AI Absorption Ledger / EXPE

Expedia Group

EXPE · Q2 2026 · reported 2026-08-05 · revenue $4.32bn

Assessment

No channel carries a management-given size this quarter either, and the quarter's disclosure is thinner than the prior one. The order-of-magnitude claim for AI in marketing, the service interaction counts and the statement that AI costs are rising were not repeated. AI hiring, partner onboarding, content enrichment and partner advertising tools, all described in the prior quarter, went unmentioned.

Several steps stand out. The 10-Q sentence that credited payments and customer service for the fall in cost of revenue against revenue now credits payments alone, so the traveler support channels lose the filing line that gave them a shape, and with the call silent on them they are read as not mentioned. Management said its natural language search and agents are not driving conversion right now, which moves that channel from an offensive reading to an exploratory one in its own words. And internal adoption moved from described to a claimed result: the release and the call say AI lets teams ship faster and operate more efficiently, while the filing attributes the change in technology and content personnel cost to previously announced cost saving initiatives, which is recorded as a confound and not as an AI saving.

The lines that had a shape in the prior quarter still move the same way. Consumer direct selling and marketing grew against consumer bookings growth of , leaving it below its prior-year share of bookings, a smaller gap than in the prior quarter and one management credits to measurement and discipline. Other revenue grew beside a repeated claim of record attach rates, with an acquisition now in the line. Licence, maintenance and cloud costs in technology and content rose .

New channels opened. An agentic voice solution now handles Vrbo partner inquiries that human agents handled before. The CEO addressed organic search for the first time, calling SEO a bit soft and stabilized and organic traffic overall stable to slightly up; the softness is credited to search page changes without naming AI, so it is context, not a channel. And the company announced the acquisition of Layla, an AI planning app, after the quarter ended, at an undisclosed price.

Ballparks. Most channels management described without a number carry the ledger's own inferred range, labelled as such and built from the reported lines with every assumption written down: the vendor bill at , creative at , internal adoption at , ranking and personalization at , the traveler-facing features at , referrals from AI platforms at , AI-platform ads at , and partner support automation at . Marketing reallocation is left unsized because AI is credited beside measurement, targeting and spend cuts with nothing to separate its part. Self-service and agent assistance are read silent: the call does not mention them and the 10-Q’s one-clause aim is context. The organic search toll is withdrawn as a channel, since management credits the SEO softness to search page changes without naming AI. Demand lost to AI planning products and the Layla acquisition stay unsized: the first has no reference class and the second fell after the quarter at an undisclosed price. Each range spans at least an order of magnitude; the points are placeholders for comparison, not findings.

Sized channels against the income statement, Q2 2026

8 of 18 channels sized

Each blue mark is one channel's dollars for the quarter; a bar is the range of an estimate. Grey marks are the company's reported lines. The distance between them is the point: how large the AI channel is next to the line it sits in.

New money and old money, Q2 2026

4 new10 expanded4 relabelled

Each channel is tagged once for whether its money existed before language models, from the company's annual report and call at the start of the period. A bar splits one flow's sized dollars by that tag. The incremental total is the part that would not be there without the models: a new channel counts in full, an expanded one only for what AI added, a relabelled one at zero.

Paid for AI$1.1mn to $11mn sized

new $1.1mn to $11mnexpanded not sized

Incremental total $1.1mn to $11mnpoint $3.2mn
Cost displaced by AI$868k to $66mn sized

expanded $868k to $66mn

Incremental total $868k to $66mnpoint $9.8mn
Revenue arriving through AI$2.2mn to $134mn sized

new $2.2mn to $22mnexpanded $0 to $26mnrelabelled $0 to $86mn

Incremental total $2.2mn to $47mnpoint $13mn
Cost imposed, or revenue lost, by others’ AI$551k to $5.5mn sized

new $551k to $5.5mnexpanded not sized

Incremental total $551k to $5.5mnpoint $3.3mn

The sized total counts every channel the company credits to AI, including relabelled money that existed before language models and the ledger's own estimates for it. The incremental total counts relabelled channels at zero. A flow is split by layer where its dollars sit at more than one: end use, compute sold to builders, and hardware. The same dollar can be a buyer's spend, a cloud's revenue and a chipmaker's revenue, so the layers are never added together.

Paid for AI3 channels · $1.1mn to $11mn sized · $1.1mn to $11mn incremental · 2 not sized

vendor bill

AI model and token bill

0.02% to 0.25% of the quarter’s revenue

Incremental total: counts in full.

our inferencedisclosure: described· motive: exploratory· before LLMs: new

The prior quarter's statement that AI costs are increasing and token costs are expected to rise was not repeated; this quarter the CEO said only that token costs are being managed while access is broad, and that marketing's agentic systems use third-party technology. The filing again shows licence, maintenance and cloud costs rising inside technology and content without naming AI. The size shown is the ledger's own estimate, , built from the licence, maintenance and cloud line and the share Booking's CFO gave for the same kind of cost. The counterparty is mixed: model providers paid for tokens, AI software licensors and the cloud providers behind AI-specific capacity.

Evidence: 4 quotes, 4 figures, 2 confounds, 2 from before coverage

What Expedia pays outside vendors for AI: model usage (tokens), AI licences and AI-specific cloud, wherever it lands in cost of revenue or technology and content.

Why this motive

The CEO describes managing token costs while giving teams broad access (c11) and gives no level, share or line, and the bill is not split by use. AI spending named with no measure and no line moving is the exploratory tell for a passing mention, as in the prior quarter; the sources do not contradict each other.

Before LLMs: new

A token or model bill is not a cost the anchor reports. The FY2024 10-K says third-party AI technology was already in use, and the line that would hold the bill, the cloud, licensing and maintenance part of technology and content, was for FY2024 with no AI share named.

“We have incorporated third-party AI technology in certain of our products, services and business operations.”
Filing, risk factors, 10-K periodic report, 2025-02-07
“Technology and content expense includes product development and content expense, as well as information technology costs to support our infrastructure, back-office applications and overall monitoring and security of our networks, and is principally comprised of personnel and overhead, including stock-based compensation, as well as other costs including cloud expense and licensing and maintenance expense.”
Filing, mdna, 10-K periodic report, 2025-02-07

Figures

  • Technology and content: other costs (licences, maintenance, cloud) · 2026-CQ2
  • Technology and content: other costs, prior-year quarter · 2025-CQ2
  • Technology and content: other costs growth, year over year · 2026-CQ2
  • Technology and content: other costs, increase year over year (whole line, all causes) · 2026-CQ2

Reported line it is matched to

On the call the CEO said AI capabilities are being scaled with broad access for teams and token costs managed. The 10-Q shows the non-personnel part of technology and content, which holds licences, maintenance and cloud, at against a year earlier, a rise of , and names higher licence, maintenance and cloud costs as the cause without mentioning AI.

2026-CQ2: 2025-CQ2: 2026-CQ2: 2026-CQ2:

What else could explain it

  • bundling: The line holds every software licence, maintenance contract and cloud charge in technology and content, not only AI, and cloud costs for the websites also sit in cost of revenue.
  • other: Cloud and licence costs scale with traffic and bookings, which grew in the quarter.

