Market Impact · AI Agents · Fact check

Will AI Agents Book Around Airbnb? What the "Book Direct" Pitch Would Actually Take

A pitch is going around host groups: AI agents will book the cheapest stay, usually the host's own website, and Airbnb and Vrbo will lose relevance. We traced where it came from, checked it against the data, and ran the numbers on how much agent booking it would take to dent Airbnb's revenue.

<1% Booking's room nights from AI chatbot referrals, Q2 2026
15.5% Most Airbnb's fee lets direct undercut it. The viral claim: 60%
~1 in 5 Bookings via agents for a 10% revenue hit, in the pitch's best case
~6 pts Margin at risk if agents charge a 4% toll on a quarter of bookings

Since late September, a version of the same post has been going around host groups: soon guests will tell an AI agent to find a rental, the agent will book whichever option is cheapest, that will usually be the host's own website, and Airbnb and Vrbo will lose their grip. There is a real event behind it. The conclusion runs well ahead of the evidence, and it misses the risk that actually matters. Numbered references in the text jump to the full list at the end.

A note before you read further: company figures, stock moves and survey numbers are reported and cited. The adoption model and the margin scenarios are our own arithmetic on stated assumptions, labeled as such. One key anecdote reached us only secondhand, and we say so. We found no booking data published for Meta's agent, so nobody outside Meta can yet measure how many stays agents book, or where. This is not financial or investment advice.

The short version

Claim 1
"The agent will book the cheapest option." Not today. Agents book through whatever they can transact on, and that is often an OTA.5, 7
Claim 2
"The cheapest option is usually the host's own website." Airbnb's fee caps the saving at about 15.5%, many listings have no direct site, and direct is losing share.11, 23
Claim 3
"Airbnb and the OTAs are already losing relevance." Not in any reported numbers. AI chatbot referrals are under 1% of Booking Holdings' room nights.19
Missed risk
The toll. If agents charge a commission on the bookings that stay on the platforms, Airbnb loses margin, not relevance.9

Where the pitch comes from

  • The trigger is real. Meta launched Muse on September 8, a personal AI agent that can send email, fill in forms and book travel on a user's behalf, in the U.S. only.1 Adoption was fast: analytics firm Apptopia estimated 642,000 U.S. mobile daily users by day 12, ahead of ChatGPT at the same point after its launch.2
  • So was the market reaction. On September 23, Expedia fell about 7%, Airbnb about 6% and Booking Holdings about 5%, with no company news. A Goldman Sachs trading desk note warned that businesses built on middleman fees face disruption from agents of this kind.3 Truist cut its price targets on Booking and Expedia, citing Muse, while calling the effect on OTA booking share "highly uncertain."4
  • The "cheapest, usually direct" version traces to one anecdote. Independent stock analyst Mostly Borrowed Ideas said on The Synopsis podcast that he sold his Airbnb shares after Muse told him a cottage he liked could be booked directly for 60% less.6 We could only find that story in secondhand summaries, several of them translated. It is one booking, one comparison, unverified.
  • The thesis is older than Muse. Direct-booking software companies have argued since 2025 that AI search and agents will level the field between OTAs and direct sites.26, 27 They sell direct-booking sites, so it is a reasonable argument from an interested party.

The pitch bundles three separate claims. Each one checks out differently.

Claim 1: "The agent will book the cheapest option"

An agent books what its user asks for, through whatever it can actually transact on. Today that often means going through an OTA, not around one.

  • How Muse books stays. Skift tested Muse at launch. Flights run through a direct connection to the travel infrastructure company Duffel. Hotels go through a browser, where Muse shops consumer sites such as Expedia and Hotels.com the way a person would.7
  • Expedia is joining it. Expedia agreed to let U.S. travelers book its hotels through Muse, with Expedia as merchant of record; no launch date has been announced.5 Booking Holdings stayed out.4
  • Platforms can say no. Amazon blocked Muse from its store, saying Meta never asked permission and the agent does not identify itself as a bot.8
  • Airbnb doesn't work well inside agents yet. Airbnb CEO Brian Chesky told TechCrunch he has tried Airbnb inside Muse and Instinct and that it doesn't work very well; in his words, hotels don't work that well, "but Airbnbs work even worse."10

That last point cuts against Airbnb, not for it. If Airbnb stays clumsy inside agents, agents will route around it, toward whatever is easy to book. This is a real vulnerability, and Chesky says Airbnb is working on it (see what Airbnb and Vrbo can do).

