The Itinerary Is the New Search Result: How AI Picks Which Hotels and Destinations to Recommend

The planning journey collapsed into a single prompt

For two decades, travel ran on the longest funnel in consumer commerce. The average leisure trip spanned 36 days from a traveler's first research visit to a first booking, across 29-plus website visits, according to Tripadvisor data cited by Forbes. Inspiration, comparison, validation, and purchase were spread across dozens of tabs, sessions, and devices. The discipline of travel marketing was the discipline of being present in as many of those tabs as possible.

That funnel is collapsing into a conversation. In Phocuswright's research, 56% of US leisure travelers used an AI tool for at least one trip within the past 12 months, more than double the level in 2024. In a Skift Research and McKinsey joint study, the share of travelers using AI tools "extensively" for trip planning rose 124% year over year, from 13% to 30%. McKinsey found that 84% of travelers who used generative AI for planning said it improved their experience.

The behavior is generational and accelerating. 58% of US millennials have used AI for trip planning, versus 45% of Gen Z and just 11% of baby boomers (Phocuswright). And 84% of travelers who have used AI for travel started doing so within the past year, which tells you the curve is still bending upward, not leveling off.

Traditional search is feeling it directly. Phocuswright found search engines' share as a trip-planning starting point dropped from 51% in late 2024 to 36% by the second half of 2025. The first tab a traveler opens is increasingly a chat box, not a search box.

The 36-day, 29-visit journey is compressing into a single conversation. The engine reads the reviews, compares the options, and hands the traveler a shortlist before they ever reach a hotel website or an OTA.

This is the inversion that defines the vertical. AI does not send a curious browser to your property. It sends a traveler who has already been told where to stay. By the time that person sees your name, the comparison is done and the recommendation is made. The question this report answers is the one every hotel, destination, and travel brand now has to ask: how does the engine decide which names to say?

Being seen vs. being selected: the biggest brand doesn't win the answer

The most important finding in travel GEO is also the most counterintuitive. The brand with the most rooms, the most flights, and the largest loyalty program is frequently not the one the AI names.

In 2026, 5W released the Airlines & Hotels AI Visibility Index, the first study to rank travel brands by citation share inside ChatGPT, Claude, Perplexity, and Google AI Overviews, using 60-plus consumer-intent prompts across leisure, business, family, luxury, and budget travel. The structural pattern was stark. In several sub-categories, the top three brands captured more than 70% of total citation share, leaving more than twenty competitors fighting over the remainder. AI does not distribute visibility. It concentrates it.

But concentration did not follow size. 5W's headline finding was that loyalty programs do not predict AI visibility. Several of the largest loyalty programs in travel underperformed their market share inside AI answers, while smaller brands with stronger earned-media footprints punched well above their weight. The brands that "win the answer" were the ones with sustained tier-1 press coverage and structured authority on the publications the engines trust, not the ones with the biggest paid-media and OTA-distribution spend.

The same dynamic shows up at the property level. A Skift analysis of travel AI visibility found that when AI agents answer questions about a specific hotel brand, third-party and user-generated sources often outrank the brand itself. Asking about a Hyatt property, NerdWallet was cited at 13.6% of citations, more than Hyatt's own website at 10.3%. The brand's own domain was not the primary source of truth about the brand.

• Earned media is the dominant signal across both airlines and hotels (5W).
• The top three brands take 70%+ of citation share in several sub-categories; everyone else is rounding error (5W).
• Third-party publishers can out-cite the brand's own site on questions about that brand (Skift).

AI visibility is a winner-take-most market, and size does not buy a seat. The engine selects a short list and reads it back. Being the biggest chain gets you seen by humans. It does not get you selected by the model.

This is the visibility-versus-selection gap, and in travel it is brutal. You can have rate parity, top OTA placement, perfect classic SEO, and a loyalty program with fifty million members, and still never be named when a traveler asks an engine where to stay. Selection runs on a different set of signals.

The signals that get a property or destination recommended

If the engine is not leading with your website and not weighting your loyalty tier, what is it using? Travel AI recommendations are assembled from a consistent set of signal classes, and they map directly to what a brand can and cannot control.

