Hotel websites are just 8% of the sources AI engines cite in Milestone's benchmark
Milestone Inc.'s Hospitality AI Visibility Benchmark 2026 analyzed 1.38 million AI answers for 1,600 hospitality locations. Hotels' own websites supplied just 8% of cited sources, while booking and review sites supplied 32%. Each engine mixes them differently, which one combined score can hide.
This story was produced through MarketScale. See how Hospitality teams put it to work with Executive Thought Leadership.
Key takeaways
A hotel's own website can show up in nearly half of its AI answers and still be just 8% of the sources the engines lean on.
Discovery is where hotels lose ground: when travelers don't name the property, median visibility falls to 55%, and location questions have visibility ranging from about 15% to 80%.
Google AI Overviews and Gemini reward different work: AI Overviews cited hotels' own sites most (15%) and booking sites least (7%), while Gemini cited booking sites most (24%).
Free workspace
Turn your Hospitality expertise into content.
Record interviews, organize footage, and write with AI on a free trial of the MarketScale platform for qualifying companies. No demo required, no credit card.
For the median hotel in Milestone Inc.'s new benchmark, the hotel's own website was cited in 45% of AI answers about it. Across every source those engines cited for hotels, owned websites made up 8%.
Those two numbers sit side by side in the Hospitality AI Visibility Benchmark 2026, which Milestone released with the AI Hospitality Alliance (AIHA) and HEDNA, with Aven Hospitality as co-sponsor. They measure different things, and the gap between them is the most useful finding in the report. A hotel's own website can show up in nearly half of its AI answers and still be just 8% of the sources the engines lean on.
The arithmetic is simple once it's spelled out. One AI answer can cite several sources at once: a booking listing, a review page, a travel magazine roundup, a nearby attraction's site, and the hotel's own page. The hotel can be present in the answer while being one voice among five. For the director of e-commerce who has spent years tuning brand.com, that means the owned site is necessary and nowhere near sufficient.
The report's definition explains how both figures can hold at the same time. In the benchmark, AI Visibility means the share of answers that either name a property or cite a website the property owns, so a hotel still counts as visible when the citation came from a booking listing. The answer format makes this matter more. According to the Milestone release, AI engines typically reply to a travel query with a brief list of hotels rather than a page of links, so the sources behind each answer carry more weight for hotels. A short list has no page two.
The dataset is large. Milestone analyzed 1,381,203 live answers from ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews, recorded between July 1 and September 30, 2026, for 1,600 of its own hospitality customer locations. All answers came from routine monitoring; none were simulated. Milestone, a founding partner of AIHA, led the research and supplied the customer data.
Where hotels drop out of the answer
Ask an engine about a hotel by name and it names the hotel in essentially every answer. Ask an open question, the report's example being the best pet-friendly hotel in Asheville, and the median property appeared in 55% of answers across all 1,600 locations. The summary page published by the AI Hospitality Alliance, which includes fuller methodology than the release, separates being known from being discovered, and the second is where the work is.
Location questions were the hardest category. Prompts like hotels near the Blue Ridge Parkway produced the lowest median visibility, with individual properties ranging from about 15% to 80%. A spread that wide says the category median tells a single property very little. For a resort that sells on proximity to a park, stadium or convention center, its own score on those exact prompts is the number that matters.
How often AI engines included properties, Q3 2026
Milestone Inc., Hospitality AI Visibility Benchmark 2026
That last figure frames the competition plainly. In roughly half of the discovery answers for the median hotel, an online travel agency is in the room. Booking sites accounted for 21% of hotel citations and review sites for 11%, together 32%, four times the owned-site share by the report's own numbers. Milestone notes that higher visibility when OTA listings are cited is an association, not proof that the listings caused it.
Hospitality Technology's editors offer one reason the intermediaries do so well. In their view, OTAs are already strong at supplying AI models with inventory and sentiment data that is structured and machine-readable. If that reading holds, the 32% reflects a data-format advantage as much as a popularity one, and a hotel's response would start with how cleanly its own facts are published.
Five engines, five source mixes
Milestone calls each engine's pattern of sources its citation fingerprint, and the fingerprints differ enough to change the to-do list. ChatGPT leaned most on other local businesses' websites, such as partners, venues and attractions, at 18% of its hotel citations, and it cited hotels' own sites more than any other chat engine, at 11%. Gemini cited booking sites more than any engine, at 24%, plus travel media at 16%. Perplexity favored booking sites (21%) and review sites (16%). Claude had the broadest mix: booking sites 21%, travel media 17%, review sites 16%.
Google AI Overviews is the outlier. It sent 15% of its hotel citations to the hotel's own website, the highest of any engine, and only 7% to booking sites, the lowest. By the report's figures, that puts the owned-site share at roughly twice the booking share on that one engine.
On Gemini, Perplexity and Claude, booking sites led all sources in citations, and Claude and Perplexity relied on review sites more heavily than the other engines did. Benu Aggarwal, Milestone's founder and president, summed up the result plainly, saying that every engine pulls from its own mix of sources, so a plan that succeeds on one engine may not succeed on the next. For a revenue team, that could mean breaking a single AI plan into separate tracks: listings and reviews for the chat engines that depend on intermediaries, owned pages for AI Overviews, and local partner sites for ChatGPT.
During the quarter, the engines also went in opposite directions. Using matched prompts and locations, Perplexity's hotel visibility climbed from 65% in July to 70% in September, while Google AI Overviews slipped from 57% to 52%. In AI Overviews answers, the average number of hotels named fell from 2.65 to 1.89. With fewer names in each answer, hotels have fewer spots to win, and that shrinkage is happening on the engine that gives a hotel's own pages the most credit.