Claim 2: "The cheapest option is usually the host's own website"

Airbnb's fee caps the gap. Airbnb now charges most hosts a single fee of about 15.5% of the booking subtotal, and the guest sees no separate service fee.11, 12 A host who prices Airbnb to keep the same payout, and passes every dollar of the saving to direct guests, can undercut Airbnb by 15.5% at most, or about 12.9% after paying card processing on their own site.

Figure 01 · Our arithmetic

What the Guest Pays for the Same $100 Host Payout

The Airbnb price that keeps a $100 payout under the 15.5% fee, two direct-site prices that pass the whole saving to the guest, and the price the viral 60% claim implies.

$100 ÷ 0.845 = $118.34. $100 ÷ 0.97 = $103.09 with about 3% card processing. 60% below $118.34 is $47.34, which would leave the host about $45.92 after processing. Excludes taxes, which apply on both channels. Illustrative.

That is the ceiling the fee can explain. A 60% gap needs something else: different dates, a different unit, a promotion, or a mistaken comparison. For the same stay, a host would have to accept less than half of their Airbnb payout to match it.

  • Many Airbnb listings have no direct site to send an agent to. Airbnb ended 2025 with more than 9 million active listings.16 Nobody publishes how many also have a bookable website of their own. Our reading is that most do not, because direct booking sites are mainly run by professional managers. The "agent finds the direct site" story only applies to the slice of supply that has one.
  • Even among professional managers, direct is losing share. KeyData's index of about 1.3 million professionally managed U.S. rentals showed Airbnb at 51% of reservations in Q2 2026, up from 47% a year earlier, while direct fell from 25% to 21%.23
  • For Vrbo-listed homes, the price gap is about to be against the rules. Vrbo's new Host Terms, effective October 29, require a host's Vrbo rates and fees to be at least as favorable as on their own website and every other channel.13 An agent shopping on price should find no cheaper direct rate for those properties.

Claim 3: "Airbnb and the OTAs are already losing relevance"

Nothing in the reported numbers shows it yet.

  • Booking Holdings: traffic from large language models, paid and unpaid, is still "significantly below 1%" of room nights, with no material change in recent quarters.19
  • Expedia Group: about two-thirds of bookings come through direct channels, and AI-driven channels are still small.21
  • Airbnb, Q2 2026: revenue up 17% to $3.6 billion, gross booking value up 16% to $27.2 billion, nights up 10%, and first-time bookers up 11%, the fastest in four years. Its app now carries 64% of nights booked.17, 18
39% plan with AI · <1% book through it Phocuswright found 39% of active U.S. travelers used AI for travel research or planning by late 2025.24 The booking numbers above have barely moved. Using AI to plan a trip is not the same as an agent booking it.

What it would take: the adoption math

The pitch is about the future, so the useful question is not "is it happening?" but "how much would have to happen before Airbnb notices?" We built a simple model.

The model
Revenue lost ≈ agent share × leakage
leakage = bookable elsewhere × agent books it there
Agent share is the share of Airbnb's booking value where an outside agent (Muse, ChatGPT and the like) does the shopping and completes the booking. Leakage is the share of those bookings that leave Airbnb.

Leakage depends on two things we can't measure: whether the home is bookable somewhere else, and whether the agent finds that option, trusts it and books it there instead. So we ran three cases.

Figure 02 · Our model

How Many Bookings Agents Would Need to Make

Share of Airbnb's booking value that outside AI agents would have to book for Airbnb to lose 5%, 10% or 20% of revenue, under three leakage cases.

CaseLeakageFor −5% revenueFor −10%For −20%
Today's world20% × 50% = 10%50%100%not possible
Middle case30% × 70% = 21%24%48%95%
The pitch's best case60% × 90% = 54%9%19%37%

Leakage = share bookable elsewhere × share the agent books there. Inputs are assumptions, not measurements. "Not possible" means agents could book every Airbnb stay and still not cost it 20% of revenue at that leakage.

For scale: Airbnb has guided to at least mid-teens revenue growth for 2026.18 A 5% hit would erase about a third of a year's growth; 10% would erase most of it.

