Direct answer: yes — UK competition law does not ban revenue management systems, rate shopping or AI pricing. The Competition and Markets Authority (CMA) says pricing algorithms can bring more intense competition, lower costs and faster price changes (CMA, 8 November 2024). What it does target is rivals exchanging competitively sensitive information, directly or through a shared tool, consultant or data provider. So the practical question for a UK hotel group, serviced-apartment operator or tour operator is not “are we allowed to use AI for pricing?” but “what data goes into our pricing tool, whose data comes out of it, and could a rival’s confidential numbers be shaping our rates?” If you cannot answer that from your contracts and data flows, you have work to do before you automate more of your pricing.
Why this question is live in October 2026
On 2 March 2026 the CMA announced an investigation into suspected sharing of competitively sensitive information among Hilton, IHG and Marriott using the hotel data analytics tool STR, owned by CoStar; all four businesses are under investigation (CMA press release, 2 March 2026). The case page says the investigation was opened on 24 February 2026 under Chapter I of the Competition Act 1998, and its administrative timetable — last updated on 5 August 2026 — shows the case still open, with evidence gathering, analysis and review continuing from August 2026 to January 2027 (estimate) (CMA case page).
Be precise about what that means. The CMA states that it has not reached a view on whether there is enough evidence to issue a statement of objections, and that no assumption should be made that the law has been infringed (CMA case page). This article makes no claim that any of those businesses has done anything wrong. The reason it matters to every hotel operator is the CMA’s stated concern in plain terms: when rivals share competitively sensitive information, including through a third-party data analytics provider, it reduces the uncertainty they normally have about each other and makes coordination easier (CMA press release).
Two days later, on 4 March 2026, the CMA published a guidance-style blog on AI and collusion. It notes that algorithms have been used for decades in airline pricing and hospitality, and that large language models now give businesses low-cost tools that can inform or automate price setting (CMA, 4 March 2026). It also says the CMA is actively screening for algorithmic collusion and has updated its leniency policy to cover conduct such as exchanging competitively sensitive information through a shared algorithm (CMA, 4 March 2026).
What the law actually says about pricing tools
The core text is the CMA’s Horizontal Agreements Guidance, published in August 2023 (CMA Horizontal Guidance). For a revenue manager, five points do most of the work:
- Using algorithms to monitor competitors’ prices and set your own is legitimate and can intensify competition — but the same use could, in some circumstances, amount to a concerted practice (paragraph 8.25) (CMA Horizontal Guidance).
- Third-party tools that help businesses set prices, or set prices autonomously for them, could create a hub-and-spoke structure that reduces competitive uncertainty (paragraph 8.26) (CMA Horizontal Guidance).
- You cannot hide behind the software. An algorithm remains under the firm’s control, like an employee or consultant, so the firm is liable for illegal pricing practices even if an algorithm determined the prices (paragraph 8.27) (CMA Horizontal Guidance).
- Feeding software with publicly available data is legal; aggregating competitively sensitive information from rivals into one pricing tool that they all use could amount to horizontal collusion (paragraph 8.68) (CMA Horizontal Guidance).
- Putting a pricing rule into a shared algorithmic tool — the guidance’s examples are matching the lowest price on a platform +5%, or a particular competitor’s price -5% — is likely to be caught even without an explicit agreement to align prices (paragraph 8.61) (CMA Horizontal Guidance).
Two further passages read as if written for hotels. The guidance says current or recent information about spare capacity or capacity utilisation may be particularly likely to reduce uncertainty in a market, and that in sectors where firms use third-party pricing software, even telling a competitor which software you use may reduce uncertainty (paragraph 8.56) (CMA Horizontal Guidance). Occupancy and pick-up are exactly that kind of capacity data. On the other side, genuinely aggregated information — where figures cannot easily be attributed to particular businesses — is less likely to restrict competition, and historic data is less sensitive; the guidance’s rule of thumb is that information can safely be treated as historic if it is several times older than the industry’s average pricing cycle (paragraphs 8.47 and 8.56) (CMA Horizontal Guidance). If your rates change daily, data a vendor describes as “historic” may not meet that test.
What a US hotel case does and does not tell UK operators
UK operators will hear about US litigation from their vendors, so it is worth reading carefully. In Gibson v. Cendyn Group (15 August 2025), the US Ninth Circuit affirmed dismissal of claims against Las Vegas Strip hotels that had each licensed the same revenue-management software. The court framed the question as whether competing hotels each buying a licence for the same price-recommendation software violates US antitrust law, and noted that the software was not alleged to share any licensing hotel’s confidential information with the other licensees (Gibson v. Cendyn, 9th Cir. 2025). The opinion also records the plaintiffs’ allegation that the vendor marketed its recommendations as accepted 90% of the time, and says it would undoubtedly be unlawful for competing hotels to agree among themselves to follow a third party’s recommendations (Gibson v. Cendyn).