Quotes

“As we scale these capabilities, we're managing token costs thoughtfully while giving broad access to our teams.”
c11 · CEO, prepared remarks, earnings call, 2026-08-05
“It's not just that measurement that allows us to better understand incrementality and returns. The team is also doing some really fantastic work around using technology to be more effective. They're developing these agentic systems to create personalized ads at scale, both using our technology and using third-party technology.”
c29 · CEO, qa, earnings call, 2026-08-05
“We also collaborate with leading AI platforms to enhance our capabilities and accelerate innovation.”
c41 · Filing, mdna, 10-Q periodic report, 2026-08-06
“Technology and content expense remained relatively consistent during the three and six months ended June 30, 2026, compared to the same periods in 2025 as higher license and maintenance costs and cloud costs were mostly offset by lower personnel costs in connection with previously announced cost saving initiatives.”
c44 · Filing, mdna, 10-Q periodic report, 2026-08-06

By quarter

  • Q1 2026direction only · our inference · exploratory · $970k to $9.7mn
  • Q2 2026described · our inference · exploratory · $1.1mn to $11mn

engineering

Hiring of AI skills

Not sized

The call, the release and the 10-Q say nothing about adding AI skills this quarter, and no headcount or cost was given when it was raised in the prior quarter.

inscrutabledisclosure: not mentioned· motive: exploratory· before LLMs: expanded

The prior CFO said the company would be adding back skills for its AI plans. The new CFO did not return to it, and nothing in the release or the 10-Q identifies AI hiring.

Evidence: 0 quotes, 1 figure, 1 confound, 4 from before coverage

Personnel cost added to carry out the AI plans: roles and skills management says it is adding back after earlier cost reductions.

Why this motive

No statement this quarter. The motive is carried from the prior quarter, when the CFO framed the hiring as skills needed for the AI plans.

Before LLMs: expanded

At the anchor this was technology and data science payroll inside technology and content, where personnel and overhead was for FY2024. The Q1 2024 call announced the cost actions that the later hiring is said to add back from. The size is the whole of an activity that existed before.

“Technology and content expense decreased $44 million for 2024 compared to 2023 primarily due to lower personnel costs in connection with previously announced cost saving initiatives, partially offset by higher stock-based compensation.”
Filing, mdna, 10-K periodic report, 2025-02-07
“During 2020, Expedia Group unified its technology, product, data engineering, and data science teams to build services and capabilities that can be leveraged across our business units to provide value-add services to our travel suppliers and serve our end customers.”
Filing, business, 10-K periodic report, 2025-02-07
“To that end, in February, we announced cost actions that will impact approximately 1,500 employees through this year.”
CFO, prepared remarks, earnings call, 2024-05-02
“As of December 31, 2024, approximately one half of our people work in technology roles.”
Filing, business, 10-K periodic report, 2025-02-07

Figures

  • Restructuring and related reorganization charges · 2026-CQ2

What else could explain it

  • transformation program: Restructuring charges continued at in the quarter, so the net movement in personnel cost cannot isolate AI hiring.

By quarter

  • Q1 2026described · described, no size · exploratory
  • Q2 2026not mentioned · inscrutable · exploratory

other

Acquisitions of AI products

Not sized

No purchase price is disclosed, the announcement came after the quarter ended, and the 10-Q does not mention the transaction. The quarter's own dollars are none, since no payment fell in it, and an undisclosed price for a private app has no reference class the ledger can defend. The ledger does not ballpark it.

inscrutabledisclosure: described· motive: exploratory· before LLMs: expanded

New this quarter: the CEO said the acquisition of Layla, an AI conversational planning app, was announced the week before the call. The release and the 10-Q are silent on it, so neither price nor timing of payment is known.

Evidence: 1 quote, 1 from before coverage

Capital paid to acquire AI-native products and teams, starting with the Layla conversational planning app.

Why this motive

The CEO frames the Layla acquisition as a way to capture new types of travelers and bring learnings into the core business (c7), a learning framing with no revenue attached.

Before LLMs: expanded

The FY2024 10-K treats acquisitions as a standing activity in its risk factors, and its FY2024 cash flow statement carries no acquisition line. No AI product had been acquired at the anchor. The size is the change AI made, not the whole line.

“We have acquired, invested in, divested or entered into significant commercial arrangements with a number of businesses in the past, and our future success may depend, in part, on such transactions, any of which could be material to our financial condition and results of operations.”
Filing, risk factors, 10-K periodic report, 2025-02-07

Quotes

“Last week, we announced the acquisition of Layla, an AI conversational planning app, allowing us to capture new types of travelers while bringing learnings into our core business.”
c7 · CEO, prepared remarks, earnings call, 2026-08-05

Cost displaced by AI6 channels · $868k to $66mn sized · $868k to $66mn incremental · 3 not sized

customer support · cheap to verify

Traveler support resolved by AI self-service

Not sized

The call is silent on self-service this quarter; the only remaining passage is the 10-Q’s one-clause aim of integrating AI into customer service operations (claim c40, now context), and a one-clause statement with no tool, deployment or measure is a context claim with no channel. The filing credits the line’s leverage to payments alone (claim c42).

inscrutabledisclosure: not mentioned· motive: efficiency· before LLMs: expanded

The call said nothing about traveler self-service this quarter, and the 10-Q’s line that AI is being integrated into customer service operations is a one-clause aim, kept as context. Cost of revenue moved from of revenue to , and the filing sentence that credited customer service efficiencies in the prior quarter now credits payments alone. The channel is read silent and unsized; the former estimate, carried on the prior quarter’s claims, is no longer used.

Evidence: 0 quotes, 4 figures, 1 confound, 4 from before coverage

Traveler service contacts resolved by the AI servicing agent and other AI-powered self-service instead of a human agent; the displaced cost sits in cost of revenue (customer support).

Why this motive

Carried from the prior quarter (efficiency); silence is not evidence about motive. The 10-Q now credits payments alone for the cost of revenue movement.

Before LLMs: expanded

The FY2024 10-K describes a virtual agent platform and AI-assisted self-service on the customer service platform, beside outsourced and in-house contact centers. The cost sat in cost of revenue, for FY2024, with no support share split out. The size is the change AI made, not the whole line.

“We provide 24-hour-a-day, seven-day-a-week traveler sales and support by our virtual agent platform, telephone, chat, or e-mail. For purposes of operational flexibility, we use a combination of outsourced and in-house contact centers.”
Filing, business, 10-K periodic report, 2025-02-07
“In addition, we have continued to invest in our customer service platform technology, which leverages technology and artificial intelligence to provide our customers with online customer service options and self-service capabilities.”
Filing, business, 10-K periodic report, 2025-02-07
“Cost of revenue primarily consists of direct costs to support our customer operations, including our customer support and telesales as well as fees to air ticket fulfillment vendors”
Filing, mdna, 10-K periodic report, 2025-02-07
“We've done a lot of experimentation in GenAI, obviously, user-facing as well as within the company from an efficiency standpoint, from customer service, all kinds of places.”
CEO, qa, earnings call, 2024-05-02

Figures

  • Cost of revenue growth, year over year · 2026-CQ2
  • Cost of revenue as a share of revenue · 2026-CQ2
  • Cost of revenue as a share of revenue, prior-year quarter · 2025-CQ2
  • Cost of revenue below what the prior-year share of revenue would give (whole line, all causes) · 2026-CQ2

What else could explain it

  • bundling: Cost of revenue fell against revenue again, but the 10-Q and the earnings presentation credit payments efficiencies only (c42, c37).

By quarter

  • Q1 2026direction only · shape match · efficiency
  • Q2 2026not mentioned · inscrutable · efficiency

customer support · cheap to verify

AI assistance for human support agents

Not sized

The call is silent on agent assistance this quarter; the only remaining passage is the 10-Q’s one-clause aim of integrating AI into customer service operations (claim c40, now context), and a one-clause statement with no tool, deployment or measure is a context claim with no channel. The filing credits the line’s leverage to payments alone (claim c42).

inscrutabledisclosure: not mentioned· motive: efficiency· before LLMs: expanded

The handling-time and onboarding claims of the prior quarter were not repeated, and the 10-Q’s line on AI in customer service operations is a one-clause aim, kept as context. Personnel and overhead inside cost of revenue was against a year earlier, and the filing attributes the line’s leverage to payments. The channel is read silent and unsized; the former estimate, carried on the prior quarter’s claims, is no longer used.