Figure 03 · Our model

Airbnb Revenue Lost as Agent Booking Grows

Revenue lost at each level of agent booking, for the three leakage cases in Figure 02. The dashed line marks a 10% revenue hit; the dotted line marks today's best public proxy for agent booking, under 1%.

Revenue lost = agent share × leakage, with leakage of 10%, 21% and 54%. Today's proxy is Booking Holdings' AI chatbot referrals, under 1% of room nights,19 which counts referrals rather than completed agent bookings.

Three readings

  1. In today's world the pitch can't produce a big hit. Agents would have to book half of all Airbnb stays just to cost it 5%, because most listings have nowhere else to go.
  2. The pitch needs two things at once. Most of Airbnb's supply (60% in our best case) would need a bookable direct site, and agents would have to find and trust it nine times out of ten. With both, one booking in five through agents costs Airbnb 10% of revenue.
  3. Where we are today is closer to 1% than 20%. Booking's sub-1% figure is the best public proxy, and it counts referrals, not completed agent bookings.19

Who uses agents matters

Agent adoption will not be even across age groups, and the two platforms draw different guests. Similarweb data cited by Alan Boswell show 44% of Vrbo's browsers are 45 or older, against 32% for Airbnb.25

We assumed travelers under 35 are 1.6 times as likely as 35-to-54-year-olds to let an agent book, and those 55 and over half as likely. We also assumed bookings split across those three age groups 45/35/20 on Airbnb, 20/45/35 on Vrbo and 30/35/35 for U.S. travelers overall. On those assumptions, Airbnb's guests are about 1.16 times as agent-prone as the average U.S. traveler, and Vrbo's about 0.94 times.

PlatformAgent-prone vs. average travelerBookable elsewhere (middle / best case)U.S. travelers booking via agents for a 10% hit (middle / best case)
Airbnb1.16×30% / 60%~41% / ~16%
Vrbo, before its new terms0.94×55% / 80%~28% / ~15%

Our model. In both cases agents book the alternative 70% of the time in the middle case and 90% in the best case. Vrbo's higher "bookable elsewhere" share reflects our assumption that more of its supply is professionally managed with a direct site.

Vrbo is the more interesting case. Its guests are older, which slows adoption, but more of its supply has somewhere else to book, which raises leakage. On our assumptions, before its new terms, a 10% Vrbo hit would need about 28% of U.S. travelers using agents in the middle case: less than Airbnb's 41%. The October 29 parity terms remove the price gap that would make an agent leave,13 which is what Vrbo needs if this ever becomes real.

The risk the pitch misses: the toll

Mark Zuckerberg has said Meta eventually plans to charge a commission on transactions made through Muse.9 If platforms end up paying agents for the bookings that stay with them, the cost shows up in margins rather than revenue. It is the same position Booking.com is in with Google search today.

We used Airbnb's 2025 revenue of $12.2 billion on $91.3 billion of booking value, a 13.4% take rate,16 and its 2026 guidance of an adjusted EBITDA margin of at least 35.5%.18

Figure 04 · Our arithmetic

Airbnb's Margin if Agents Charge a Toll

Adjusted EBITDA margin after paying a toll on agent bookings that stay on Airbnb, by agent share of bookings and toll rate (as a share of booking value). Middle leakage case.

Agent shareTollRevenue lostToll cost (% of revenue)EBITDA margin
10%2%−2.1%1.2%~34.3%
25%2%−5.2%3.0%~32.5%
25%4%−5.2%5.9%~29.6%
40%4%−8.4%9.5%~26.0%

Toll as % of revenue = agent share × (1 − 21% leakage) × toll ÷ 13.4% take rate. The margin column subtracts that from 35.5% and ignores the revenue loss itself. Illustrative.

5.9% vs. 5.2% At 25% agent share and a 4% toll, the toll costs Airbnb more of its revenue than the bookings it loses. That is the realistic version of this story: not Airbnb disappearing, but Airbnb paying a new middleman.

