Direct answer: yes — a UK hotel group, tour operator, DMC or villa company can use AI around guest reviews, but only on one side of a clear line. AI can help you read reviews, spot themes, route complaints, detect suspicious submissions and draft replies a person approves. It must not write reviews for guests, decide which guests get asked for a public review based on how happy they seem, quietly hold back negative reviews, or produce “what guests say” summaries that are more flattering than the reviews behind them. Since 6 April 2025, fake reviews, concealed incentivised reviews and publishing reviews or review-based information in a misleading way have been banned practices under paragraph 13 of Schedule 20 to the Digital Markets, Competition and Consumers Act 2024 (DMCC Act, Schedule 20). The Competition and Markets Authority (CMA) can now decide breaches itself and fine up to 10% of global turnover (CMA, 27 March 2026). The major travel platforms are stricter still: Expedia Group says a review “must not be generated using AI” (Expedia Group content guidelines, August 2026), and Tripadvisor removed more than 214,000 reviews in 2024 it believed contained AI-generated text (Tripadvisor Transparency Report 2025).
Kaize is a UK AI product studio for travel and hospitality. This article is based on the DMCC Act, the CMA’s fake reviews guidance (CMA208), CMA case pages and blog, and the review policies of Google, Tripadvisor, Expedia Group and Booking.com, read on 14 October 2026. It is not legal advice or a case study; controls below are our recommendations.
Why this question is live in October 2026
Reviews are where hospitality AI arrives quietly — reply drafting, survey sentiment scoring, automated review requests, AI summaries — and each feature touches a part of the review chain the DMCC Act now regulates.
Enforcement has moved from warning to investigation. In July 2025 a CMA review of more than 100 businesses’ websites found 54 could be failing to comply, some with no published policy banning fake reviews (CMA online consumer reviews case). On 27 March 2026 it opened investigations into five businesses across how reviews are obtained, moderated and displayed — including whether 1-star reviews moderated by Feefo were left off Autotrader’s platform and ratings, and whether Dignity asked staff to write positive reviews (CMA, 27 March 2026). The CMA has reached no conclusions, and the cases remained open at its 28 September 2026 update (CMA Feefo case page). None is a travel business and we suggest no wrongdoing; the point is that every stage being examined has an AI feature attached to it in today’s hospitality software.
The CMA has also said what comes next. In its April 2026 review of the first year of direct enforcement, it listed fake reviews — “have robust policies in place so reviews give consumers an honest picture – and make sure your employees are implementing them” — and the way businesses use AI, including agents, among its year-two compliance priorities (CMA blog, 17 April 2026).
What the law actually bans
Paragraph 13 is short. It bans submitting, or commissioning someone to submit or write, a fake consumer review or a review that conceals it was incentivised; publishing consumer reviews or consumer review information in a misleading way; publishing reviews without taking “reasonable and proportionate steps” to prevent and remove fake reviews, concealed incentivised reviews and false or misleading review information; and offering services to traders that do or facilitate any of that (DMCC Act, Schedule 20). Four definitions do most of the work for a travel business (DMCC Act, Schedule 20; CMA208):
- A consumer review is a review of a product, a trader or anything relevant to a transactional decision. The CMA says it does not have to be written by a consumer, and includes star ratings and thumbs-up icons as well as text (CMA208, paragraphs 2.2 and 2.4).
- A fake review is one that purports to be, but is not, based on a person’s genuine experience — positive or negative (paragraph 2.5).
- A review conceals an incentive if someone was commissioned to write it and that is not made apparent. Commissioning includes discounts, vouchers, free stays and asking a member of staff to write one (paragraphs 2.9 and 2.10).
- Consumer review information is anything derived from or influenced by reviews — overall ratings, review counts, rankings and “overall summaries” (paragraph 2.11).
That last definition is the one most AI projects miss. An AI-written paragraph on your website that says “guests love our spa and rave about breakfast” is consumer review information. So is a chatbot answer to “what do people say about this villa?”. If it misrepresents the reviews behind it, it can fall within the ban even if every individual review is genuine.
Five places AI meets reviews in a hotel or tour operator
1. AI that writes or “polishes” the review for the guest
Some post-stay tools turn a guest’s star rating or a few words into a finished review. The CMA says encouraging reviews is fine “without predetermining the contents or sentiment expressed in the review” (CMA208, paragraph 3.6) (CMA208); AI-drafted text supplied by the business is predetermination by design. Platforms are starker: Expedia Group says a review “must not be generated using AI” and partners “must not write reviews on behalf of travellers” (Expedia Group content guidelines); Tripadvisor said in its 2025 report that, for the time being, it will not allow reviews identified as AI-generated (Tripadvisor Transparency Report 2025); Google tells merchants not to “request that specific content be included” (Google Maps user-generated content policy). A writing assistant in your review email gets genuine reviews removed — or manufactures ones that are not really the guest’s.