Evidence: 0 quotes, 2 figures, 1 confound, 4 from before coverage

AI used alongside human agents: shorter wait and handling time, automated conversation summaries across languages, faster onboarding of new agents; the displaced cost sits in cost of revenue (customer support).

Why this motive

Carried from the prior quarter (efficiency); silence is not evidence about motive. The 10-Q now credits payments alone for the cost of revenue movement.

Before LLMs: expanded

At the anchor this was investment in contact center technologies aimed at agent efficiency, and on the Q1 2024 call generative AI in the support organization was described as work still under way. The cost sat in cost of revenue, for FY2024. The size is the change AI made, not the whole line.

“We provide 24-hour-a-day, seven-day-a-week traveler sales and support by our virtual agent platform, telephone, chat, or e-mail. For purposes of operational flexibility, we use a combination of outsourced and in-house contact centers.”
Filing, business, 10-K periodic report, 2025-02-07
“We invested significantly in our contact center technologies, with the goal of improving customer experience and increasing the efficiency of our contact center agents and we expect to continue reaping the benefits of these investments going forward.”
Filing, business, 10-K periodic report, 2025-02-07
“Cost of revenue primarily consists of direct costs to support our customer operations, including our customer support and telesales as well as fees to air ticket fulfillment vendors”
Filing, mdna, 10-K periodic report, 2025-02-07
“You know, there's work, as Peter said, with our customer support organization, how do they use it to be more effective? Our development teams, even our commercial teams are looking at, you know, are there pilots for how we can use those to be more effective as well.”
CEO, qa, earnings call, 2024-05-02

Figures

  • Cost of revenue: personnel and overhead · 2026-CQ2
  • Cost of revenue: personnel and overhead, prior-year quarter · 2025-CQ2

What else could explain it

  • bundling: The 10-Q and the earnings presentation credit the cost of revenue movement to payments efficiencies only (c42, c37).

By quarter

  • Q1 2026direction only · our inference · efficiency · $0 to $16mn
  • Q2 2026not mentioned · inscrutable · efficiency

marketing · cheap to verify

AI-assisted allocation of marketing spend

Not sized

The CEO lists the team’s use of technology, including agentic ad systems, beside measurement and targeting as causes of the marketing leverage (claim c29), and the CFO credits reductions and removal of inefficient spend (claim c20); nothing separates AI’s part (the methodology rule for AI named beside another cause).

described, no sizedisclosure: described· motive: exploratory· before LLMs: expanded

The prior quarter’s order-of-magnitude claim for AI in marketing was not repeated, and neither was the CFO’s statement that AI helped the reallocation. The marketing leverage is now credited to measurement and targeting, with the CEO adding in Q&A that the team uses technology, including agentic ad systems, to be more effective. The filing line still moves the same way, below its prior-year share of bookings, a smaller gap than in the prior quarter. Nothing separates AI’s part, so the channel is left unsized.

Evidence: 10 quotes, 5 figures, 4 confounds, 2 from before coverage

AI used in targeting, measurement, testing and reallocation of paid marketing across channels and markets; the displaced cost is direct selling and marketing in the consumer business, paid to search engines and other traffic sources.

Why this motive

The displaced line grew while consumer bookings grew , so it shrank per booking dollar, but the CEO lists the team’s use of technology beside measurement as a cause (c29) and the CFO credits spend reductions (c20). A result credited to several changes together does not meet the efficiency tell for any one of them; AI named with no measure of its own is the exploratory tell.

Before LLMs: expanded

The FY2024 10-K describes marketing investment managed across the brand portfolio on a unified marketing technology platform to optimize spend, with no AI label. Consumer direct selling and marketing, the line the saving sits in, was for FY2024. The size is the change AI made, not the whole line.

“For example, we now manage our marketing investments holistically across the entire brand portfolio, allowing us to optimize our spend to achieve better returns, and run on a unified marketing technology platform, improving our performance by scaling our marketing capabilities.”
Filing, business, 10-K periodic report, 2025-02-07
“Selling and marketing - direct costs primarily include traffic generation costs from search engines and internet portals, television and print spending, private label and affiliate program commissions, public relations and other costs.”
Filing, mdna, 10-K periodic report, 2025-02-07

Figures

  • B2C selling and marketing - direct, change year over year · 2026-CQ2
  • B2C gross bookings growth, year over year · 2026-CQ2
  • B2C selling and marketing - direct below what the prior-year share of B2C gross bookings would give (whole line, all causes) · 2026-CQ2
  • Selling and marketing - direct, all segments · 2026-CQ2
  • Selling and marketing - direct, all segments, prior-year quarter · 2025-CQ2

Reported line it is matched to

The call says measurement, targeting and the team's use of technology are producing marketing leverage. The 10-Q segment table shows consumer direct selling and marketing at against a year earlier while consumer gross bookings rose to from , taking the line from of bookings to . The 10-Q and the earnings presentation call this marketing leverage and discipline and do not mention AI.

2026-CQ2: 2025-CQ2: 2026-CQ2: 2026-CQ2: 2025-CQ2: 2026-CQ2: 2026-CQ2: 2025-CQ2: 2026-CQ2:

What else could explain it

  • transformation program: The CFO credits the past year's marketing contribution to substantial reductions in spend and removal of inefficient spend (c20), which will be lapped in the second half. It is not attributed to AI.
  • other: Partner-funded offers were in more than of Vrbo bookings (c3), and loyalty and a healthy U.S. market also bring bookings without paid traffic.
  • fx: The CFO said foreign exchange added slightly to reported bookings growth (c12).
  • other: The CEO credits measurement and targeting beside the team’s use of technology for the same leverage (claim c29), and gives no AI share.

Quotes

“More than 40% of Vrbo bookings last quarter included partner-funded offers, and our May sale was the first campaign to exceed $1 billion in bookings for participating properties.”
c3 · CEO, prepared remarks, earnings call, 2026-08-05
“In our consumer business, our sharper measurement and targeting capabilities are allowing us to invest in the highest return opportunities and leverage our marketing spend. We're investing in newer surfaces like AI and social platforms, where more consumers are starting their trip planning.”
c4 · CEO, prepared remarks, earnings call, 2026-08-05
“We expanded margins by nearly two points in the quarter, driven by tight expense management and the consumer marketing leverage I just mentioned.”
c8 · CEO, prepared remarks, earnings call, 2026-08-05
“Foreign exchange was a tailwind for the business, contributing nearly half a point to bookings growth four points to revenue growth.”
c12 · CFO, prepared remarks, earnings call, 2026-08-05
“The margin expansion was driven by a combination of cost efficiencies, consumer marketing leverage, the flow-through of higher volume.”
c13 · CFO, prepared remarks, earnings call, 2026-08-05
“We had the substantial reductions in marketing spend and driving out inefficient spend, but also redirecting spend to more productive channels.”
c20 · CFO, qa, earnings call, 2026-08-05
“We grew our bookings 8% while leveraging marketing spend, which was only up 1%.”
c28 · CEO, qa, earnings call, 2026-08-05
“It's not just that measurement that allows us to better understand incrementality and returns. The team is also doing some really fantastic work around using technology to be more effective. They're developing these agentic systems to create personalized ads at scale, both using our technology and using third-party technology.”
c29 · CEO, qa, earnings call, 2026-08-05
“B2C leverage driven by marketing discipline and improved returns across channels”
c36 · Filing, press release, 8-K earnings release, 2026-08-05
“Selling and marketing - direct increased $199 million and $298 million during the three and six months ended June 30, 2026, compared to the same periods in 2025, primarily driven by an increase in B2B partner commissions to support revenue growth. The increase in the year-to-date period was partially offset by a decrease in marketing spend at B2C, which had significant marketing leverage.”
c43 · Filing, mdna, 10-Q periodic report, 2026-08-06