What Airbnb and Vrbo can do (and some are doing)

  • Become the inventory agents prefer, or block them. Expedia chose to plug into Muse;5 Amazon blocked it.8 Chesky told TechCrunch he wants the Airbnb app to become more agentic, probably interoperable with MCP, the standard that lets agents connect to other software.10
  • Own the front door. Airbnb's app carries 64% of nights booked, and its Reserve Now, Pay Later option accounted for more than 20% of booking value in Q2.17 A small direct site can't easily match either.
  • Sell the trust layer. Verified identities, payment protection, damage cover and reviews tied to real stays are exactly what an agent can't confirm about a website it found ten seconds ago.
  • Co-opt "direct." Airbnb is piloting direct booking links that cut the host fee from 15.5% to 6% or 10% when the host brings the guest, with the booking still on Airbnb.14
  • Member pricing. At Booking, top-tier Genius members are more than 30% of active customers and drive a high-50s share of room nights.20 Supplier-funded deals passed 40% of Vrbo bookings in Q2.22 Airbnb has no loyalty program, which is its clearest gap here.
  • Parity. Vrbo is enforcing it from October 29.13 In the EU, the Digital Markets Act bars Booking.com from using parity clauses.15

What we don't know yet

  • How many stays agents actually book, and where. We found no Muse booking data published by Meta.
  • How Airbnb will handle agents. It has not said whether it will partner with Muse, block it, or build its own connectors first.
  • Whether hosts will multi-list more. The pitch assumes AI makes it trivial to list everywhere with a direct site. KeyData's data shows the opposite drift so far.23
  • Who pays the toll. Zuckerberg has said there will be a commission, not who bears it or how large it is.9

What would tell us the pitch is right

  • Booking Holdings reporting LLM-driven room nights well above 1% in its Q3 or Q4 results.
  • Meta disclosing travel bookings made through Muse, and which sites they landed on.
  • Direct share rising in KeyData's Q4 2026 index above Q4 2025's 21% of reservations, in more than one quarter.
  • Airbnb reporting a falling share of direct and app traffic, or a rising cost to acquire guests, in its Q3 or Q4 results.

What to do as a host

  1. Don't plan around Airbnb shrinking. The current data points the other way, and agents still read your Airbnb listing, reviews and photos when they shortlist.
  2. Make your direct site readable by machines. Live rates and availability without a login, the same property name, address and amenities on every channel, and structured listing data. An agent can only book what it can parse and confirm.
  3. Don't undercut Vrbo after October 29. If you're listed there, a cheaper direct rate is a terms risk.13
  4. Start with repeat guests. They are the bookings where a direct site, or Airbnb's lower-fee link, actually wins.
  5. Track where bookings come from. Ask, and tag any booking where the guest says an AI assistant found you, so you'll see the shift in your own data before it shows up in anyone's earnings call.

Frequently asked questions

Will AI agents replace Airbnb?

Not on current evidence. AI chatbot referrals are still below 1% of Booking Holdings' room nights, and Airbnb's bookings grew 10% in Q2 2026. On our model, even under the pitch's most favorable assumptions, about one in five Airbnb bookings would have to run through outside agents to cut its revenue by 10%. The more realistic risk is Airbnb paying agents a commission, which hits margins rather than relevance.

Is it cheaper to book an Airbnb directly with the host?

Sometimes, by up to about 15.5% if the host passes Airbnb's entire fee on to direct guests, or about 13% once the host pays card processing. A 60% difference can't be explained by Airbnb's fees. Many Airbnb listings have no direct site at all.

Can Meta's Muse book Airbnb stays?

Muse can browse and book through consumer websites, but Airbnb's CEO says Airbnb works poorly inside current agents. Expedia has agreed to offer its hotels through Muse; Airbnb has not announced a partnership.

Should I lower my direct price so AI agents pick my website?

Check each platform's terms first. From October 29, 2026, Vrbo's Host Terms require your Vrbo rates and fees to be at least as favorable as your own website's. Clear, consistent, machine-readable listing data is a safer way to be found than a discount.

Related reading · Vrbo Vrbo's 12% Commission Comes With a Catch: You Can't Just Raise Your Vrbo Rate → Related reading · Airbnb Airbnb's Single 15.5% Fee: How the Rollout Unfolded (and What Vrbo Is Doing Next) →
Free Listing Teardown

Agents Read Your Listing Before Guests Do

Whether a guest scrolls or an agent shortlists, the listing that converts is the one that wins. If you operate in the Southeast, we'll audit your Airbnb listing against your comp set and show you where you are leaving bookings on the table.

