Airbnb is having its fastest revenue growth in more than three years, and its chief executive does not think general-purpose AI agents are the reason. On the company’s Q3 2026 earnings call on November 5, CEO Brian Chesky said agents such as Meta’s Muse and Instinct are sending “not a lot of traffic, and even fewer bookings.” Of nearly half a million Airbnb bookings a day, he said, a “very non-material” number come from those products. Traffic from the agents “kind of peaked in early October” and had been stable for about four weeks. The comments, first widely picked up by Business Insider and The Verge, landed alongside a shareholder letter filed with the SEC that reported Q3 revenue of $4.869 billion, up 19% year over year, and laid out Airbnb’s plan to ship its own specialized travel agent in the first half of 2027.
Quick Take
- The claim: Chesky said general AI agents, naming Muse and Instinct, produce almost no Airbnb bookings today — “very non-material” against nearly 500,000 bookings a day — and that inbound agent traffic peaked in early October.
- The quarter: Revenue was $4.869 billion, up 19% (16% excluding FX). Nights and seats booked rose 16% to 155.0 million, and gross booking value rose 21% to $27.7 billion.
- Airbnb’s answer: A first-party trip-planning agent for the first half of 2027, built around a generative interface of maps, photos, comparisons, calendars and itineraries rather than a text chatbot.
- The argument: 80% of travelers enjoy researching trips, and most guests already use maps, photos and reviews, often deciding with other travelers. Chesky said he sees outside agents as a discovery handoff, “kind of like Google,” not as a threat.
What happened
Chesky’s remarks came in answer to an analyst question about data showing that a lot of commercial activity on agents such as Muse and Instinct is travel-related. He did not dispute that people use those products. He disputed that they close Airbnb bookings.
“We’re not seeing a lot of traffic, and even fewer bookings from Muse and Instinct yet,” he said. “We do have nearly half a million bookings a day, and a very non-material number of bookings happen on Muse and Instinct. So, we’re talking very low numbers. We’re not seeing a lot of traffic today, and it’s not really increased over the last four weeks. It kind of peaked in early October.”
His broader stance was explicit: “I’m generally excited about them. I don’t see them as a threat. I see them kind of like Google, as an opportunity for more people to discover Airbnb. I think they’ll be great handoff to us.”
The financial backdrop was strong. Chesky opened the call by saying “Airbnb is back in growth mode” and that Q3 was “our fastest revenue growth in more than three years.” Revenue grew 19% year over year to $4.9 billion, he said; the letter’s quarterly table lists $4,869 million, or $4.869 billion, up 19% from $4,095 million a year earlier. Gross booking value grew 21% to $27.7 billion, and nights and seats booked grew 16% to 155 million. The company raised its full-year outlook to high-teens revenue growth and an adjusted EBITDA margin of about 36%, and announced a new $12 billion share-repurchase authorization on top of $2.7 billion remaining under the existing program.
Background: why Airbnb thinks travel is a poor chatbot task
The shareholder letter makes the strategic case at length. “We don’t believe people will use one AI agent for everything,” it says. “We believe the best agents will be specialized, built by companies that deeply understand the people they serve, have earned their trust, and can help them take action — not just answer questions in a chatbot.”
Today’s agents, the letter says, are “good at simple, routine tasks, like making a restaurant reservation or scheduling an appointment.” Planning a trip is different because “discovering where to go and where to stay is part of the experience.” The letter cites Skift and McKinsey’s seventh annual research collaboration for the finding that 80% of travelers enjoy spending time researching trips.
Guest behavior on Airbnb, as the letter reports it, matches that claim: more than 75% of guests use maps when deciding where to stay, over 90% browse multiple listing photos, and 70% read reviews before booking. The average booking is for 3 guests, so choosing a place is often a shared decision. More than 90% of stays involve direct communication between guests and hosts. On the call, Chesky added color that does not appear in the letter: people “typically look at a dozen listings or more” and “look at 25 photos per listing often,” and “most of our traffic happens during 9:00 A.M. to 5:00 P.M. hours.” “These people are supposed to be working, and they’re not,” he said. “The idea is that it’s highly aspirational, so I don’t think a lot of people want to offload all that to an agent.”