By quarter

  • Q1 2026bounded · described, no size · exploratory
  • Q2 2026described · described, no size · exploratory

marketing · cheap to verify

AI in marketing creative and workflow

0.01% to 0.77% of the quarter’s revenue

Incremental total: counts in full.

our inferencedisclosure: described· motive: exploratory· before LLMs: expanded

The CEO said the marketing team is developing agentic systems to create personalized ads at scale on in-house and third-party technology. This is more specific than the prior quarter's remark about AI improving creative and still carries no number. The size shown is the ledger's own estimate, , built from consumer paid media, an industry production-to-media ratio and an assumed share of creative cost displaced by AI. The channel this reading was counted inside (marketing-spend-efficiency) is unsized this quarter, so the overlap no longer holds and this reading counts in totals on its own evidence (overlap rule, 2026-10-06).

Evidence: 1 quote, 2 from before coverage

AI used to produce and vary marketing creative and to automate marketing workflows; the displaced cost is creative production and marketing personnel time.

Why this motive

The CEO offers agentic systems that create personalized ads as part of the marketing leverage (c29), but no creative cost line is visible and indirect selling and marketing was relatively consistent. AI named with no measure and no line moving is the exploratory tell; the sources do not contradict each other.

Before LLMs: expanded

The FY2024 10-K shows brand advertising and creative campaigns, with creative production work being built in-house. No creative cost is split out of consumer direct selling and marketing, for FY2024, or of marketing personnel. The size is the change AI made, not the whole line.

“We also license content from suppliers for our creative campaigns. As we continue to build our creative production work and introduce new services that incorporate new technologies, content and brands, we may be required to license additional technology, content, and brands.”
Filing, risk factors, 10-K periodic report, 2025-02-07
“Our marketing channels primarily include brand advertising through online and offline channels, loyalty programs, mobile apps, search engine marketing and optimization as well as metasearch, social media, direct and/or personalized traveler communications on our websites as well as through direct e-mail communication with our travelers.”
Filing, business, 10-K periodic report, 2025-02-07

Quotes

“It's not just that measurement that allows us to better understand incrementality and returns. The team is also doing some really fantastic work around using technology to be more effective. They're developing these agentic systems to create personalized ads at scale, both using our technology and using third-party technology.”
c29 · CEO, qa, earnings call, 2026-08-05

By quarter

  • Q1 2026described · our inference · exploratory · $414k to $31mn
  • Q2 2026described · our inference · exploratory · $440k to $33mn

engineering · cheap to verify

Internal AI adoption (engineering and internal processes)

0.01% to 0.46% of the quarter’s revenue

Incremental total: counts in full.

our inferencedisclosure: direction only· motive: unknown· before LLMs: expanded

The CEO said the agentic technology stack lets teams design and ship products faster and that AI use in the technology teams shortens the development cycle; the release calls AI a force multiplier for operating more efficiently. That is a claimed result where the prior quarter had only adoption. The filing shows the personnel line falling and gives the cost program as the reason, so the line has the shape the claim predicts and the filing credits no saving to AI. The size shown is the ledger's own estimate, , built from technology and content personnel cost, a productivity gain taken from public developer studies and an assumed share of the gain that reaches the cost line.

Evidence: 13 quotes, 4 figures, 2 confounds, 6 from before coverage

Employees' own use of AI tools in engineering and other internal processes, with operational efficiency as the stated goal; the displaced cost would be personnel in technology and content and overhead lines.

Why this motive

Management now claims a result, shipping faster and operating more efficiently (c10, c33), and the personnel line fell, but the filing attributes the fall to previously announced cost saving initiatives (c44). The tells conflict and none separates AI from the cost program.

Before LLMs: expanded

At the anchor this was technology payroll: technology and content personnel and overhead was for FY2024, already falling on the cost actions announced that year. The Q1 2024 call described generative AI use by development and commercial teams as pilots. The size is the change AI made, not the whole line.

“You know, there's work, as Peter said, with our customer support organization, how do they use it to be more effective? Our development teams, even our commercial teams are looking at, you know, are there pilots for how we can use those to be more effective as well.”
CEO, qa, earnings call, 2024-05-02
“We've done a lot of experimentation in GenAI, obviously, user-facing as well as within the company from an efficiency standpoint, from customer service, all kinds of places.”
CEO, qa, earnings call, 2024-05-02
“Technology and content expense decreased $44 million for 2024 compared to 2023 primarily due to lower personnel costs in connection with previously announced cost saving initiatives, partially offset by higher stock-based compensation.”
Filing, mdna, 10-K periodic report, 2025-02-07
“To that end, in February, we announced cost actions that will impact approximately 1,500 employees through this year.”
CFO, prepared remarks, earnings call, 2024-05-02
“As of December 31, 2024, approximately one half of our people work in technology roles.”
Filing, business, 10-K periodic report, 2025-02-07
“Technology and content expense includes product development and content expense, as well as information technology costs to support our infrastructure, back-office applications and overall monitoring and security of our networks, and is principally comprised of personnel and overhead, including stock-based compensation, as well as other costs including cloud expense and licensing and maintenance expense.”
Filing, mdna, 10-K periodic report, 2025-02-07

Figures

  • Technology and content: personnel and overhead, change year over year · 2026-CQ2
  • Restructuring and related reorganization charges · 2026-CQ2
  • Total adjusted overhead expenses (indirect selling and marketing, technology and content, general and administrative, excluding stock-based compensation) · 2026-CQ2
  • Total adjusted overhead expenses, prior-year quarter · 2025-CQ2

Reported line it is matched to

The release and the call say AI is letting teams ship faster and operate more efficiently. The 10-Q shows technology and content personnel cost at against a year earlier, a change of , and the release shows adjusted overhead at against while revenue grew . The 10-Q attributes the personnel decline to previously announced cost saving initiatives and does not mention AI.

2026-CQ2: 2025-CQ2: 2026-CQ2: 2026-CQ2: 2025-CQ2: 2026-CQ2:

What else could explain it

  • transformation program: The 10-Q attributes lower technology and content personnel cost to previously announced cost saving initiatives (c44), the earnings presentation credits overhead leverage to a series of cost reductions (c38), and restructuring charges of continued in the quarter. None of this is attributed to AI and none of it is credited to this channel.
  • fx: The CFO said foreign exchange added to reported revenue growth (c12), which flatters overhead measured against revenue.

