AI traffic to US travel sites keeps climbing. What hotels and operators should make machine-readable

What the travel numbers say

AI is sending more visitors to US travel sites every month, and those visitors are unusually engaged. Adobe reported that traffic from AI sources to US travel sites grew 194% year on year in May 2026, measured on more than 8 million visits to US travel sites.¹ Once on site, AI-referred travellers were 21% more engaged than other visitors, stayed 70% longer per visit and bounced 41% less often.¹

The newer reading is softer on growth and better on bookings. Adobe published its July travel figures on 19 August: AI traffic to US travel sites grew 119% year on year in July 2026, and AI-referred visits converted only 1% below other traffic, against 47% below in July 2025.² Engagement held up, with AI visitors 21% more engaged, staying 67% longer and bouncing 42% less often.²

Adobe sells analytics and AI visibility tools, which is worth knowing when a vendor reports strong growth in a channel it measures.

US travel sites, May and July 2026

More AI visitors, staying longer, closing the booking gap

+194%AI traffic, year on yearAdobe, May 2026 ¹

70%longer visits than non-AI trafficAdobe, May 2026 ¹

1%lower conversion than non-AI traffic, down from 28% in MayAdobe, Jul 2026 ²

Why travel answers link out more than most

Travel questions end in a booking, so answers point somewhere. Similarweb measured how often US ChatGPT prompts carried a citation between June 2025 and May 2026. Across all prompts the share rose from about 1.6% to about 6.8%; in travel and hospitality it was about 23%, against under 4% in professional services.³

That makes travel one of the sectors where an answer most often hands the user a link. Our note on Similarweb's sector data covers the method and why a single month should not be read as the trend.

Our view: a high citation rate raises the stakes on accuracy. When the answer links to a page, that page has to confirm what the answer said about price, fees and cancellation. If it cannot, the traveller bounces back to the chat and asks about someone else.

The habit is not only American. In a Marriott Bonvoy survey of 4,017 adults in the UAE and Saudi Arabia, published in November 2025 with fieldwork dates not stated, 72% said they had used AI to plan or research a holiday.⁴

How readable are travel sites today?

Partly. Adobe's AI Content Visibility Checker scores how much of a page a large language model can read. In the July release, hotel homepages scored 71% on average, cruise lines 70%, car rental 64% and airlines 42%; on product pages, cruise lines led at 76% and airlines trailed at 59%.² In May, Adobe had said more than a third of hotel and car rental pages remained unreadable by AI, even for sector leaders.¹

The gaps follow a pattern. Rates load from a booking engine after the page renders. Fees appear only at checkout. Cancellation terms sit in a modal. Amenities are icons without text. A crawler sees a hero photo and a date picker. Check robots.txt too: a rule meant for training bots can also shut out the search and user bots that answers rely on, as our note on blocking training crawlers explains.

What to make machine-readable first

Publish the facts an answer repeats, as text, on pages that load without a booking session. In order of how often they matter to a traveller choosing between options:

Five facts every property or operator page should state in text

1. Rates with fees

   A typical rate range by room or product type, with resort, cleaning or service fees named.

2. Cancellation and deposit terms

   The rule in one sentence on the page, not only in a pop-up at checkout.

3. Check-in, age and pet rules

   The questions travellers ask an assistant before they ask you.

4. Amenities as words

   Parking, breakfast, accessibility, airport transfer: text next to every icon.

5. Availability signals

   Seasons, minimum stays and closures stated, with a crawlable link to book.

Live availability cannot be static, and it should not be. What helps is a stable page that says what exists, what it typically costs and how to check dates, with the booking engine one link away.

Structured data is worth adding for hotel and offer details, but do not expect it to do the job alone. Our review of the schema evidence found it helps machines parse a page without being a citation switch. The text has to be there first.

How to tell whether it worked

Measure the answers, not only the traffic. Pick the prompts travellers ask in your segment, such as family hotels with free cancellation, and run them on each engine on several days. Record whether you are named, what the answer says about your price and terms, and which page it cites.

A single run is noise; our buyer's spec for an AI visibility score explains how many runs a number needs before you trust it. Our page for travel and hospitality lists the prompts we measure for hotels and operators.

Our view: travel is the sector where AI answers already behave most like a booking channel. Treat your rooms and product pages as source material for those answers, written so a machine can quote them without guessing.