Your reviews used to be a reputation question: what past guests said, and how the page looked when a prospective guest scrolled it. That has changed. When a traveler asks ChatGPT, Gemini, or Perplexity for a hotel, the engine assembles a shortlist before your website is ever seen, and review content is part of the raw material it reads to build that shortlist. A property with a thin, stale, or vague review record is not argued against. It is simply left off the list, and the room nights it would have taken go to the comp set instead.

AI can read reviews before a traveler reaches your website.

AI travel answers are not built from your brochure copy. They are assembled from what the wider web says about you, and reviews carry a large share of that record. Cloudbeds reports that over 40 percent of brand impressions in engines like ChatGPT and Perplexity come from third-party citations, a pool that includes forums, blogs, news, and expert reviews. The engine is quoting other people's account of your hotel, not yours.

The review platforms themselves are feeding the shift. Tripadvisor's engineering team explains that its AI review summaries are generated from recent traveler reviews and that it only creates summaries for properties with enough recent reviews. Read that condition the way an owner should. A hotel without a steady flow of fresh reviews does not get a worse summary. It gets no summary. Volume and recency are already gatekeeping which properties receive the richer AI treatment on one of the largest review platforms in travel.

The same summaries are built attribute by attribute: cleanliness, value, location, drawn from what guests actually wrote. A review that names the quiet rooms or the breakfast gives the machine something to work with. A three-word "lovely stay, thanks" gives it nothing. Specifics in reviews have become machine-readable inventory descriptions, written by your guests, that you never see in a content plan.

What one controlled audit found.

In June 2026, researchers published an algorithm audit of exactly this question: which reputation signals move a language model's hotel choice. The study on arXiv ran randomised five-hotel choice sets through twelve language models, 3,024 trials per model and more than 60,000 model calls in total, with rating, review volume, recency, management response, price, and other attributes varied independently.

The results are worth an owner's attention. In the audit, a top guest rating raised a hotel's chance of being selected by 31.7 percentage points, the largest positive effect of any signal tested. High review volume added 8.3 points. Recent reviews added a further 1.6. A high price cut selection by 30 points. The two heaviest levers in the machines' choices were the guest rating and the rate, which is to say the two numbers an owner already watches, now being weighed by software the owner has never met.

One honest caveat from the same audit: management responses to reviews had no detectable effect on the models' choices. The machines are mostly reading what your guests wrote, not what you wrote back. Responses still matter, for reasons covered below, but nobody should sell you review replies as a direct lever on AI shortlists.

A fair reading of the study also notes its limits. It used synthetic hotels in a controlled test, not live booking traffic, so the percentage points describe model behavior under experiment conditions, not a revenue forecast for any real property. What it establishes is direction: when these systems choose between comparable hotels, review-derived signals do a large share of the deciding.

Reviews now influence visibility as well as reputation.

Put those pieces together and the frame changes. Reputation is about how you look once a guest is already considering you. Distribution is about whether you are put in front of the guest at all. Review volume, recency, sentiment, and specifics now do both jobs, and the second one is new.

A review is no longer only what a past guest said about you. It is the data an engine reads to decide whether the next guest hears about you at all.

The cost of losing this channel is invisible in your reports. A hotel that never makes the AI shortlist is not rejected. It is never considered, and the query that would have become a direct booking resolves quietly at a competitor. That is revenue left behind with no line item, because the inquiry never reached your booking engine to be counted.

Your target guests are comparing against your comp set, and now the engines are too, on the same review signals, continuously. If a neighboring property runs a steadier flow of recent, specific reviews, it occupies shortlist positions your property could have held, at whatever rate its positioning supports.

What an owner should do differently.

None of this requires new software or a new department. It requires treating the review record as a sales channel with a pipeline to keep full.

  • Keep the review flow steady. Volume and recency are the two signals a property most directly influences. Tripadvisor withholds its AI summaries from hotels without enough recent reviews, and the audit measured high volume at 8.3 points of selection effect. A consistent, honest ask at checkout and post-stay is now channel maintenance, not vanity.
  • Respond, for the layer AI reads from. Google's own guidance states that more reviews and positive ratings can help a business's local ranking, and that helpful replies can help a business stand out. The AI engines lean heavily on that local layer when they answer travel questions. Replies also shape what the next guest writes, and it is the guests' words the machines weigh.
  • Keep the facts current everywhere. Google states plainly that businesses with complete and accurate information are more likely to show up in local search results. An engine reading listings that disagree about your pool hours, room count, or address will describe with confidence the competitor whose record is consistent. Listings that disagree undermine credibility with machines exactly as they do with guests.
  • Watch the comp set's review record, not just their rate. Most owners check comp set rates weekly. Review volume, recency, and rating deserve a place in the same habit, because they now influence who enters the shortlist where rate is compared at all.

How HTB supports the work.

HTB treats the review record as part of your distribution, because that is what it has become. We identify where your review volume, recency, and rating sit against your comp set, and where listings disagree about the basic facts of your property. We monitor those signals over time, on the platforms the engines read. We report what we find in plain terms, as dated observations of your property, not generalities. We recommend the actions that sit with you, including how the review ask fits your front desk and post-stay flow.

Where the work is website content, we do it directly: keeping your site's facts current, consistent with your listings, and structured so that machines can read what your guests already know. The rating stays yours to earn and the rate stays yours to set. Our job is to make sure that when an engine reads the record on your hotel, the record is full, current, and consistent. The shortlist stays the engine's decision, as the rate stays yours.