Quotes

“We expanded margins by nearly two points in the quarter, driven by tight expense management and the consumer marketing leverage I just mentioned.”
c8 · CEO, prepared remarks, earnings call, 2026-08-05
“Importantly, we're continuing to deploy AI to innovate faster and operate more effectively. On Vrbo, we launched an agentic voice solution to support partner inquiries previously handled by human agents. Early results are promising, with faster resolution and lower contact propensity.”
c9 · CEO, prepared remarks, earnings call, 2026-08-05
“More broadly, our agentic technology stack is allowing us to design and ship products faster, ultimately unlocking new capabilities for both travelers and partners.”
c10 · CEO, prepared remarks, earnings call, 2026-08-05
“As we scale these capabilities, we're managing token costs thoughtfully while giving broad access to our teams.”
c11 · CEO, prepared remarks, earnings call, 2026-08-05
“Foreign exchange was a tailwind for the business, contributing nearly half a point to bookings growth four points to revenue growth.”
c12 · CFO, prepared remarks, earnings call, 2026-08-05
“The margin expansion was driven by a combination of cost efficiencies, consumer marketing leverage, the flow-through of higher volume.”
c13 · CFO, prepared remarks, earnings call, 2026-08-05
“In addition, I would say, just as an aside, is all the work we're doing in using AI in our technology teams in order to increase our cycle time is allowing us to innovate a lot faster.”
c15 · CEO, qa, earnings call, 2026-08-05
“We've also continued to make incremental progress on the rest of the cost structure, and I'd just point to the fact that overheads were flat year-over-year in the most recent quarter, even as revenue rose 14%.”
c21 · CFO, qa, earnings call, 2026-08-05
“We continued to strengthen our marketplace through more personalized consumer product experiences and expanded supply across our business, while leveraging AI as a force multiplier to innovate faster and operate more efficiently.”
c33 · CEO, press release, 8-K earnings release, 2026-08-05
“Leverage benefited from ongoing savings generated by a series of cost reductions”
c38 · Filing, press release, 8-K earnings release, 2026-08-05
“We continue to integrate artificial intelligence across our platform, including in traveler-facing experiences, customer service operations and internal processes, with the goal of improving conversion, personalization and operational efficiency.”
c40 · Filing, mdna, 10-Q periodic report, 2026-08-06
“Technology and content expense remained relatively consistent during the three and six months ended June 30, 2026, compared to the same periods in 2025 as higher license and maintenance costs and cloud costs were mostly offset by lower personnel costs in connection with previously announced cost saving initiatives.”
c44 · Filing, mdna, 10-Q periodic report, 2026-08-06
“We have continued to recalibrate resources and expand the restructure efforts that began in 2024 due to the significant completion of the Company’s organizational and technological transformation.”
c45 · Filing, mdna, 10-Q periodic report, 2026-08-06

By quarter

  • Q1 2026described · our inference · exploratory · $318k to $20mn
  • Q2 2026direction only · our inference · unknown · $307k to $20mn

partner support · cheap to verify

Partner support handled by AI agents

0% to 0.3% of the quarter’s revenue

Incremental total: counts in full.

our inferencedisclosure: direction only· motive: exploratory· before LLMs: expanded

New this quarter: on Vrbo an agentic voice solution now supports partner inquiries that human agents handled before, with early results described as faster resolution and lower contact propensity. Nothing in the release or the 10-Q refers to it. The size shown is the ledger's own estimate, , built from an assumed partner support share of cost of revenue, an assumed share of inquiries handled by the voice solution and an assumed saving per handled inquiry. The channel this reading was counted inside (traveler-self-service) is unsized this quarter, so the overlap no longer holds and this reading counts in totals on its own evidence (overlap rule, 2026-10-06).

Evidence: 1 quote, 2 from before coverage

Supply partner inquiries handled by AI agents instead of human agents, starting with an agentic voice solution for Vrbo partners; the displaced cost is partner support labor.

Why this motive

The CEO places the agentic voice solution under operating efficiency and says it handles inquiries human agents handled before, which points to efficiency, but she calls the results early (c9) and no displaced line is visible. The tells conflict and the less durable motive is chosen.

Before LLMs: expanded

The anchor does not describe partner support staffing by name. It shows account and market managers working directly with suppliers and supplier operations among the costs in cost of revenue, for FY2024. The size is the change AI made, not the whole line.

“other costs, primarily including data center and cloud costs to support our websites, supplier operations, destination supply, certain transactional level taxes as well as related personnel and overhead costs”
Filing, mdna, 10-K periodic report, 2025-02-07
“Our strategic account managers and local hotel market managers work directly with travel suppliers to optimize the exposure of their travel products and brands through our points of sale”
Filing, business, 10-K periodic report, 2025-02-07

Quotes

“Importantly, we're continuing to deploy AI to innovate faster and operate more effectively. On Vrbo, we launched an agentic voice solution to support partner inquiries previously handled by human agents. Early results are promising, with faster resolution and lower contact propensity.”
c9 · CEO, prepared remarks, earnings call, 2026-08-05

Revenue arriving through AI6 channels · $2.2mn to $134mn sized · $2.2mn to $47mn incremental · 3 not sized

partner support · cheap to verify

AI in onboarding lodging partners

Not sized

The quarter's sources report supply growth and say nothing about AI in partner onboarding.

inscrutabledisclosure: not mentioned· motive: exploratory· before LLMs: relabelled

The 10-Q reports approximately lodging properties and the CEO spoke of expanded supply, with no mention of AI in onboarding. The claim made in the prior quarter was not repeated.

Evidence: 0 quotes, 1 figure, 2 from before coverage

AI used to bring lodging properties onto the marketplace faster; the money arrives as bookings on supply added sooner.

Why this motive

No statement this quarter; carried from the prior quarter, where faster AI onboarding was a passing mention with no measure (exploratory). Silence is not evidence about motive.

Before LLMs: relabelled

Supplier onboarding and direct work with suppliers are in the FY2024 10-K with no AI label; the marketplace held over lodging properties at the end of 2024.

“Our strategic account managers and local hotel market managers work directly with travel suppliers to optimize the exposure of their travel products and brands through our points of sale”
Filing, business, 10-K periodic report, 2025-02-07
“For example, when onboarding suppliers to our websites, we may fail to identify falsified or stolen supplier credentials, which may result in fraudulent bookings or unauthorized access to personal or confidential information of users of our websites and mobile applications.”
Filing, risk factors, 10-K periodic report, 2025-02-07

Figures

  • Lodging properties on the marketplace, approximate · as-of 2026-06-30

By quarter

  • Q1 2026described · described, no size · exploratory
  • Q2 2026not mentioned · inscrutable · exploratory

other · cheap to verify

AI enrichment of property content

Not sized

The quarter's sources say nothing about AI enrichment of property content.

inscrutabledisclosure: not mentioned· motive: product-defensive· before LLMs: relabelled

Content came up only as something the brands need to be complete in for answer engines; AI enrichment of supply content was not mentioned.

Evidence: 0 quotes, 3 from before coverage

Withdrawn from the ledger on 2026-10-06 as not an AI channel, under the rule that a one-clause statement with no tool, deployment or measure is a context claim with no channel. Registered in Q1 2026 as conversion lift from AI-enriched property content, on one clause in the prepared remarks; the passage is kept as context.

Why this motive

No statement this quarter; the motive is carried from the prior quarter, when content enrichment was framed as underpinning traveler trust.

Before LLMs: relabelled

At the anchor, property content was a cost inside technology and content and something Vrbo partners edited with listing tools; on the Q1 2024 call generative AI for showing partner inventory was a question still being worked on. No content line is split out.