Across all engines combined, the change stayed below the study's two-point threshold. That is the report's case for tracking engines separately in a single sentence: the blended number barely moved while two of its parts swung five points apart. A combined score would have shown a quiet quarter.
A hotel's own website can show up in nearly half of its AI answers and still be just 8% of the sources the engines lean on.
Other scorecards, other units
Milestone's benchmark arrives in a year already crowded with AI visibility rankings, and they don't measure the same thing. eMarketer's AI Visibility Index for travel and hospitality, published October 28 by analyst Kelsey Voss, tracks brand mentions on ChatGPT and Google Gemini across seven categories; Marriott International led its September leaderboard with a 26% mention rate, ahead of Hilton Worldwide at 20%.
In May, Hotel Management's Esther Hertzfeld reported on 5W's Airlines and Hotels AI Visibility Index, which found that in several subcategories the top three brands took more than 70% of citation share, and that earned media, more than paid media or loyalty program size, tracked with how often brands were cited. The same index found luxury structurally exposed: premium hotel brands cited weaker than expected in general travel prompts, a gap 5W tied to limited third-party editorial coverage for the engines to retrieve.
Those are brand-level views. Milestone's is property-level, and on brand affiliation it found no reliable edge: brand-affiliated hotels appeared in a median 47% of unbranded answers and independents in 55%, with overlapping ranges. The two findings can both hold. A parent brand can dominate mentions in a category question while an individual flagged hotel still has to earn its place in a question about a specific town or landmark. The 5W emphasis on third-party editorial coverage does line up with Milestone's finding that Gemini and Claude draw heavily on travel media.
Milestone's sample is its own customer base: 1,128 hotels and resorts plus 472 RV parks, campgrounds and communities. RV properties are 30% of locations but 54% of answers, so the 63% and 55% headline figures cover all 1,600 locations while the citation and trend findings use hotels and resorts only. Prompts were in English, 86% of locations are in the US, and monitoring used APIs plus live AI Overviews, not signed-in chatbot sessions.
The strongest objection to reading too much into the benchmark comes from Milestone itself: visibility measures inclusion. The study did not measure clicks, bookings or whether the engines' statements were accurate. For now, the link from visibility to revenue rests on separate studies run by vendors. Cendyn's Hotel Visibility Report 2026, covered by Hospitality Technology, found that 148 hotels combining generative engine optimization with SEO fixes saw organic bookings rise 66% on average, with organic revenue up 116%. Because the cohort and method are different, the report suggests a payoff but does not confirm one for Milestone's sample.
Changing the weekly report
The report's five priorities for marketers translate into a measurement change more than a content sprint. Map the questions travelers actually ask, including the follow-ups engines generate. Track each engine on its own. Keep booking listings complete and consistent, earn travel-media coverage and answer reviews. Publish accurate facts on key website pages and keep details matched across channels. Test content formats by question type and engine before spending on them.
Cendyn's report offers one data point for that last step. Some individual properties in its cohort rewrote keyword-driven pages as conversational question-and-answer content and recorded gains in visibility and engagement as high as 60%. That suggests format is worth testing, though Milestone's engine-by-engine split implies the result could differ by engine.
For hotels whose reporting still rolls AI into a single visibility number, the per-engine split is the first fix, and the tools are arriving. Hotel Dive reported in April that Operto launched a free GEO Consultant that scans a property against prompts it chooses, compares it with direct competitors and shows whether visibility comes from the hotel's own site or from intermediaries such as OTAs. The outlet's example is a Miami hotel that wants to surface for "a hotel with ocean access in Miami," a location-style question of the kind that scored lowest in Milestone's data.
Operto's separate Marketing Agent goes a step further, identifying when OTAs are "intercepting high-intent searches" and where independent hotels are losing direct bookings. Demand is not the open question: the Milestone release cites Phocuswright data from March 2026 showing 56% of active US travelers used AI for at least one trip in the past year, and Adobe data showing AI-referred visits to US travel sites up 194% year over year in May 2026.
Before buying any AI visibility dashboard, ask three things: does it report each engine separately, does it split branded from unbranded prompts, and does it show which source (own site, OTA, review or media) put the hotel in the answer?
The number worth watching next is the 1.89 hotels per Google AI Overviews answer. If that count keeps falling, the engine that gives a hotel's own pages the most weight will also be the one with the fewest open slots, and the case for owned-site work on that engine gets harder to prove.
Sources
- Hospitality AI Visibility Benchmark 2026 ↗ · Hospitality Net
- Hospitality AI Visibility Benchmark 2026 ↗
- 5W releases hotels AI visibility index | Hotel Management ↗ · Hotel Management
- Cendyn Report Links Generative Engine Optimization to Direct ... ↗ · Hospitality Technology
- Hotel Tech-in: The tool giving hotels visibility into gen AI ... ↗ · Hotel Dive
- AI Visibility Index: Travel and Hospitality Insights Q4 2026 ↗ · eMarketer
Featured companies
Your experts belong here
Every story in MarketScale Hospitality starts with a company putting its general managers, operations leads, and brand teams on the record. Buyers are already reading this topic. The only question is whose experts they find.
Owners and operators buy from people who understand the property, and your teams prove it in their own words.
About the author
The MarketScale Newsroom reports on the companies, technologies, and trends shaping 16 B2B industries. It turns primary sources and expert commentary into clear, useful coverage for the people doing the work.