The direct answer: 56% of travelers used AI for planning, booking, or in-destination assistance on at least one trip in the past twelve months — the fastest behavioral shift in a decade, according to Phocuswright’s March 2026 report The AI Surge. More than half of them (51%) used AI for real-time recommendations on what to do, and 95% rated it helpful for in-destination tasks like navigation, understanding neighborhoods, and managing reservations. Yet almost every hotel and tour operator GEO program targets pre-booking prompts (“best boutique hotel in Lisbon”). The queries travelers ask during the trip — “rooftop bar open now near me,” “things to do in Split when it rains,” “is the spa at [your hotel] worth it” — decide same-day, high-margin spend: dinners, spa treatments, upgrades, and last-minute tours. If you are invisible in that second window, you lose revenue from guests you already paid to acquire.
Most of the industry is still fighting the first visibility window: the planning-phase prompt where assistants shortlist hotels and tour operators weeks before travel. That fight matters, and we have covered it in depth — from how ChatGPT recommends hotels to the holiday booking window and AI travel planners. But the same research that made planning-phase GEO undeniable now shows a second window opening — smaller per query, closer to the wallet, and almost entirely uncontested.
The 2026 data: in-destination AI use went mainstream
Four data points from the last six months define the shift:
- 56% of travelers used AI for planning, booking, or in-destination assistance on at least one trip in the previous 12 months — up from 43% in H2 2025 and 33% in H1 2025 (Phocuswright, The AI Surge: Travel’s Fastest Behavioral Shift in a Decade, March 2026). The update published by Bedsonline for World Tourism Day (September 23, 2026) confirms the figure held through 2026.
- 51% used AI for real-time recommendations on what to do — not inspiration, not itinerary drafts: live, on-the-ground suggestions while traveling (Phocuswright, March 2026).
- 95% of travelers who used AI in-destination rated it helpful for navigation, learning about neighborhoods, and managing reservations (Phocuswright, March 2026). Satisfaction this high is the strongest predictor that the behavior repeats.
- 44% of travelers say they would book directly inside an AI platform, and 40% would let an agentic assistant handle flights and hotel bookings (Phocuswright data cited by Bedsonline, September 2026). On the supply side, 61% of travel businesses are already experimenting with or scaling agentic AI, and 62% expect technology budgets to increase during 2026.
Scale makes this concrete rather than theoretical: UN Tourism counted 1.52 billion international arrivals in 2025, up 4% year over year — and every one of those travelers spends three to seven nights asking questions in a destination. The majority of them now ask an assistant.
What travelers actually ask mid-trip
In-destination queries cluster into five families. Read them as a demand map, because each family maps to a revenue line:
- “Tonight” and “near me” intent — “best rooftop bar open now near me,” “where to eat in Trastevere tonight without a reservation.” Decides restaurant covers and bar spend within hours.
- Weather and disruption pivots — “things to do in Split when it rains,” “indoor activities Lisbon with kids.” Decides which attraction, spa, or tour operator absorbs a rained-out beach day.
- Worth-it questions that name you directly — “is the breakfast at Hotel X worth 28 euros,” “Hotel X vs Hotel Y spa.” Your property is already the subject; the answer decides ancillary revenue you hold.
- Logistics and reservations — “how do I book a table at,” “do I need to reserve Uffizi in advance,” “best way to get from X to Y.” Assistants that cannot resolve these against your live data default to an aggregator that can.
- Upgrade and extension decisions — “should I book the sea-view room at Hotel X,” “is it worth staying a third night in Dubrovnik.” These are direct-booking conversations happening inside a chat window.
The pattern across all five: the decision horizon collapses from weeks to minutes, and the assistant’s answer — built from maps data, review platforms, and whatever structured facts it can verify — is usually the only input.
The two visibility windows, compared
| Dimension | Planning window (before booking) | In-destination window (during the trip) |
|---|---|---|
| Example prompt | “best boutique hotel in Lisbon” | “rooftop bar open now near me” |
| Decision horizon | Weeks; comparison-heavy | Minutes to hours; same-day |
| Revenue at stake | Room nights, package price | Ancillary and walk-in: F&B, spa, upgrades, last-minute tours |
| Who wins today | OTAs, big brands, aggregators | Whoever has the freshest entity data: maps, reviews, hours, live availability |
| Freshness that matters | Evergreen authority and coverage | Tonight’s hours, seasonal closures, current menus and prices |
| Cost of invisibility | A future booking you never see | Lost revenue from a guest already in-market — often already on property |
The strategic asymmetry is in the last row. Losing the planning window costs you a booking. Losing the in-destination window costs you margin on guests you already acquired — the most expensive revenue to replace, because acquisition is sunk.