“Vrbo's alternative accommodation listing services offers our supply partners a set of tools to help them manage an availability calendar, respond to customer inquiries, edit the content of a property listing, as well as provide various other services to manage reservations or drive incremental sales volume.”
Filing, business, 10-K periodic report, 2025-02-07
“How do we help partners use GenAI to better, you know, show their inventory in our apps or in our brands? How do we allow them to use GenAI to improve their advertising with us?”
CEO, qa, earnings call, 2024-05-02
“Technology and content expense includes product development and content expense, as well as information technology costs to support our infrastructure, back-office applications and overall monitoring and security of our networks, and is principally comprised of personnel and overhead, including stock-based compensation, as well as other costs including cloud expense and licensing and maintenance expense.”
Filing, mdna, 10-K periodic report, 2025-02-07

By quarter

  • Q1 2026not mentioned · inscrutable · product-defensive
  • Q2 2026not mentioned · inscrutable · product-defensive

product ranking · cheap to verify

AI ranking, recommendation and personalization

0% to 2% of the quarter’s revenue

Incremental total: counts at zero.

our inferencedisclosure: direction only· motive: offensive· before LLMs: relabelled

The CEO again tied record attach rates on Expedia to AI personalization and said recommendations and ranking are improving conversion now. The line that holds attached products grew faster than total revenue for a second quarter, and this quarter the filing adds activities revenue, which includes an acquired business, to its explanation. No attach rate, conversion rate or revenue amount is attributed to AI. The size shown is the ledger's own estimate, , built as revenue times an assumed exposed share times an assumed conversion lift.

Evidence: 10 quotes, 1 figure, 5 confounds, 4 from before coverage

Ranking and recommendation models trained on traveler interactions; the money arrives as conversion and attach-rate lift on the consumer brands.

Why this motive

Management attributes record attach rates and improving conversion to AI personalization and recommendations (c2, c14), which is the offensive tell. No rate is given, so the attribution rests on management's word.

Before LLMs: relabelled

On the Q1 2024 call management credited machine learning with product and conversion wins and with choosing what to attach, and the FY2024 10-K credits machine learning and artificial intelligence for the improved product. Other revenue, the line that holds attached products, was for FY2024.

“There's also a lot of impact, as we've talked about before, from old-fashioned machine learning, driving all kinds of wins across the product, and as I mentioned, much more at scale because they can be deployed much more readily across all brands and across different lines of business.”
CEO, qa, earnings call, 2024-05-02
“It is true that with machine learning, you can get a lot smarter in understanding what is the next best thing to propose to a traveler. What are they most likely to attach, given what we know about them, given what we know about what they're doing in that trip plan?”
CEO, qa, earnings call, 2024-05-02
“For example, a recommendation algorithm gets smarter faster because of our scale, but it has to be trained on the differences between a traveler shopping on Vrbo compared to one on Expedia.”
CEO, prepared remarks, earnings call, 2024-05-02
“With greatly improved product driven by the latest in machine learning and artificial intelligence capabilities, we believe we will continue to drive greater retention, repeat and direct business.”
Filing, business, 10-K periodic report, 2025-02-07

Figures

  • Other revenue growth, year over year · 2026-CQ2

Reported line it is matched to

The call says AI personalization produced another quarter of record attach rates on Expedia. The 10-Q shows other revenue, which holds insurance, car, cruise and activities, at against a year earlier, up against total revenue growth of , and credits higher insurance and activities revenue without mentioning AI.

2026-CQ2: 2025-CQ2: 2026-CQ2: 2026-CQ2:

What else could explain it

  • acquisition: The Tiqets acquisition was absorbed in the quarter (c32) and the 10-Q credits activities revenue alongside insurance, so part of the line's growth is acquired.
  • other: The CEO credits design and placement of attach messaging alongside recommendations (c31).
  • other: At Vrbo, partner-funded offers were in more than of bookings (c3), a non-AI driver of conversion.
  • fx: The CFO said foreign exchange added to reported revenue growth (c12).
  • bundling: The other revenue line also holds car, cruise and activities, and attach can rise from product and pricing changes unrelated to ranking models.

Quotes

“Our AI-powered personalization and recommendations keep getting smarter across all three of our consumer brands. On Expedia, our fastest-growing brand, this translated into another quarter of record attach rates.”
c2 · CEO, prepared remarks, earnings call, 2026-08-05
“More than 40% of Vrbo bookings last quarter included partner-funded offers, and our May sale was the first campaign to exceed $1 billion in bookings for participating properties.”
c3 · CEO, prepared remarks, earnings call, 2026-08-05
“Foreign exchange was a tailwind for the business, contributing nearly half a point to bookings growth four points to revenue growth.”
c12 · CFO, prepared remarks, earnings call, 2026-08-05
“In that first bucket, it's using AI for better recommendations, for better ranking, for personalizing the UX and the content. There, we are seeing immediate impacts, and it's not only just improving conversion, but you can imagine, as we're getting traffic, whether it's direct traffic or paid traffic, that is improving.”
c14 · CEO, qa, earnings call, 2026-08-05
“Actually, part of what's behind the record attach rates is the personalization we're able to drive by understanding what is the next best thing to recommend to a traveler.”
c30 · CEO, qa, earnings call, 2026-08-05
“In addition, they've done a great job really optimizing the UX and the design, figuring out where is it that we put the attach messaging, where is it in the app, where is it in the post-sale communications. It's really a combination of all of the above.”
c31 · CEO, qa, earnings call, 2026-08-05
“One aspect of that is that we had the digestion of Tiqets and that acquisition in Q2, that would have weighed a little bit on the cost that you're seeing.”
c32 · CFO, qa, earnings call, 2026-08-05
“We continued to strengthen our marketplace through more personalized consumer product experiences and expanded supply across our business, while leveraging AI as a force multiplier to innovate faster and operate more efficiently.”
c33 · CEO, press release, 8-K earnings release, 2026-08-05
“We continue to integrate artificial intelligence across our platform, including in traveler-facing experiences, customer service operations and internal processes, with the goal of improving conversion, personalization and operational efficiency.”
c40 · Filing, mdna, 10-Q periodic report, 2026-08-06
“All other revenue, which includes insurance, car, cruise and activities, increased 23% and 20% for the three and six months ended June 30, 2026, compared to the same periods in 2025 primarily due to higher insurance and activities revenue.”
c46 · Filing, mdna, 10-Q periodic report, 2026-08-06

By quarter

  • Q1 2026direction only · our inference · offensive · $0 to $69mn
  • Q2 2026direction only · our inference · offensive · $0 to $86mn

search discovery · cheap to verify

Traveler-facing AI search features

0% to 0.6% of the quarter’s revenue

Incremental total: counts in full.

our inferencedisclosure: described· motive: exploratory· before LLMs: expanded

The company launched natural language search on the Vrbo homepage and updated Property Expert and AI Compare. In Q&A the CEO said these are not driving conversion right now and yield over more information about traveler intent. The prior quarter's statement that AI filter users convert at higher rates was not repeated. The size shown is the ledger's own estimate, , built as revenue times an assumed exposed share times a conversion lift set near zero on management's own words.

Evidence: 4 quotes, 1 figure, 2 from before coverage

AI-powered filters and conversational shopping experiences inside the consumer apps and sites; the money arrives as repeat visits and conversion among travelers who use them.

Why this motive

Management tags the features itself: the natural language search and agents are not driving conversion right now, are valued for what they reveal about traveler intent, and are expected to pay off over time (c47, c16).

Before LLMs: expanded

At the anchor travelers shopped through ordinary search and filters, with machine learning already in the product. The Q1 2024 call describes generative AI search as an early experiment with modest use and little effect on conversion; no line holds it. The size is the change AI made, not the whole line.