1. Reviews, read for substance not stars. Modern engines process thousands of reviews at once, using natural-language processing to scan for recency, sentiment, and specific experiences, tagging themes like "quiet for business travel," "perfect for families," or "dated decor" through pattern recognition (Mediaboom). The engine does not average your star rating. It reads what guests actually said and looks for consistent, specific, recent signal. A property with a slightly lower average but rich, current, detailed reviews can be selected over a higher-rated property with thin or stale review content.

2. Multi-source consensus and consistency. AI recommendation engines build trust by cross-referencing a property across many sources at once: the website, Google Business Profile, OTA listings, Tripadvisor, review platforms, travel blogs, local guides, and social. They look for consistency, credibility, and clarity (Mediaboom). Contradictory data is a disqualifier. Hospitality Today reported that hotels with unclear, incomplete, or inconsistent data risk losing direct bookings outright, because AI agents favor sources with structured, reliable information, and that data-quality gaps could accelerate OTA dependence.

3. Reddit and community consensus. Nearly all AI-recommended hotels appear across YouTube, travel blogs, and Reddit, and the more a property shows up in trusted third-party content described consistently and positively, the more confident the engine becomes (TakeUp AI). The Skift analysis confirmed that user-generated platforms like Reddit are among the most-cited sources in travel answers, frequently out-citing the brands themselves. Community discussion is not a side channel. It is a primary input.

4. Third-party "best of" lists and editorial. For "best hotels in [city]" or "best food tour in Rome" queries, the engine's shortcut is to defer to the publishers humans already trust for that judgment. A placement in a credible roundup, a guidebook, or a destination editorial feature is worth more to AI visibility than any amount of on-site copy claiming you are the best.

5. Structured data and machine-readable content. The controllable foundation. Hotels and destinations that publish content matching how travelers actually phrase questions, structured so generative models can extract and assemble it, perform measurably better (Bored Hotelier). Natural-language queries are now the norm: "pet-friendly hotel near the temple with early check-in," not "hotel in Varanasi" (AxisRooms).

AI does not rank your property in isolation. It ranks it relatively, against alternatives, across reviews, cross-source consistency, community sentiment, and editorial consensus, then names a winner.

Notice the pattern. Four of the five signal classes live off your own domain. The job has shifted from optimizing a website you own to seeding and shaping a consensus you can influence but not author. For an independent hotel, that is the entire game.

Earned media beats paid placement: where the engines actually look

The single clearest takeaway from 2026 travel research is that the source mix AI engines draw on is dominated by earned and owned content, not advertising or social.

Promodo's analysis of travel AI sources found that ChatGPT relies on earned media most heavily, 61% of sources, followed by owned content at 44%, while social media accounts for less than 1%. That ordering is decisive. The PR mention, the guest contribution, the third-party feature, and the genuinely useful owned guide carry the weight. The paid social campaign and the influencer post carry almost none.

This is why the 5W index found brands leaning on paid media and OTA distribution underperforming brands with sustained press coverage. The engine is not impressed by spend. It is persuaded by the consensus that forms when credible third parties write about you consistently over time. Recomaze's travel research reached the same conclusion from the other direction: brand mentions correlate more strongly with AI travel visibility than backlinks alone, and investment in digital PR, guest contributions, thought leadership, and community engagement builds the footprint AI recognizes as authoritative.

There is a structural reason travel rewards explanation. AI uses fan-out querying, breaking a complex travel request into sub-queries, then assembling an answer from the best source for each (Recomaze). A request like "five-day family trip to Lisbon with a pool hotel and kid-friendly food" fragments into hotel, neighborhood, dining, and activity sub-questions. Content that answers a specific sub-question wins that fragment, even from a small operator. This is how a boutique hotel or a niche tour operator outranks a global platform on a narrow query the platform only answers generically.

