Travelers face a daunting task during the research phase of trip planning: sifting through hundreds or even thousands of reviews to separate the selling points from the dealbreakers.
But artificial intelligence (AI) offers an opportunity to cut through that volume and surface the insights travelers actually need. At the top of many review sections and within large language models (LLMs), travelers now see AI-generated review summaries, which—in theory—simplify the decision-making process by aggregating
common traveler sentiments.
Phocuswright data found that these summaries are indeed directly tied to action. When asked what most increases their likelihood to act when presented with an AI trip recommendation, 33% of U.S. travelers said review summaries that highlight pros and cons.
Travel platforms use various strategies to structure these summaries but generally all aim to keep them objective.
A listing on GetYourGuide for an Illuminated Evening Guided Kayaking Tour in Split, Croatia, for example, highlights supportive guides and on-the-water photos but concedes that some reviewers find the
tours long and the late timing tiring.
An AI-generated Google review summary of the Waldorf Astoria New York states that reviewers appreciate the “stunning, newly renovated interiors” and spacious rooms. But while many praise the personalized
service from the staff, some also say “service can be inconsistent.”
It's a seemingly simple formula. However, crafting an AI-generated overview that adequately reflects travelers' high and low points, includes recent information and acknowledges genuine
concerns can be tricky.
Tripadvisor recently came under fire for allegedly downplaying traveler concerns within its AI-generated review summaries.
An investigation conducted by U.K. consumer
advocacy organization Which? suggested that for some hotels, Tripadvisor’s summaries were downplaying traveler concerns about food poisoning, hygiene, cleanliness and sexual harassment.
The organization said its findings suggested Tripadvisor’s
technology is capable of identifying negative comments and concerns but then questioned why they weren't consistently included in summaries and why language appeared to “minimize the significance” when they were.
A spokesperson for Tripadvisor
said the company “fundamentally disagrees” with Which?’s characterization of its AI review summaries, stating that the platform is “built on transparency” and that AI features are used to help travelers navigate information.
“Our AI features ... use large language models and natural language processing to analyze both positive and negative traveler reviews from the past 12 months and provide a high-level snapshot of themes from these reviews. Supporting traveler quotes are
included for transparency, and every summary links directly to the underlying content,” the spokesperson told PhocusWire.
“Nothing is hidden or removed by these features ... We also have built-in safeguards, including suppressing AI review summaries on listings with reports of serious safety incidents.”
But the situation with Tripadvisor poses a larger question for the traveler industry: How are companies ensuring AI-generated summaries accurately reflect varying review sentiments?
Strike a balance
Across the board, travel companies said they want to feature both positive and negative feedback.
GetYourGuide’s summaries, for example, are “structured to always include the most common piece of constructive feedback
alongside the positives, rather than only the favorable reviews,” according to director of product Arjun Muralidharan.
And Google told PhocusWire it won’t generate a summary on Maps if it can’t identify a consistent theme across reviews, whether
that be positive or negative.
Booking.com, on the other hand, focuses on creating AI-generated summaries for properties with 15 or more text reviews from
the past three years, surfacing both positive and negative themes, said David Adamczyk, VP of product and accommodations.
Personalization is another factor.
The same property can mean entirely different things to a solo business traveler versus a family or a couple planning a trip.
David Adamczyk, Booking.com
Juanjo Rodriguez, chief AI officer at market intelligence platform Lighthouse, said the company serves different hotel review summaries to different travelers.
“The
moment we know something about you, for example, let's say that we already know that you are a couple looking for a hotel because you said so ... then the reviews that you will see would be a summary of reviews but specifically from people who were
[traveling as] couples before.”
Adamczyk also noted that Booking.com summaries are “built to surface what is most accurate and relevant to the specific traveler searching, rather than just a generic snapshot.”
“The same property
can mean entirely different things to a solo business traveler versus a family or a couple planning a trip,” he said.
Consider the source
It’s also crucial for travel platforms to ensure AI has legitimate reviews to pull from and summarize.
Booking.com only allows travelers who have stayed at a property to leave reviews and does not accept reviews from anyone who didn’t complete
a stay booked through its platform.
“Every single review is strictly verified against a real booking,” Adamczyk said. “This verified-only standard ensures that our AI overviews retrieve data solely from authentic, first-hand experiences.”
GetYourGuide,
too, said it only allows reviews from verified customers.
Google said it also has systems in place to detect content such as fake reviews from people who haven’t visited a place, with those reviews removed and not used in review summaries.
Users
can flag reviews across platforms too. If Booking.com sees an uptick in “not helpful” feedback, humans adjust the algorithm for the summaries. People and businesses on Google can use the “Report this summary” button to prompt an analyst to review.
It’s worth
noting that Tripadvisor has historically battled fake reviews on its platform, where users aren't currently required to provide proof of
stay or a receipt before submitting a review. In 2024, it rejected or removed more than 2.7 million fraudulent reviews, up from 2 million in 2023, according to its 2025 Transparency Report. It also flagged and removed 214,000 AI-generated reviews. Moderators reviewed approximately 4.2 million reviews before or after they were posted, the report stated.
Trust and differentiation
Review summaries are far from perfect, and trust in AI lags.
Looking ahead, it'll be tough to establish and hold travel companies to an enforceable standard for AI-generated summaries.
“I’m not even sure we are at a point of an expected standard of transparency and neutrality in general search results. How many times have you clicked on a link and it wasn’t what you were hoping for because the result heading/title was misleading?” said Alicia Schmid, director of research for Phocuswright.
“The other challenge is, how do you police it? I believe there will be best practices that will become the norm rather than the exception.”
For now, summaries are typically labeled as AI generated and often include a note to help travelers understand
how they’re created. And while travel companies say summaries are intended only to give travelers a broader understanding rather than replace actual reviews, their abbreviated length might actually be a hinderance.
People realize that you could say the exact same thing about the hotel next door, so, where’s the value added?
Juanjo Rodriguez, Lighthouse
“A review summary, by definition, is pretty short. It's impossible to be nuanced. You don't have the space,” Rodriguez said.
“It’s a major problem with the summaries, that they all look very similar. If you look at them, it could be any of them, [they’re] interchangeable almost,” he said, adding that he believes the industry will
evolve.
“People realize that you could say the exact same thing about the hotel next door, so, where’s the value added? That’s why you need, in a way, to preserve some of the nuance or the quirks from specific reviews.”
And when LLMs generate summaries, they’re looking for commonality as opposed to differences, Rodriguez said.
“That's kind of the opposite of what you want as a guest. You want to understand how things are different. And now reviews are saying
what's common.”
While Rodriguez was specifically discussing hotels, it points to a broader branding issue: Guests want to know what sets an experience or a property apart from others, and there probably isn’t enough information in a short summary to
capture that.
“By definition, summarizing, you lose what makes a property different. And the whole story about any brand—but clearly hospitality brands—is, 'I want my brand to be different,'” he said.
“The moment you make a summary
of my brand, I lost again, I lost already. I don't want to be summarized. If I'm Ritz-Carlton or I'm Shangri-La, the last thing I want is for you to make a summary of my brand.”
Phocuswright's Travel Marketing AI Summit London
Join Phocuswright in London on Sept. 8 to learn more about the intersection of travel, marketing and AI.