“We've done a lot of experimentation in GenAI, obviously, user-facing as well as within the company from an efficiency standpoint, from customer service, all kinds of places.”
CEO, qa, earnings call, 2024-05-02
“But basically, you know, it is early days as far as the search approach goes. It's still pretty modest in terms of how many people use it, certainly in terms of its impact on conversion or anything else.”
CEO, qa, earnings call, 2024-05-02

Figures

  • Additional traveler-intent information from natural language features, lower bound · 2026-CQ2

Quotes

“AI is unlocking new ways to make these experiences simpler and more personalized. Last quarter, we launched new and updated conversational experiences, introducing natural language search on the Vrbo homepage and updating Property Expert and AI Compare in the hotel shopping flow.”
c1 · CEO, prepared remarks, earnings call, 2026-08-05
“The second is, what are we doing in our product that's not necessarily delivering conversion right now, but we know is helping us better understand travelers and is going to have compounding benefits over time?”
c47 · CEO, qa, earnings call, 2026-08-05
“Also in the product, we're introducing these natural language experiences like Vrbo natural language search on the homepage, or these agents like Property Expert or AI Compare. Those are not driving conversion right now, but what we're finding is that you get over 60% more information about traveler intent, and that allows us to deepen the relationships with the traveler.”
c16 · CEO, qa, earnings call, 2026-08-05
“We continue to integrate artificial intelligence across our platform, including in traveler-facing experiences, customer service operations and internal processes, with the goal of improving conversion, personalization and operational efficiency.”
c40 · Filing, mdna, 10-Q periodic report, 2026-08-06

By quarter

  • Q1 2026direction only · our inference · offensive · $0 to $34mn
  • Q2 2026described · our inference · exploratory · $0 to $26mn

other · cheap to verify

AI-powered partner insights, advertising and promotion tools

Not sized

The CFO discussed advertising growth without any reference to AI, and the 10-Q credits sponsored listings.

inscrutabledisclosure: not mentioned· motive: exploratory· before LLMs: relabelled

EG Advertising revenue was against a year earlier, up . Asked about advertising, the CFO spoke of geographic reach and new surfaces and did not mention AI. The prior quarter's claim about AI-powered partner tools was not repeated.

Evidence: 0 quotes, 3 figures, 3 from before coverage

Withdrawn from the ledger on 2026-10-06 as not an AI channel, under the rule that a one-clause statement with no tool, deployment or measure is a context claim with no channel. Registered in Q1 2026 as advertising revenue from AI-derived partner insight and promotional tools, on one clause in the prepared remarks; the 10-Q credits advertising growth to sponsored listings. The passages are kept as context.

Why this motive

No statement this quarter; carried from the prior quarter, where AI in partner advertising tools was a passing mention with no measure (exploratory). Silence is not evidence about motive.

Before LLMs: relabelled

The FY2024 10-K lists marketing tools and a revenue management product giving pricing insights to lodging partners, and reports Expedia Group Media Solutions advertising revenue of for FY2024. On the Q1 2024 call generative AI for partner advertising was still a question.

“How do we help partners use GenAI to better, you know, show their inventory in our apps or in our brands? How do we allow them to use GenAI to improve their advertising with us?”
CEO, qa, earnings call, 2024-05-02
“Marketing tools assist hotels and alternative accommodations in tailoring demand to their requirements and our revenue management product provides pricing insights.”
Filing, business, 10-K periodic report, 2025-02-07
“In 2024, we generated $639 million of advertising revenue from Expedia Group Media Solutions, a 32% increase from 2023.”
Filing, mdna, 10-K periodic report, 2025-02-07

Figures

  • EG Advertising revenue · 2026-CQ2
  • EG Advertising revenue, prior-year quarter · 2025-CQ2
  • EG Advertising revenue growth, year over year · 2026-CQ2

By quarter

  • Q1 2026not mentioned · inscrutable · exploratory
  • Q2 2026not mentioned · inscrutable · exploratory

distribution · cheap to verify

Traffic and bookings arriving from AI platforms

0.05% to 0.5% of the quarter’s revenue

Incremental total: counts in full.

our inferencedisclosure: direction only· motive: exploratory· before LLMs: new

Answer engine optimization is now described as one of the fastest-growing channels, alongside social, where the prior quarter called it the fastest. Agentic traffic remains small. The CEO placed AI experiences inside the part of consumer bookings that comes through paid channels and said it is too early to say who is getting more or less from the channel. The size shown is the ledger's own estimate, , built from revenue and the share Booking's CFO gave for traffic from large language models, the peer reference class.

Evidence: 10 quotes

Bookings that start on an AI platform (ChatGPT, Gemini, Claude and similar) and land on an Expedia Group brand through answer engine optimization, integrations or ads.

Why this motive

Management again calls it early days and describes testing and participating everywhere (c18, c19). It also describes making sure the brands show up, which is the channel-defensive tell; the less durable motive is chosen.

Before LLMs: new

The anchor reports no traffic or bookings from AI platforms. Its indirect channels are search engines, metasearch, social media and affiliates.

Quotes

“AEO and social are two of our fastest-growing channels. While agentic traffic remains small, it's a promising channel with high purchase consideration.”
c5 · CEO, prepared remarks, earnings call, 2026-08-05
“We're deepening our partnerships with leading AI platforms. During the quarter, we're an early adopter of ChatGPT's latest ad product and expanded our work across Google's newer AI services.”
c6 · CEO, prepared remarks, earnings call, 2026-08-05
“I also believe there's a big growth opportunity in getting access to travelers who are starting outside of our brands in these AI experiences. As a reminder, two-thirds of our bookings in our consumer brands come direct.”
c17 · CEO, qa, earnings call, 2026-08-05
“Of the third that is coming through paid channels, obviously AI experiences, whether they are with ChatGPT or Claude or Google's new experiences, are new opportunities for our brands to show up there. It is early days.”
c18 · CEO, qa, earnings call, 2026-08-05
“We are testing and participating everywhere that things are evolving, and I see that over time as opportunities to bring more travelers into our business.”
c19 · CEO, qa, earnings call, 2026-08-05
“When I think about organic traffic, and I'm going to bucket together SEO and AEO, for us, organic search traffic is stable to slightly up. Now, as I mentioned in my prepared remarks, AEO is one of our fastest-growing channels, and SEO has remained a bit soft, but it's stabilized over the last few quarters, and I think it's due to the great work that our team has done.”
c22 · CEO, qa, earnings call, 2026-08-05
“A number of quarters ago, we organized a small team to look at organic across the board, AEO and SEO, and they've increased testing velocity, both technical and content. As I said, they've been doing more tests. They've been using AI across the board in order to improve our performance in that channel.”
c23 · CEO, qa, earnings call, 2026-08-05
“Look, AEO is still a small channel. As I said, it's one of our fastest-growing channels.”
c25 · CEO, qa, earnings call, 2026-08-05
“Are there opportunities with these connectors or micro apps, for example, in Claude and ChatGPT? It's fascinating to see there are some cases in which we can control a bit more what the interface is, where our brand shows up, and others where we can't.”
c26 · CEO, qa, earnings call, 2026-08-05
“I think it's too early to declare anyone's getting a lot more versus less.”
c27 · CEO, qa, earnings call, 2026-08-05

By quarter

  • Q1 2026direction only · our inference · exploratory · $1.7mn to $17mn
  • Q2 2026direction only · our inference · exploratory · $2.2mn to $22mn

Cost imposed, or revenue lost, by others’ AI3 channels · $551k to $5.5mn sized · $551k to $5.5mn incremental · 2 not sized

distribution

Paid placement on AI platforms

0.01% to 0.13% of the quarter’s revenue

Incremental total: counts in full.

our inferencedisclosure: described· motive: channel-defensive· before LLMs: new

The company was an early adopter of ChatGPT's latest ad product in the quarter and expanded its work across Google's newer AI services. No spend, pricing or return is given. The size shown is the ledger's own estimate, , built as a share of consumer direct selling and marketing capped by the peer bound on traffic from AI platforms.