Airbnb says it has three advantages in building a first-party agent. First, nearly 20 years of marketplace data on how guests search and how hosts operate. Second, it is part of the reservation itself — payments, changes, refunds, rebooking and 24-hour safety support — which an outside agent cannot easily do on its own. Third, it is building “a new interface specifically for travel,” drawing on “a library of hundreds of travel-specific interface elements — maps, photos, comparisons, calendars, itineraries — and compos[ing] them in real time.” Chesky called that “a little bit like generative UI.” The company plans to start rolling the agent out to guests in the first half of 2027. Over time, it wants the agent to plan an entire trip across homes, hotels, services and experiences, and Chesky said it will integrate outside travel supply through MCP connections while staying focused on travel rather than becoming a general agent.
How Airbnb is already using AI
The same letter that downplays outside booking agents describes a company leaning hard on AI internally. Airbnb says it shipped more than 50 products and features in Q3, nearly 70% more than a year ago, “because of AI.” Guest-facing features already include natural-language and voice search, AI-suggested filters, AI listing descriptions in search results, and side-by-side AI comparisons of saved homes. Hosts got 3D floor plans, dynamic pricing recommendations and AI-generated earnings insights.
Customer support is the most quantified win. Airbnb’s AI assistant is available in more than 50 languages. In Q3, approximately 40% of all customer-support issues — not just those that begin in the assistant — were resolved without a human agent, up from a third in Q2. Support cost per booking declined more than 17% year over year. The company is expanding the assistant to voice support in the United States.
Business Insider separately reported that executives at Booking.com and Expedia also said on earnings calls the same week that AI agents had so far resulted in very little traffic, and that DoorDash CEO Tony Xu said the previous week that third-party agents had generated “several hundred orders in total.” Those are other companies’ remarks, relayed by Business Insider, not Airbnb’s.
| Claim or figure | Source | Notes |
|---|---|---|
| Q3 2026 revenue | $4.869 billion, up 19% year over year (16% excluding FX) | Shareholder letter quarterly summary; the letter also headlines the figure as $4.9 billion, matching Chesky’s wording on the call |
| Nights and seats booked / GBV | 155.0 million nights and seats, up 16%; $27.7 billion GBV, up 21% | Shareholder letter and Chesky’s opening remarks |
| Bookings from general AI agents | A “very non-material” number from Muse and Instinct, out of nearly half a million Airbnb bookings a day | Chesky on the Q3 call, answering a question about early agent adoption |
| Agent traffic trend | Peaked in early October and had not increased over the following four weeks | Same call remarks |
| Airbnb’s own agent | Specialized trip-planning agent, first half of 2027; generative UI of maps, photos, comparisons, calendars and itineraries | Shareholder letter and prepared remarks |
| Why travel is different | 80% of travelers enjoy researching trips (Skift and McKinsey, cited in the letter); more than 75% of guests use maps; over 90% browse multiple photos; 70% read reviews; average of 3 guests per booking; more than 90% of stays involve guest-host communication | Letter figures; on the call Chesky also said users often look at a dozen listings and about 25 photos each, and that most traffic arrives between 9 a.m. and 5 p.m. |
| AI customer support | About 40% of all Q3 customer-support issues were resolved without a human, up from a third in Q2; support cost per booking down more than 17% year over year | Shareholder letter and call |
Pros and cons
For users
- Pro: Airbnb is building a trip planner around maps, photos and comparisons — the tools guests already use — rather than asking them to hand the whole decision to a chatbot.
- Con: The first-party agent is slated for the first half of 2027, so travelers who want agent help on Airbnb will wait, and today’s general agents still close almost no stays.
For developers
- Pro: Chesky’s interface argument is a design brief: specialized agents can generate custom UI components, while a general agent is pushed toward text because it has to do everything.
- Con: If bookings stay “very non-material,” consumer agent platforms still lack a reliable travel checkout path, and Airbnb has not published technical details of its 2027 agent or its MCP integrations.