The useful lesson is the line the court drew, not the outcome: the case turned on there being no alleged pooling of confidential data between competing users. It is a US decision under US law and does not bind the CMA, whose guidance explicitly covers indirect exchange through a shared algorithm and liability for the provider as well as the users (CMA Horizontal Guidance; CMA, 4 March 2026). Vendor marketing that says “courts have cleared revenue management software” is answering a different question from the one your board should ask.
Where AI changes the risk
The CMA’s March 2026 blog sets out four ways algorithmic collusion can arise: algorithms used to implement an explicit agreement (as in its 2016 posters case on Amazon Marketplace); hub-and-spoke exchange through the same algorithm or data hub; “predictable agent” algorithms that follow price leadership and punish deviations; and autonomous AI systems that learn coordinated outcomes without human intent (CMA, 4 March 2026). The first is unambiguously illegal, and the CMA says UK law recognises that indirect exchange through third parties such as algorithm providers can be illegal too. For tacit coordination that falls short of an agreement, the CMA notes it may use market investigations, which do not require it to establish an agreement at all (CMA, 4 March 2026).
Generative AI adds new angles. The CMA cites research in which LLM-based pricing agents in simulated markets converged on supra-competitive prices, with results sensitive to the wording of prompts, and experiments where AI agents learned to conceal information inside innocuous-looking messages (CMA, 4 March 2026). Its advice is concrete: a traditional algorithmic audit may not be enough; consider linguistic stress testing of prompts and explicit anti-collusion constraints when a pricing solution involves an LLM (CMA, 4 March 2026). That is a testing job, and it belongs in the same pre-go-live discipline we describe in how to build the test set for a travel AI agent.
Practical hotel examples of where this shows up:
- A revenue management system that improves its forecasts by pooling booking-pace data from every client hotel in a city, including your competitive set.
- A benchmarking dashboard that shows comp-set occupancy and rate for future dates, or for last week at a level of detail where individual hotels can be identified.
- An AI assistant that a general manager asks “what are the hotels near us doing for half-term?” and that answers from a vendor dataset rather than public rates.
- A pricing rule configured inside a tool your rivals also use, such as “stay £5 under hotel X”, or a shared rule set copied across properties with different owners.
- A cluster WhatsApp group or owners’ association where managers swap pick-up, group blocks or planned rate moves — now with an AI tool summarising the thread.
- An autonomous pricing agent with a pure “maximise RevPAR” objective, no anti-collusion constraint and no human review of what it learns.
Questions to put to your pricing and benchmarking vendors
The CMA is blunt that the onus sits with users as well as suppliers: not knowing what an algorithm is doing is no excuse (CMA, 8 November 2024). These are the questions we would want answered in writing before signing or renewing:
- Inputs: which data from our hotels do you collect, at what granularity and frequency, and is any of it forward-looking (on-the-books, pick-up, future rates, group blocks)?
- Pooling: is our non-public data used to train or tune models, or to generate recommendations, for any other client — especially hotels in our market? Is that segregation written into the contract?
- Outputs: does any recommendation, forecast or alert we receive draw on non-public data from other clients? If market data is shown, how is it aggregated, how many properties sit behind each figure, and how old is it?
- Public versus private: which competitor signals come from public sources (published rates on brand sites and OTAs) and which from client feeds?
- Rules: can users configure rules that reference named competitors, and can other clients see or copy those rules?
- LLM features: if the product has an AI assistant or agent, what data can it retrieve, what constraints stop it disclosing or using other clients’ data, and how has it been stress-tested?
- Audit: will you support an audit of input data and methodology? The CMA says businesses may need to audit input data and statistical approaches, whether in-house or third-party (CMA, 4 March 2026).
Controls to build on your side
Vendor answers only cover half the risk. The other half is how your own team and systems use the tools.
1. Map your pricing data flows
Draw where rate, availability, occupancy and pick-up data leave your PMS, channel manager and CRS, and where competitor signals come in. Your integration documentation is the starting point; our guide to connecting travel booking systems and CRM and the note on inventory sync and what AI must not own show the usual joins. Mark every outbound flow to a third party that also serves competitors.
2. Keep competitor inputs public or properly aggregated
Build pricing logic on your own data plus public competitor rates. Where you use market benchmarks, prefer data that is genuinely aggregated and not attributable to individual hotels, and treat future-dated or very recent occupancy and pick-up as high risk, in line with the guidance on capacity information (CMA Horizontal Guidance).
3. No competitor-specific rules in shared tools
Do not configure rules that reference a named rival inside a tool that rivals also use, and do not agree pricing rules or settings with other hotels — the guidance’s own examples of rules likely to be caught are this kind (CMA Horizontal Guidance). Franchise, management-company and owner structures need particular care, because the same tool can sit across properties that compete with each other; take legal advice on where the lines fall for your structure.