• ChatGPT travel sources: 61% earned media, 44% owned content, under 1% social (Promodo).
• Brand mentions correlate with AI travel visibility more strongly than backlinks alone (Recomaze).
• Fan-out querying lets niche operators win specific sub-queries large platforms answer generically (Recomaze).

In travel, the engine reads the press, the reviews, and the genuinely useful guides. It barely reads social. Earned authority is the currency, and you cannot buy your way to it with media spend.

The corollary is uncomfortable for marketing teams built around paid acquisition. The lever that moves AI visibility is the slowest, least-controllable lever in the toolkit: being genuinely good enough that credible people write about you, then making sure the engine can read what they wrote.

The booking funnel is moving into the chat, and the economics favor it

The reason this matters now, rather than someday, is the conversion economics. AI is not just a small new traffic source in travel. It is, per visit, the most valuable one.

Adobe's 2026 data found that revenue per visit from AI-source traffic is 37% higher than non-AI traffic, and that AI referrals converted 31% more than other traffic sources during the 2025 holiday season, nearly doubling year over year. Travel was among the fastest-growing sectors for AI visit share, up 233% year over year in the first quarter of 2026, and AI-driven travel traffic rose 539% during the 2025 holiday season. Adobe earlier documented generative-AI-driven travel site traffic up 3,500% off its small base. The slope is exponential and the quality is rising at the same time.

The intent explains the quality. The engine pre-qualifies the traveler. It has already filtered for fit, compared options, read the reviews, and surfaced your property as the answer to a specific need. There is no top-of-funnel browsing to discount. The traveler who clicks through has been handed a recommendation, and recommendation traffic closes.

For independent hotels, this reframes the entire OTA economics question. An OTA booking carries a base commission of 15% to 18% on Booking.com, but the all-in cost reaches 18% to 30% once Genius and Visibility Booster are added, and 17% to 23% on Expedia with Accelerator (Cloudbeds and industry analysis, 2026). A direct booking, by contrast, costs roughly 5% to 12% all-in. GEO is the mechanism to intercept guest intent at the most profitable stage, before the traveler ever reaches an OTA, by being the property the engine names in the first place.

• AI travel traffic: 37% higher revenue per visit, 31% better conversion vs. other sources (Adobe).
• Travel AI visit share up 233% YoY in Q1 2026; holiday AI travel traffic up 539% (Adobe).
• OTA all-in cost 18–30% (Booking.com) vs. 5–12% for direct (2026 industry analysis).
• The global travel market is roughly 40% OTA-dominated; AI visibility now shapes that split (AxisRooms).

A channel that delivers 37% higher revenue per visit, converts better, and is growing triple digits a year is not a side experiment. For an independent hotel, it is the most direct path off the 18-to-30% OTA tax.

The platform that built modern travel discovery is itself being disrupted by this shift. Tripadvisor acknowledged on earnings calls that traditional search traffic is falling and will play a smaller role in future bookings, and it partnered with OpenAI to build AI itinerary generators using its review corpus (Hospitality Today, Hotel Dive). When the company that defined travel discovery concedes the model is changing, the bellwether is unambiguous.

Agentic booking is coming, but the checkout retreat is the real lesson

The widely predicted endgame is agentic booking, where an AI agent does not just recommend a hotel but books it, pays for it, and manages the trip end to end. Phocuswright found more than 60% of travel businesses surveyed are experimenting with or scaling agentic AI, with 6% already scaling actively. The infrastructure is being built. Expedia partnered with OpenAI on access connectors and with Perplexity for its Comet browser, and ChatGPT launched in-chat apps starting with Booking.com and Expedia.

But 2026 delivered a clarifying reversal. OpenAI spent most of 2025 building an in-chat "Buy Now" capability so travelers could research and book in the same window. In March 2026, it quietly pulled checkout for travel, concluding the category was too complex (Skift). The reason matters: travelers were happy to ask the engine for ideas, but when it came time to enter a credit card, they left and booked somewhere they trusted. Expedia Group's own survey found only 8% of travelers are comfortable letting AI handle the actual booking, even as a majority use AI to plan.