For companies
- Pro: A travel marketplace can treat outside agents as a discovery channel, the way it treated Google, while using AI to cut support costs (40% of Q3 issues resolved without a human) and ship product faster.
- Con: If visual, specialized agents later learn to complete bookings — Chesky did not say that will never happen — Airbnb will be competing with them on the same trip-planning surface it is now building.
What X is saying
The reaction split between people who saw confirmation that travel is too aspirational to delegate and people who saw a threat deferred, not disproved.
Airbnb gets nearly 500,000 bookings a day. CEO Brian Chesky says AI agents like Meta’s Muse and Instinct still bring in a “very non-material” number. Travel may be harder to automate when discovery is part of the fun.
— Foadue (@foadue) November 7, 2026 · View on X
Uber is done. Doordash, done. Now AirBnB says very non material traffic from Muse/Instinct. BTW, these are the things that every influencer was talking about Muse/Instinct being useful for. Turns out, people dont need it at all.
— nutanc (@nutanc) November 6, 2026 · View on X
Airbnb CEO Brian Chesky said bookings from AI agents Muse and Instinct remain “very non-material.” He also said 80% of travelers enjoy planning. A useful product question: are you removing a chore—or automating the part customers enjoy?
— Ajay (@i_amajay018) November 8, 2026 · View on X
Andy West put the same earnings comments in a wider frame: Airbnb and Expedia had both said personal AI agents were completing an immaterial portion of bookings, with people using them for research and itineraries but booking manually.
Critical analysis
Chesky is making two claims that are easy to blend together. The first is empirical and narrow: today, Muse and Instinct are not a meaningful source of Airbnb bookings, and their traffic has not grown since early October. That is a company metric only Airbnb can see, and nothing in the public letter independently audits it. It is still the most direct data point yet from a large travel marketplace.
The second claim is strategic: people will not want a general agent to choose their trip, because research is part of the fun. The 80% figure, the map/photo/review stats and the “secretly browsing at work” anecdote all support that story. They do not prove it will remain true once agents can show the same maps, photos and comparisons Chesky wants his own product to generate. His sharpest argument may be the interface one: a general agent is stuck with a general UI. A travel company can build the components. That is a product thesis, not yet a shipped product.
There is also a tension inside Airbnb’s own AI story. The company is happy to automate the parts of travel guests do not enjoy — customer support, filter-picking, listing summaries — and reluctant to automate the part they do enjoy. That is coherent. It also means Airbnb’s 2027 agent has to help people browse without taking the browsing away. If it becomes another chatbot, Chesky will have argued against himself.
What’s still open
- Will Muse, Instinct or other general agents start converting more Airbnb traffic after this early-October plateau?
- What will Airbnb actually ship in the first half of 2027, and how much of a trip will it book natively versus via MCP handoffs?
- Can a first-party agent stay specialized and still cover flights, cars and activities well enough that guests do not go back to a general agent?
- Will other travel companies publish comparable booking-share figures, so “very non-material” can be compared rather than taken on trust?
About the analysis
Researched and written with AI assistance under the editorial responsibility of the i10X News team. The title image is an AI-generated illustration. Read our AI Use Policy. Spotted an error? Request a correction.
This independent, research-based i10x analysis draws on Airbnb’s Q3 2026 shareholder letter (SEC Exhibit 99.1), the November 5 earnings call transcript, and reporting by Business Insider and The Verge. Revenue is taken from the letter’s quarterly table ($4,869 million, up 19%). Chesky’s Muse and Instinct comments are from the call. The 80% research-enjoyment figure is cited in the letter to Skift and McKinsey.
i10x Perspective
The interesting sentence on the call was not “very non-material.” It was Chesky describing outside agents as a handoff, like Google. Marketplaces have heard that story before: a new front door appears, traffic looks small, then the front door starts completing the transaction. Airbnb’s hedge is to build a richer travel surface than a chatbot can offer and to stay inside the reservation. That is a reasonable plan. It is also a race against the same visual agents Chesky says are not yet a threat. The next useful disclosure is not another metaphor about IKEA versus Walgreens. It is a number: what share of bookings, if any, outside agents hold a year from now.