2. Sentiment routing: asking only the happy guests
The most common AI pattern in review tooling is also the riskiest: a survey scores sentiment, happy guests get a Google or Tripadvisor link, unhappy guests get a private complaints form. The CMA lists “encouraging just those who are satisfied to leave reviews” as cherry-picking (CMA208, paragraph 4.5) and says publishers should not steer consumers to a complaint instead of a review (paragraph 8.11) (CMA208). Google bans merchants from selectively soliciting positive reviews (Google Maps user-generated content policy). If an AI score decides who is invited to review publicly, switch that logic off: invite every eligible guest the same way and handle complaints in parallel, not instead.
3. AI moderation of reviews on your own website or app
If you publish reviews or testimonials on your own site, you are a publisher with a positive duty: a published, plain-English policy banning fake reviews and setting out your approach to incentives, an evidence-led risk assessment, and processes for detection, investigation and response (CMA208, paragraphs 8.5 to 8.27) (CMA208). Automated detection is encouraged — the guidance cites reviews from different reviewers sharing one email or IP address, or a sudden spike of highly positive or negative reviews (paragraph 8.19). But removing a genuine negative review “may mislead consumers as much as publishing a fake positive review”, and checks must apply to positive and negative reviews alike (paragraph 8.16). An AI moderator tuned to “protect the brand” will drift towards suppressing complaints; tune and test it on “is this a genuine experience?”, and keep a person on every removal of a low-rated review. The open investigation into both Autotrader and its moderation provider Feefo over the same alleged conduct (no finding made) suggests outsourcing moderation does not outsource the duty (CMA, 27 March 2026).
4. AI summaries and chatbot answers about reviews
AI review summaries now sit on booking pages, OTA listings and guest assistants. The CMA names them: when a banned review is removed, publishers must correct “review counts, overall ratings, rankings and AI generated review summaries” that relied on it (CMA208, paragraph 8.24) (CMA208). Old reviews may need different treatment after a material change such as a refurbishment or new operator (paragraph 4.10), and merging reviews of materially different products is “catalogue abuse” (paragraphs 4.8 and 4.9) — one summary across two hotels, or across premium and entry-level versions of a tour, is the travel equivalent. Build summaries from all eligible reviews for one product, show the count and date range, and regenerate when reviews are removed.
5. AI-drafted replies to reviews
Replies are the lowest-risk use of AI here, with two traps. The CMA’s examples of commissioning a fake review include contacting a customer who left a negative review and offering a refund or gift card if they change it (CMA208, Chapter 3 examples) (CMA208); Expedia Group bans asking a guest to revise a review to receive compensation (Expedia Group content guidelines). An AI that “helpfully” offers a goodwill voucher in a public reply to a 2-star review is heading for that line. And platforms moderate replies: Booking.com reviews them before publication (up to 72 hours) and asks for original, specific responses that are “true and accurate” (Booking.com Partner Hub); Expedia Group may remove responses containing personal data (Expedia Group content guidelines). Draft with AI if it helps; approve with a person; never let the model offer compensation, repeat booking details or argue facts it cannot check.
Where platform rules are stricter than the law
The DMCC Act allows incentivised reviews if the incentive is made apparent and the review is still genuine (CMA208, paragraph 3.7) (CMA208). The main travel platforms go further, and their rules decide whether your reviews survive:
- Google: no incentives in exchange for posting a review or for revising or removing a negative one; no discouraging negative reviews; no selective solicitation; no pressure to review on the premises; no requests for specific content, including naming a staff member (Google Maps user-generated content policy).
- Booking.com: property partners “can't post on behalf of guests or offer incentives in exchange for reviews” (Booking.com review guidelines).
- Expedia Group: no AI-generated reviews, no partner-written reviews, no partner incentives, and a ban on AI-generated or AI-edited content used to misrepresent a property (Expedia Group content guidelines).
- Tripadvisor: its 2025 report says identified AI-generated reviews are not allowed for the time being; its 2025 report attributes 54% of fraudulent review submissions in 2024 to “review boosting” by owners, staff or affiliates, and links 360,000 removed reviews to employee incentive programmes (Tripadvisor, 18 March 2025).
Kaize reading: design to the strictest platform you depend on. A review journey that is lawful but gets your Google or Tripadvisor reviews stripped is still a commercial failure.