The market read the retreat as good news for the transactional layer. Expedia's share price jumped 12% and Booking Holdings rose 8% on the news (Skift). The emerging consensus: OTAs may lose their grip on the discovery interface, but they remain the trust layer and transactional intermediary. The current model routes a traveler through conversational research in the chat, then hands them off to Booking.com or a hotel's own site to complete the reservation.

The strategic lesson is not to wait for agents. It is that the recommendation, not the transaction, is the contested ground today. The engine decides who to name long before anyone enters a card. And one durable risk hangs over the whole system: AI-generated fake travel reviews rose 137% from 2019 to 2024 (Originality.ai). If the engines increasingly learn from reviews that are themselves machine-generated, the trust foundation under every travel recommendation erodes, which raises the premium on verifiable, first-party, structured proof a property can actually stand behind.

Don't optimize for a buy button that keeps moving. Optimize to be the name the engine says, on whatever surface the traveler is standing. The recommendation is the contested ground. The checkout is downstream.

The execution playbook for hotels, destinations, and travel brands

The work divides into what you own, what you influence, and what you measure. In travel, the influence column is the largest, and it is where the wins are.

Make your property data machine-readable and consistent everywhere. This is the controllable foundation and the fastest disqualifier to fix. AI agents drop properties with contradictory data.

• Implement complete Hotel, LocalBusiness, Review, AggregateRating, and FAQPage schema, kept dynamic so rates, availability, and amenities stay current.
• Enforce data consistency across your site, Google Business Profile, OTA listings, and Tripadvisor. Same name, address, amenities, and descriptions everywhere. Inconsistency reads as unreliability.
• Render critical content server-side. If amenities, location detail, and policies only load after JavaScript, assume the engine never sees them.

Earn the third-party signals you cannot author. Roughly 61% of what ChatGPT draws on is earned media, and the engine weights consensus across sources it did not get from you.

• Pursue placements in the "best of" roundups, guidebooks, and destination editorial the engines defer to for "best X" queries. One credible third-party recommendation outweighs a page of self-description.
• Build genuine, recent, specific review depth. The engine reads reviews for substance and recency, so a steady flow of detailed, current guest stories matters more than a frozen high average.
• Participate authentically in travel communities, especially Reddit. The strategy is real contribution and being genuinely worth writing about, not seeded spam, which engines detect and discount.

Write for the question, not the keyword. Travelers ask in natural language and the engine fans the query out into sub-questions.

• Build content clusters: a pillar page plus related articles that demonstrate topical depth on a destination, neighborhood, or experience (Recomaze recommends 5 to 10 supporting pieces per pillar).
• Answer the specific long-tail questions travelers actually ask: "pet-friendly hotel near the old town with early check-in," "best beachfront hotels for families in [city]," "quiet hotel for business travel near the convention center."
• Lead with verifiable specifics, location, amenities, policies, distances, that the engine can extract and quote, not atmospheric copy it cannot.

Use GEO as your direct-booking strategy. Intercept intent before the OTA. Package experiences with local partners that are only available on your direct channel, differentiating your listing from the commoditized OTA version (Revinate), and make sure those differentiators are described in machine-readable terms.

Measure selection, not just traffic. Classic analytics tell you who arrived. They cannot tell you whether the engine named you in the first place, and in travel the recommendation happens before any click. Track your share of voice in AI answers, the itinerary and "best of" prompts you appear in and miss, and which competitors get named when you do not. Dedicated travel AI-visibility tools have already emerged, a signal of how fast the category is maturing.

The old job was ranking a page you control. The new job is being selected by a system you don't, built on reviews, consensus, and earned authority you mostly don't own. Visibility was a placement. Selection is a reputation.

The hotels, destinations, and brands winning this transition are not the ones with the biggest loyalty programs or the deepest ad budgets. They are the ones the engine can read clearly, that credible third parties describe consistently, and that answer the traveler's actual question better than anyone else. The funnel collapsed into a conversation. The only question that matters now is whether your name is in the answer.

Every statistic in this piece follows Visigent's research standards: named sources, dated figures, corrections on request to help@visigent.com.