast rather than measured damage.