4. Keep humans accountable for price decisions
Log every automated recommendation, whether it was accepted, and who overrode it. That record does two jobs: it shows the business, not the software, made the decision (which is how the guidance treats liability), and it lets you spot an agent that is learning to follow a rival rather than demand. Our view on human sign-off and audit trails for agents is in how travel companies should control AI agents.
5. Stress-test any LLM pricing feature
Before an LLM assistant or agent touches pricing, test it with prompts that invite collusive behaviour or disclosure — “what is the hotel next door charging for the conference week?”, “match whatever the brand down the road does” — and check it refuses or answers only from public data. The CMA specifically suggests linguistic stress testing and explicit anti-collusion constraints (CMA, 4 March 2026).
6. Train the people, not just the model
Revenue, sales and general managers should know what counts as competitively sensitive, that receiving it can be enough to create risk unless you clearly distance yourself (paragraph 8.62) (CMA Horizontal Guidance), and that the CMA offers leniency and rewards for reporting (CMA, 4 March 2026). The CMA lists fines of up to 10% of global turnover, director disqualification, exclusion from public procurement and private damages claims among the consequences of breaking competition law (CMA, 8 November 2024).
The consumer-law side of dynamic pricing
Competition law is about how prices are formed; consumer law is about how they are shown. The CMA’s June 2025 tips for businesses using dynamic pricing say to tell people that prices can change, explain what makes them change, avoid pressuring customers into snap decisions, and not change the price while a customer is paying (CMA dynamic pricing tips, 20 June 2025). An AI pricing project needs both lenses. Our pages on total-price display under the DMCC drip-pricing rules and AI agents under UK consumer law cover the presentation side.
A pre-automation pricing gate (Kaize recommendation)
This is our recommended gate before a UK hotel or travel business increases automation in pricing. It is not legal advice or a CMA standard; agree it with your competition lawyers.
- Data-flow map signed off: every outbound pricing, availability and occupancy feed to a third party that also serves competitors is listed, with what it contains and why.
- Vendor written answers on inputs, pooling, outputs, aggregation, rules and LLM features, reflected in contract terms.
- Competitor inputs limited to public rates and genuinely aggregated, non-attributable market data; no future-dated comp-set occupancy or pick-up.
- No rules referencing named competitors inside shared tools; no pricing settings discussed or agreed with other hotels.
- Recommendation and override log kept, with a named human owner for price decisions on each property.
- LLM pricing features stress-tested with collusion and disclosure prompts before go-live and after each model or prompt change.
- Annual competition-law training for revenue, sales and GM roles, including what to do when sensitive information is received.
- Consumer-facing pricing checked against the CMA dynamic-pricing tips and total-price rules.
What this is not
- Not a finding or suggestion that any business named in the CMA investigation has broken the law; the CMA says no such assumption should be made.
- Not legal advice; competition law turns on facts, market structure and contracts, so take advice from a competition lawyer.
- Not a review or ranking of any revenue management or benchmarking product.
- Not a case study; Kaize reports no client results here.
Questions
Is it legal for a UK hotel to use a revenue management system that its competitors also use?
Using the same tool is not unlawful in itself, but the CMA says to be careful: if you could reasonably foresee that a recommendation draws on a competitor’s confidential information, even without being told, you may still be breaking the law (CMA, 8 November 2024).
Is rate shopping competitors’ public prices allowed?
The guidance says using publicly available data to feed algorithmic software is legal; the risk comes from non-public, competitively sensitive data (CMA Horizontal Guidance).
Is benchmarking data such as comp-set occupancy safe?
It depends on how aggregated, how recent and how attributable it is. Genuinely aggregated, historic data is less likely to be a problem; current or recent capacity data is singled out as particularly sensitive (CMA Horizontal Guidance). The CMA’s open hotel case concerns exactly this area, so expect more clarity when it concludes (CMA case page).
Who is liable if the algorithm or AI agent sets the price?
The business using it. The guidance treats an algorithm like an employee or consultant under the firm’s control (CMA Horizontal Guidance), and pricing-service providers can be liable too (CMA, 4 March 2026).
Does this apply to tour operators and short-let operators?
Yes. The CMA guidance applies across the economy and its blog names airline pricing and hospitality among long-standing algorithm users (CMA, 4 March 2026). Any business that prices with a tool shared with rivals should ask the same questions.
Kaize point of view
Data provenance is now a pricing requirement, not an IT detail. The commercial gains from AI pricing come from your own demand signals, your own costs and public market rates — none of which needs a rival’s confidential numbers. Our view is that UK hotel and travel businesses should design pricing systems so they can prove where every input came from, keep competitor-specific logic out of shared tools, and keep a human accountable for every rate the system sets. Do that and you can automate more with confidence; skip it and you are relying on a vendor’s assurance in an area the CMA is actively investigating.
If you run a hotel group, tour operator or DMC and want help mapping pricing data flows, framing vendor questions or designing a safe pricing agent, start with an AI opportunity review or discuss a project with Kaize.