Staff, vendors and synthetic testimonials
Two older problems get easier with AI. Staff review schemes — asking or rewarding employees to post reviews — are commissioning under CMA208 (paragraph 2.10) and banned by Google (CMA208; Google Maps user-generated content policy). AI-written “guest quotes” on a landing page are fake reviews if they purport to reflect a real guest’s experience, because a consumer review need not be written by a consumer to count (CMA208, paragraphs 2.4 and 2.5).
Vendors are in scope too. Offering traders services that submit fake reviews or help them evade platform detection is itself banned; the CMA’s examples include selling bot-generated reviews and an SEO service that improves ratings by generating fake ones (CMA208, Chapters 6 and 7) (CMA208). If you syndicate a third party’s rating onto your own site, check that provider’s policies on reasonable grounds before relying on them (paragraph 8.32(b)).
A worked example: a 40-room coastal hotel
A hotel’s new guest-experience platform surveys guests at checkout; 9s and 10s get a Google link with a suggested review drafted from their answers, lower scores get a private feedback form; the website shows an AI “what guests love” summary; an assistant drafts every reply. Three of the four features need changing before go-live: the sentiment gate, the drafted review text, and the summary if it is built only from reviews the model liked. The reply assistant can stay with human approval and a block on compensation offers. The fixed journey invites every eligible guest the same way, lets guests write their own words, labels the summary with its review count and date range, and alerts the duty manager to low scores as well as — not instead of — inviting a review.
A pre-go-live gate for AI in your review journey (Kaize recommendation)
This is our recommended gate before any AI feature touches guest reviews. It is not a CMA standard or legal advice; agree it with your legal and compliance owners.
- A published, plain-English reviews policy, linked from every page that shows reviews, that bans fake reviews and states your approach to incentives (CMA208, paragraphs 8.9 and 8.10).
- Review invitations sent to every eligible guest on the same rules; no AI sentiment score or survey result decides who is asked for a public review.
- No AI-generated or pre-filled review text offered to guests; the guest writes their own words.
- No incentives tied to review content or rating; any incentive that remains is disclosed and permitted by every platform it touches.
- AI moderation tested on genuine positive and negative reviews before go-live; a named person approves every removal or withholding of a low-rated review; reasons logged.
- AI summaries and chatbot answers built from all eligible reviews, showing count and date range, scoped to one property or product, and regenerated when reviews are removed or the product materially changes.
- AI reply drafts approved by a person; the model cannot offer refunds, vouchers or compensation in a reply, and cannot include guest names or booking details.
- No staff review schemes and no AI-generated testimonials; vendor contracts that allocate DMCC compliance and give you audit access to moderation logs.
- A regression test set rerun on every review-related prompt or model change, including the cases where the right answer is no.
What this is not
- Not a finding or suggestion that any business named in a CMA investigation has broken the law.
- Not legal advice, and not a review or ranking of any reputation-management product.
- Not a claim that AI is banned in review operations — the CMA itself encourages automated fake-review detection.
- Not a case study; Kaize reports no client results here.
Questions
A guest used ChatGPT to tidy up their own review. Is that a fake review?
Under UK law, a review is fake if it does not reflect a genuine experience, so a guest polishing their own words is not automatically fake. Platforms may still remove it: Tripadvisor’s 2025 report says it does not allow reviews identified as AI-generated, and Expedia Group requires that reviews are not generated using AI (Tripadvisor Transparency Report 2025; Expedia Group content guidelines). The risk changes if your business supplies the AI tool or the text.
Can we show an AI summary of our reviews on our website?
Yes, if it fairly reflects the reviews it is based on. It is consumer review information, so it must not be misleading, and it must be corrected when reviews behind it are removed (CMA208, paragraphs 2.11 and 8.24) (CMA208).
Kaize point of view
Use AI to read reviews, not to write, choose or hide them. The value of AI in reputation work is real — theming thousands of comments, routing a broken shower to housekeeping before checkout, catching suspicious review spikes, drafting a careful reply in minutes — and none of it needs the model to touch what a guest says or who gets asked to say it. Our view is that UK hotels, tour operators and DMCs should run the review journey as a regulated system: one invitation rule for every guest, guests’ own words, human approval for removals and replies, and summaries traceable to the reviews behind them. Do that and AI makes reputation work faster and more honest; skip it and today’s time-saving features become tomorrow’s evidence.
Related reading: AI agents and UK consumer law, DMCC drip-pricing controls, Booking.com Messaging API controls, AI disclosure in guest chat, ABTA complaint handling and AI and building a travel AI agent test set.
If you run a hotel group, tour operator or DMC and want help designing a review journey, moderation rules or a reply assistant that stays on the right side of the DMCC Act and platform policies, start with an AI opportunity review or discuss a project with Kaize. For how review tools join your PMS, booking system and CRM, see connecting travel booking systems and CRM and how travel companies should control AI agents.