There are mixed results from aggregators around how AI is impacting their business.
OpenTable’s integration inside ChatGPT is turning into one of the restaurant booking platform’s fastest-growing channels, according to its parent company Booking Holdings. On Booking Holdings’ Q3 2026 earnings call, CFO Ewout Steenbergen said OpenTable’s native integration with OpenAI generated 190,000 seated diners in September alone, calling it one of the clearest early examples of AI driving incremental demand rather than just assisting bookings that would have happened anyway.
“OpenTable’s native integration with OpenAI has become one of the brand’s fast[est] and growing channels, and had in September 190,000 seat covers,” Steenbergen told analysts, adding that “OpenTable’s agentic interactions with Muse and other horizontal agents have brought a significant number of completely new customers to our platform.”
The 190,000 figure comes from OpenTable’s own integration built directly with OpenAI, where diners can search and reserve a table from within ChatGPT itself. The company separately flagged bookings coming through third-party AI agents such as Meta’s Muse, without putting a number against them. So the standout stat isn’t really a case of an autonomous AI agent going out and booking a table on someone’s behalf; it’s closer to OpenTable showing up as a built-in option inside a conversational product that hundreds of millions of people already use, with the diner still doing the choosing and confirming. Booking.com’s broader AI referral traffic tells a similar story: it’s growing fast off a small base, with daily AI search visitors to the company’s Trips feature up tenfold between the second and third quarters, but AI referrals overall still account for significantly less than 1% of Booking.com’s room nights.
That distinction is worth sitting with, because it comes just days after some of OpenTable’s closest comparables in the consumer-booking world reported the opposite problem: agents that technically exist, but that almost nobody is actually using.
Uber’s CFO Balaji Krishnamurthy disclosed on the company’s own Q3 earnings call that during Halloween week, one of Uber’s busiest periods of the year with roughly 320 million trips and orders, AI agents accounted for only about 500 Uber Eats orders and four Uber rides. DoorDash’s Tony Xu offered a similarly modest figure days earlier, putting third-party agent orders from the likes of Muse, Instinct and Fo at just a few hundred against the company’s roughly 1 billion quarterly orders. Airbnb CEO Brian Chesky went a step further, saying plainly that traffic and bookings from AI agents like Muse and Instinct remain “non-material” against the roughly half a million bookings Airbnb does organically every day, and that the trickle it does see peaked in early October and has stayed flat since.
In that context, Booking Holdings’ 190,000 number looks far more impressive than Uber’s four rides or Airbnb’s rounding-error traffic, but the comparison isn’t quite apples to apples. Uber, DoorDash and Airbnb were all talking specifically about third-party autonomous agents transacting on a user’s behalf, the kind of agentic commerce that companies like OpenAI, Meta and various startups are trying to build out. OpenTable’s 190,000 seated diners, by contrast, came through a native integration where OpenAI’s own product surfaces OpenTable inventory to a person who is still driving the conversation and making the booking decision themselves, much closer to a well-placed distribution channel than to an agent independently completing a transaction. Booking Holdings’ own description of a separate, smaller stream of bookings from “agentic interactions with Muse and other horizontal agents” effectively draws that same line.
Even allowing for that distinction, Booking Holdings executives were careful not to overstate where things stand. Steenbergen said AI-referred traffic across the company “is still very small in aggregate,” and CEO Glenn Fogel acknowledged that “the referral numbers are pretty darn small” even as he argued the opportunity could grow meaningfully over time. The company also pointed to efficiency gains elsewhere in the AI story that are already showing up in its numbers: Agoda’s blended cost per 1 million AI tokens fell by more than 25% between June and September, and companywide cost per merge request declined by double digits as developer productivity improved.
Taken together, the picture emerging from this round of earnings calls is consistent even where the numbers diverge. Native, first-party AI integrations that sit inside products people are already using, like OpenTable’s ChatGPT integration, appear to be finding real traction early. Fully autonomous third-party agents booking rides, meals and stays on a user’s behalf, the kind Uber, DoorDash and Airbnb were asked about, are still barely registering.