Why hotels are invisible in the second window
Five gaps account for most in-trip invisibility, and none of them require more content volume:
- Stale entity facts. Assistants resolve “open now” against maps listings and structured data. Seasonal closures (pool shut in October, restaurant dark on Mondays), renovation noise, or changed spa hours make you factually unrecommendable for tonight-queries — the AI cannot risk being wrong, so it picks the competitor whose data is current.
- Google Business Profile neglect. Review velocity, Q&A answers, and post recency are the strongest freshness signals assistants and local packs read. A profile last touched in spring is invisible in October.
- Content shaped only for planning keywords. A page targeting “luxury hotel Dubrovnik” does nothing for “what to do in Dubrovnik on a rainy Sunday.” In-trip questions need answer-first, neighborhood- and situation-shaped pages.
- No structured amenity data. Spa hours, breakfast price, parking availability, walk-in policy — if these live only in a PDF or a photo, no assistant can cite them. Schema markup (openingHours, amenityFeature, priceRange) is what makes amenities machine-verifiable.
- No same-day availability surface. When the answer is “book tonight’s cooking class” or “reserve the spa at 6pm,” the booking link the assistant finds decides who converts. If that link is an OTA widget, the second window monetizes someone else.
The playbook: eight fixes, ranked by effort-to-impact
- Audit your in-destination prompt set. List 20–30 in-trip prompts for your destination — tonight, near-me, weather-pivot, and property-named worth-it queries — and test them monthly across ChatGPT, Gemini, Perplexity, and Google AI Mode. This extends a travel AI audit with a query family most audits skip; it is also the entry point of our AI visibility program for hotels.
- Fix entity facts everywhere assistants read. Google Business Profile, Apple Business Connect, Bing Places, Wikidata: hours, seasonal closures, prices, amenities — identical and current.
- Publish answer-first amenity pages. One page per monetizable amenity (spa, breakfast, terrace, parking) with hours, prices, and booking path stated in the first sentence, marked up with schema.
- Keep review velocity in season. Ask every departing guest; respond publicly. Recency weighs more than rating for in-trip prompts. We detailed the mechanics in how hotel reviews shape AI travel planner visibility.
- Build situation pages, not just destination pages. “Split in the rain,” “Lisbon with toddlers,” “three hours in Dubrovnik” — written as direct answers with your property woven in as the obvious stay.
- Expose same-day booking links. Every in-trip answer should resolve to your direct booking engine, tour widget, or restaurant reservation link — not an intermediary.
- For tour operators and DMCs: own last-minute. Tours and activities are the most in-trip-purchasable product in travel. Real-time availability, instant confirmation, and meeting-point clarity in your structured data make you the citable answer to “what tour near me.” For programs built around this, see how a travel GEO agency structures in-destination coverage.
- Measure the second window separately. Track citation share on your in-trip prompt set, GBP calls and direction requests, ancillary revenue per occupied room, and last-minute tour bookings — then compare plans and pricing on the pricing page against the margin you protect.
FAQ
What are in-destination AI queries? Questions travelers ask AI assistants during the trip itself — “open now near me,” weather pivots, worth-it comparisons naming a specific property, and logistics like reservations. Phocuswright (March 2026) found 51% of AI-using travelers asked for real-time recommendations on what to do.
Do in-trip AI queries actually produce revenue? Yes — same-day revenue. In-trip queries precede restaurant covers, spa appointments, upgrades, extensions, and last-minute tours, all purchased within hours of the answer. With 95% of in-destination AI users rating assistants helpful for exactly these tasks (Phocuswright, March 2026), the assistant’s answer is effectively the decision.
How is in-destination AI visibility different from local SEO? Local SEO optimizes for map packs and “near me” search results. In-destination AI visibility covers the same freshness signals plus conversational worth-it and situation queries, structured amenity data assistants can verify, and same-day booking paths — because the answer, not the ranking, is the conversion.
Which fixes matter most this month? Entity-fact accuracy (hours, closures, prices) across maps listings, answer-first amenity pages with schema, and an in-destination prompt audit. All three are days of work, not quarters.
Where do I start if we have never done GEO? With a baseline: run an audit across both windows — planning and in-trip — so you know which prompts cite you, which cite competitors, and what assistants get factually wrong about your property.
The window nobody is defending
The planning window gets the conference talks because the stakes are bookings. But the in-destination window decides the revenue those bookings were supposed to deliver — dinner covers, spa slots, upgraded rooms, and the last-minute tours that tour operators live on. The data says travelers are already there: asking, trusting, and acting on answers within hours.
The properties that will win it are not the ones with the biggest content budgets. They are the ones whose facts are current tonight, whose amenities are machine-verifiable, and whose booking link is the easiest one for an assistant to find.