Evidence: 4 quotes

Fees paid to appear on AI platforms, starting with ChatGPT ads; the cost sits inside direct selling and marketing.

Why this motive

The CEO describes investing in AI platforms because that is where more consumers are starting their trip planning (c4), which is the channel-defensive tell.

Before LLMs: new

The anchor names no payment to an AI platform. Paid traffic was bought from search engines, internet portals and metasearch inside consumer direct selling and marketing, for FY2024.

Quotes

“In our consumer business, our sharper measurement and targeting capabilities are allowing us to invest in the highest return opportunities and leverage our marketing spend. We're investing in newer surfaces like AI and social platforms, where more consumers are starting their trip planning.”
c4 · CEO, prepared remarks, earnings call, 2026-08-05
“We're deepening our partnerships with leading AI platforms. During the quarter, we're an early adopter of ChatGPT's latest ad product and expanded our work across Google's newer AI services.”
c6 · CEO, prepared remarks, earnings call, 2026-08-05
“Of the third that is coming through paid channels, obviously AI experiences, whether they are with ChatGPT or Claude or Google's new experiences, are new opportunities for our brands to show up there. It is early days.”
c18 · CEO, qa, earnings call, 2026-08-05
“We are testing and participating everywhere that things are evolving, and I see that over time as opportunities to bring more travelers into our business.”
c19 · CEO, qa, earnings call, 2026-08-05

By quarter

  • Q1 2026described · our inference · channel-defensive · $518k to $5.2mn
  • Q2 2026described · our inference · channel-defensive · $551k to $5.5mn

distribution · expensive to verify

Competition from AI planning and booking products

Not sized

The filing and release name the competition and quantify nothing: no lost traffic, share or cost is disclosed. No reference class exists for demand that never arrived: the channel is a counterfactual against which nothing is measured, and the CEO reports organic traffic stable to slightly up, so no observed loss can be decomposed. The ledger does not ballpark it.

inscrutabledisclosure: described· motive: imposed· before LLMs: new

The 10-Q and the release repeat the prior quarter's language on AI-powered competition. On the call the CEO said the company controls the interface and brand placement on AI platforms in some cases and not in others. No traffic loss or cost is reported.

Evidence: 3 quotes, 2 from before coverage

Demand that other companies' AI planning, assistance or checkout products could take from Expedia Group's own sites and apps, or make more costly to reach.

Why this motive

The 10-Q names generative AI planning and assistance tools as competing offerings (c39) and the release lists AI-powered platforms among competitors (c34); the company did not choose this cost.

Before LLMs: new

The FY2024 10-K already names generative AI digital assistants as possible competing offerings, beside the older risk of search engines disintermediating online travel agencies. No lost demand is reported at the anchor.

“In addition, technological developments in generative artificial intelligence ("AI") tools may be increasingly used to create competing offerings such as AI powered digital assistants, which may further increase competition.”
Filing, risk factors, 10-K periodic report, 2025-02-07
“Search engines may also continue to expand their voice and AI capabilities. To the extent these actions have a negative effect on our search traffic or the cost of acquiring such traffic, our business and financial performance could be adversely affected.”
Filing, risk factors, 10-K periodic report, 2025-02-07

Quotes

“Are there opportunities with these connectors or micro apps, for example, in Claude and ChatGPT? It's fascinating to see there are some cases in which we can control a bit more what the interface is, where our brand shows up, and others where we can't.”
c26 · CEO, qa, earnings call, 2026-08-05
“intense competition from online travel agencies, suppliers, search engines, B2B businesses offering competing travel technology solutions and services, and emerging AI-powered platforms;”
c34 · Filing, press release, 8-K earnings release, 2026-08-05
“Technological developments in generative artificial intelligence (“AI”) tools are increasingly being used to create competing offerings, such as AI powered digital planning and assistance, further increasing competition.”
c39 · Filing, mdna, 10-Q periodic report, 2026-08-06

By quarter

  • Q1 2026described · inscrutable · imposed
  • Q2 2026described · inscrutable · imposed

search discovery

Organic search traffic exposed to AI answers and search page changes

Not sized

Withdrawn from the ledger as not an AI channel: a channel is registered only where the company names AI as a cause of the money. Management credits the softness in SEO to algorithm and search page changes in general (claims c24 and c35, now context) and names AI only as the team’s own tool for improving organic performance.

inscrutabledisclosure: not mentioned· motive: imposed· before LLMs: expanded

Asked about search changes, the CEO said organic search traffic, SEO and AEO together, is stable to slightly up, that SEO has been a bit soft and has stabilized, and that search page changes are coming faster; a small team uses AI to test and improve organic performance. None of it names AI answers as the cause of lost traffic, so the passages are kept as context and the channel is read silent and unsized; the former estimate is no longer used.

Evidence: 0 quotes, 4 from before coverage

Withdrawn from the ledger on 2026-10-06 as not an AI channel, under the rule that a channel is registered only where the company names AI as a cause of the money: management credits the softness in unpaid search to algorithm and search page changes in general. Registered in Q2 2026 as paid marketing spent to replace lost unpaid search traffic; the passages are kept as context.

Why this motive

No source names AI as the cause of the traffic change, so no motive is read from this quarter; imposed is kept from the channel’s registration.

Before LLMs: expanded

Dependence on unpaid search is in the FY2024 10-K risk factors, which already name the cost of replacing unpaid traffic with paid traffic and search engines expanding AI capabilities; on the Q1 2024 call the CEO said Google kept pushing SEO traffic down. No unpaid-traffic volume or replacement cost is reported. The size is the change AI made, not the whole line.

“Our marketing channels primarily include brand advertising through online and offline channels, loyalty programs, mobile apps, search engine marketing and optimization as well as metasearch, social media, direct and/or personalized traveler communications on our websites as well as through direct e-mail communication with our travelers.”
Filing, business, 10-K periodic report, 2025-02-07
“Search engines may also continue to expand their voice and AI capabilities. To the extent these actions have a negative effect on our search traffic or the cost of acquiring such traffic, our business and financial performance could be adversely affected.”
Filing, risk factors, 10-K periodic report, 2025-02-07
“If Google or other search or metasearch companies continue to pursue these or similar strategies, which is out of our control, or we do not successfully manage our paid and unpaid search strategies, we could face a significant decrease in traffic to our websites and/or increased costs related to replacing unpaid traffic with paid traffic.”
Filing, risk factors, 10-K periodic report, 2025-02-07
“they continue to operate pretty much how they have in terms of looking for new ways to monetize and push SEO traffic down, et cetera.”
CEO, qa, earnings call, 2024-05-02

Reported lines, year-over-year growth

Revenue +14.0%

Q2 2026. Growing slower than revenue: cost of revenue (+6.9%), selling and marketing, direct (+10.4%), general and administrative (+3.6%). A displaced cost shows up as a line that stays under the dashed revenue line. These are the audited lines, as first reported; nothing here is attributed to AI by the filing.

Cost of revenueSelling and marketing, directGeneral and administrativeRevenue
-20%-10%0%10%20%Q1 2025Q2 2025Q3 2025Q4 2025Q1 2026Q2 2026RevenueSelling and marketing, directCost of revenueGeneral and administrative
Reported values and filings
LineQ1 2025Q2 2025Q3 2025Q4 2025Q1 2026Q2 2026
Revenue$2.99bn$3.79bn$4.41bn$3.55bn$3.43bn$4.32bn
Cost of revenue$357mn$377mn$376mn$346mn$377mn$403mn
Selling and marketing, direct$1.76bn$1.92bn$1.98bn$1.70bn$1.86bn$2.12bn
General and administrative$180mn$197mn$186mn$202mn$196mn$204mn