Get my free teardown → View methodology
About this analysis

How to Read These Numbers

Claim originTrade and financial press published September 9 to October 1, 2026, and a secondhand summary of the podcast where the 60% anecdote appeared
Company figuresAirbnb, Booking Holdings and Expedia Group earnings calls and their coverage, Q1 and Q2 2026
Channel sharesKeyData U.S. Summer Index, Q2 2026, as published; we have not seen the underlying data
Our modelFee arithmetic, adoption thresholds, age-propensity indexes and margin scenarios; every input is an assumption stated in the text
Read onOctober 14, 2026. Agent products and platform policies are changing quickly
LimitationGeneral information, not legal, financial or investment advice

The model shows what would have to be true for the pitch to hurt Airbnb, not a forecast of what will happen. Its inputs are the parts nobody can measure yet, which is why we show three cases rather than one answer.

References

Numbered to match the brackets in the text. Click a number in the text to jump to its entry.

Muse and the market reaction

  1. Relevant Audience, "Meta launches Muse, an AI agent that books, fills forms and buys," Sep 2026.
  2. TechCrunch, "Meta's Muse is outpacing ChatGPT's early mobile launch," Sep 21, 2026.
  3. 24/7 Wall St., "Travel Booking Stocks Tumble as Muse Threatens to Bypass Them," Sep 23, 2026.
  4. 24/7 Wall St., "How Booking Holdings and Expedia Are Taking Opposite Bets on AI Shopping Agents," Oct 1, 2026.
  5. TravelPulse, "Expedia to Offer Hotel Bookings Through Meta's Muse AI Agent," Sep 24, 2026.
  6. TechFlow summary via KuCoin, "Analyst Sells Airbnb Stock After Testing Meta's AI App Muse," Oct 1, 2026. Secondhand summary of The Synopsis podcast.
  7. Skift, "Meta Says Its Muse Agent Books Travel. Here's How It Works," Sep 9, 2026.
  8. VKTR, "Amazon Blocks Meta's Muse AI Agent From Shopping on Its Platform," Sep 2026.
  9. CNN via CTV News, "AI agents promise to do everything for you. There may be a big wrinkle in that plan."
  10. TechCrunch, "Brian Chesky interview: AI agents need their own operating system," Oct 1, 2026.

Fees and platform rules

  1. Airbnb Help Center, "Airbnb service fees."
  2. STR Intel by Specrom, "Airbnb's Single 15.5% Fee: How the Rollout Unfolded (and What Vrbo Is Doing Next)," Oct 6, 2026.
  3. STR Intel by Specrom, "Vrbo's 12% Commission Comes With a Catch: You Can't Just Raise Your Vrbo Rate," Oct 3, 2026.
  4. Houfy, "Airbnb Direct Booking Links: What the Pilot Actually Changes."
  5. European Commission, DMA factsheet on Booking.com pricing freedom, Sep 28, 2026.

Company results

  1. Short Term Rentalz, "Airbnb Q4 2025: what the results really show property managers."
  2. MarketBeat, "Airbnb Q2 Earnings Call Highlights," Aug 6, 2026.
  3. Quartr, "Airbnb (ABNB) Q2 2026 earnings summary."
  4. PhocusWire, "Booking Holdings Q2 2026 earnings."
  5. Travelers Today, "Hotel Bookings via AI Chatbots Stay Below 1%, Booking Holdings Confirms," Aug 5, 2026.
  6. Expedia Group Q1 2026 earnings call transcript.
  7. Rental Scale-Up, "Vrbo Q2 2026 Performance: Are Hosts Paying for Its Growth?"

Market data and surveys

  1. KeyData U.S. Summer Index press release via Hotel Online, "US vacation rental revenue pacing 26% higher YoY for September," Jul 28, 2026.
  2. PhocusWire, Phocuswright's Travel Forward report for 2026, Dec 10, 2025.
  3. Alan Boswell, "Airbnb vs Vrbo: Which is best for your rental property?" (citing Similarweb).

The direct-booking thesis

  1. Lodgix, "Agentic AI and the Future of Vacation Rental Bookings."
  2. Boom, "SEO, Make Way for Intent."

Published October 14, 2026 · STR Intel by Specrom

STR Intel is independent and not affiliated with or endorsed by Airbnb, Vrbo, Expedia Group, Booking Holdings or Meta. This article is general information, not legal, financial or investment advice.