If AI shapes where travellers spend one dollar in ten online, it could influence $120 billion in annual bookings. Raven’s research across 101 airlines shows why the company that earns the recommendation may be different from the one that makes the sale.
“What’s the cheapest way to fly from London to Geneva next Friday and back on Sunday with my skis?” Ask an AI assistant and follow its recommendation. Does the fare include a ski bag in both directions? Can you reserve space for it? And whose website are you on when you reach for your card?
At Raven, we set up the Customer Growth Index to help travel and tourism brands understand how AI is changing their business and where they can benefit. We examined 9,454 answers from ChatGPT, Claude and Perplexity across 101 airlines. The assistants named the airline in 78 per cent of answers, recommended it in 55 per cent and linked to its own website in 27 per cent. For 85 of the 101 carriers, online travel agencies and comparison sites collectively received more links than the airline itself.
That is the finding I would take to the boardroom. You can operate the flight, offer the right fare and earn the recommendation, while somebody else gets the first opportunity to sell your seat or still send that passenger back to you, for a price. For airlines that have spent years bringing customers back to direct and NDC channels, that is work being undone without anyone deciding it should be. Our research shows where the assistant points. Your booking data can tell you what happens next.
How do we get to $120 billion?
Phocuswright forecast $1.2 trillion in global online travel bookings for 2026. Every percentage point is worth $12 billion. Suppose travellers consult AI on purchases representing a quarter of online booking value, and its advice changes what or where they buy on two fifths of that spend. That puts $120 billion of annual bookings within AI’s influence. The question is how quickly occasional use becomes a habit at the point of purchase.
The pace is hard to ignore. Anthropic’s 2 November prospectus reported $16.96 billion in revenue for the September quarter, up 1,127 per cent year on year. First Page Sage independently estimates that Claude’s monthly active users grew 7.4 per cent in August, 6.7 per cent in September and 7.9 per cent in October, reaching 312.4 million. In travel specifically, Similarweb estimates that AI platforms sent 534 million visits to travel websites in the year to May 2026, up about 116 per cent, and the pace of change has quickened since then.
How much of that activity can travel companies see in their tills today? A study of 41 UK travel websites published by ABTA attributed 0.37 per cent of visits to AI assistants. Net Affinity attributed 0.50 per cent of revenue across its independent hotel sample to those assistants in September. Applying those benchmarks to the global online market gives a rough estimate of $4.4 billion to $6 billion in annual bookings associated with a traceable AI click. That extrapolation assumes the samples are representative and, for the traffic figure, that AI visits produce comparable booking value per visit.
The gap between that figure and $120 billion is the influence your analytics can’t see. The skier who takes an assistant’s advice on Monday and books in an airline app on Friday illustrates what the click misses. Travel has seen this before. In 1999, online sales accounted for about 3 per cent of US travel sales, yet airline tickets already represented 73 per cent of online travel bookings. Later, a guest could discover a hotel on social media, compare rooms on a laptop and book by phone. The reservations desk got the sale. AI takes that attribution question a step further. It can decide which hotels make the shortlist before the guest knows their names. A hotel left out sees no abandoned basket. The guest never arrived.
Who wins the customer?
Similarweb’s June 2025 snapshot put Kayak, Rome2Rio, Airbnb, Booking.com and Expedia at the head of its travel ranking for AI referrals. Raven’s research shows how much the choice of assistant matters. Of 2,350 links to Trip.com in our sample, 2,209 came from Perplexity. ChatGPT supplied one. A commercial team testing only one LLM could miss a distributor that features prominently when its passengers ask the same questions elsewhere.
Booking Holdings can see what happens once a traveller reaches a seller. On its November earnings call, the group said external AI referrals still accounted for well under 1 per cent of Booking.com’s room sales. Inside Priceline and Agoda, customers who used the companies’ own AI assistants booked at roughly twice the rate of other visitors. Those customers may already have been closer to paying. What matters for our skier is that an assistant connected to live inventory can turn a conversation about baggage into a fare they can buy.
The agencies got there first. OpenAI launched apps inside ChatGPT on 6 October 2025, with Booking.com and Expedia as its travel partners. Virgin Atlantic became the first airline with its own ChatGPT app six months later, and easyJet and Southwest followed this October. Google’s AI Mode now lets travellers book hotels without leaving the conversation, with Booking.com, Expedia and Trip.com among the first booking platforms, although flight purchases still move to external sites. When an assistant looks for a live fare to show a traveller, the agencies have had a head start of between six months and a year. Sellers are also building assistants into their own websites. Virgin Atlantic’s ChatGPT app builds on its own AI-powered Concierge, Malaysia Airlines launched its Mavis assistant in early 2026, and hotel booking engines now offer chat connected to live availability.
There are two sides to the conversation I would have with a travel company today. Where is an assistant sending someone who already wants your product to another seller? And where could it introduce you to someone who has never heard of you? An agency that brings an airline a new passenger may earn every penny of its fee. The cost arises when the airline could have sold the same seat directly for less.
Take the $120 billion scenario. If AI shifted ten percentage points of that booking value from direct sales to agencies, $12 billion would change channel. If the agency channel cost ten percentage points more after allowing for direct marketing and servicing costs, suppliers would retain $1.2 billion less a year. The figures that matter to an individual airline are the passengers it won, the sales that switched channel and the margin it kept on each.
Where to start?
Start with reservations. Take ten questions customers ask just before they pay, in their own words. Put them to several assistants in a market you serve, first without your brand name to see who makes the shortlist, then with it to check the advice. Record who gets recommended, which source supports the answer and where the booking link leads. For 22 of the 101 airlines in Raven’s research, the searches behind the answers failed to surface a single page from the carrier for any question tested.
Then follow the link on a phone. For the Geneva flight, put the allowance, the charge for each leg and the means to reserve ski carriage on a page the assistant can find. If a comparison site shows the total while your website sends the passenger to a baggage policy and asks them to start again, you know what to fix. A hotel can run the same test with guaranteed connecting rooms. A shore excursion operator can try it with a refund question for a late-arriving ship. That is answer engine optimisation in practice.
After booking, ask a sample of customers two questions. Did you use an AI assistant before you bought? What did it change? Match the answers to booking value and channel. Ask people who considered your brand and booked elsewhere the same questions. They can tell you whether you lost on price, whether another seller made the purchase easier or whether the assistant left you out altogether.
If $120 billion sounds ambitious, remember how small online bookings once looked. Travel and tourism grew faster than the global economy in 2025 and contributed $11.6 trillion to world GDP. If that contribution were a national economy, it would rank third, behind only the United States and China. In a market of that scale, a small change in how people choose can move billions.
We cannot put a date on the day AI influences one booking dollar in ten. But the skier asking how to get to Geneva is already choosing from the answers on the screen. When your next customer asks, are you ready?
By Dr Louise Croft Baker, Senior Correspondent, Breaking Travel News, and Research Director, Raven
