AI Absorption Ledger / BKNG

Booking Holdings

BKNG · Q2 2026 · reported 2026-08-04 · revenue $7.35bn

Assessment

Q2 2026 is the first quarter in which management put bounds on Booking's AI channels. The CFO said AI costs (tokens and licence fees) are a low single-digit share of technology spend, which applied to Information technology expense of gives a vendor bill of . He also said traffic from large language models, paid and unpaid, is significantly below of room nights, which applied to revenue gives . Both are small against revenue of , and the second was described as not moving.

Customer service and unpaid search are each matched to a reported line whose movement the 10-Q explains by the channel's mechanism, without mentioning AI. Sales and other expenses fell from to of revenue, below the prior-year rate, which the filing puts primarily on efficiencies in third-party customer service costs; the CEO said customer service cost per booking is falling at a double-digit rate and credited AI. Marketing expenses rose from to of gross bookings, above the prior-year rate, which the filing puts primarily on declines in unpaid search traffic; asked whether that is a shift to AI search, the CEO said Google's AI overview probably caused it. These are movements of whole lines, with their other causes listed. Neither is a size for the channel. Customer service carries the ledger's own inferred range; unpaid search is left unsized, since AI is named beside display changes with nothing to separate its part.

Steps since Q1: the vendor bill and demand through AI platforms moved from described to bounded. Internal productivity moved from described to directional, as the CFO cited falling technology cost per merge request. The company's own assistants moved from directional to described, as the CEO declined to give conversion figures and said the effect is still very small. Paid placement on AI platforms is now in practice as a cost-per-click test with OpenAI, with no spend disclosed. A channel for corporate functions opened. The Transformation Program's expected run-rate savings rose from to , mostly from procurement, and remain a confound on both savings channels.

What the filing shows that the call did not attribute to AI: Information technology expense rose , again on cloud computing costs and software licence fees, while the CFO said AI cost is not what drives that line's growth. Personnel expense grew while headcount rose , which the CFO credits to cost management actions.

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: customer service savings of , internal productivity of , corporate functions of , AI-platform fees of , generative AI build spend of , partner tools of , and the company's own assistants of . The unpaid-search toll is unsized because the CEO names the AI overview beside display changes with a hedge, and personalization because its generative AI benefit is placed in the future. Each range spans at least an order of magnitude; the points are placeholders for comparison across quarters and companies, not findings.

Sized channels against the income statement, Q2 2026

9 of 11 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

3 new6 expanded2 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$11mn to $98mn sized

new $2.6mn to $7.9mnexpanded $8.1mn to $90mn

Incremental total $2.6mn to $98mnpoint $5.3mn$29mn in 1 channel has no traced baseline
Cost displaced by AI$3.7mn to $121mn sized

expanded $3.7mn to $121mn

Incremental total $3.7mn to $121mnpoint $21mn
Revenue arriving through AI$0 to $147mn sized

new $0 to $37mnexpanded $0 to $74mnrelabelled $0 to $37mn

Incremental total $0 to $110mnpoint $22mn
Cost imposed, or revenue lost, by others’ AI$0 to $12mn sized

new $0 to $12mnexpanded not sized

Incremental total $0 to $12mnpoint $2.4mn

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 AI2 channels · $11mn to $98mn sized · $2.6mn to $98mn incremental

vendor bill

AI model, licence and compute bill

0.04% to 0.11% of the quarter’s revenue

Incremental total: counts in full.

implied by managementdisclosure: bounded· motive: exploratory· before LLMs: new

The CFO said AI costs are rising, are a low single-digit share of technology spend and are not what drives that line's growth. Applied to Information technology expense of , the methodology's range for the phrase gives . The line itself rose (), which the 10-Q puts on cloud computing costs and software licence fees. The base the CFO has in mind and the AI share of cloud costs are left open. The counterparty is mixed: model and token providers, AI software licensors and the cloud providers behind cloud computing.

Evidence: 8 quotes, 6 figures, 2 confounds, 4 from before coverage

What the company pays outside vendors to run AI: model and token usage, AI licences and the AI share of cloud computing. It sits inside Information technology expense, which the filing does not split.

Why this motive

The tells conflict. Management says AI spending must earn a return or stop and reports falling AI cost per merge request (claims c42 and c46), an efficiency tell for the internal uses. The customer-facing tools the bill also pays for are still in testing and scaled only when benefits are clear (claim c1). The less durable motive is used, unchanged from the prior quarter.

Before LLMs: new

At the anchor the 10-K says the company may use third-party foundational models and names no model, token or AI licence cost. Any such spend sat unsplit inside Information technology expense of for FY2024, whose growth the filing put on cloud computing, data center and software licence costs.

“Our Gen AI initiatives will require increased investment in infrastructure and headcount.”
Filing, risk factors, 10-K periodic report, 2025-02-20
“In some instances we may make use of third-party foundational models that have been pre-trained on data which may be insufficient, erroneous, stale, contain biased information, or infringe intellectual property rights.”
Filing, risk factors, 10-K periodic report, 2025-02-20
“Information technology expenses increased year-over-year in 2024 due to an increase in expenses related to cloud computing costs and outsourced data center costs, as well as software license and system maintenance fees.”
Filing, mdna, 10-K periodic report, 2025-02-20
“We expect our more fixed OpEx to grow faster than revenue in the second quarter, due primarily to faster IT expense growth as we have been investing in new tech platforms and in line with the full year guidance we provided last quarter.”
CFO, prepared remarks, earnings call, 2024-05-02

Figures

  • Information technology expense · 2026-CQ2
  • Information technology expense, prior-year quarter · 2025-CQ2
  • Information technology expense, year-over-year growth as printed in the 10-Q · 2026-CQ2
  • Information technology expense, year-over-year increase · 2026-CQ2
  • Information technology expense as a share of revenue · 2026-CQ2
  • Information technology expense as a share of revenue, prior-year quarter · 2025-CQ2

What else could explain it

  • other: The CFO's base is overall technology spend, which may be wider than the reported Information technology line; if so the bill is larger than the estimate.
  • bundling: The CFO names token cost and licence fees. AI compute bought inside general cloud contracts may sit in cloud computing costs and outside what he counts as AI costs.

Quotes

“Our approach remains focused on deploying this technology in measurable ways and scaling when we see clear benefits.”
c1 · CEO, prepared remarks, earnings call, 2026-08-04
“What we do not want is any surprise because people went bonkers with spending money on tokens or somehow doing something that costs a lot more than we thought it would.”
c41 · CEO, qa, earnings call, 2026-08-04
“We are already seeing an ROI that is positive on our AI investments today as a company.”
c42 · CFO, qa, earnings call, 2026-08-04
“If I'm looking specifically about our AI costs, yes, they're going up. It could be token cost or license fees. It's still at a low single-digit level of our overall technology spend. It's not really a driver of the growth of that line item, but it is going up.”
c43 · CFO, qa, earnings call, 2026-08-04
“The most important one from my perspective is that we are having a cost-aware model routing so that we make sure that for simple tasks, we are using cheaper models, and that for more complex tasks, we're using more expensive models.”
c45 · CFO, qa, earnings call, 2026-08-04
“We're measuring that, for example, by looking at our AI costs over merge requests and our technology cost over merge requests. Actually, those are coming down in a meaningful way.”
c46 · CFO, qa, earnings call, 2026-08-04
“Information technology expenses increased year-over-year for the three and six months ended June 30, 2026 due primarily to an increase in cloud computing costs and software license fees, as well as changes in foreign currency exchange rates.”
c49 · Filing, mdna, 10-Q periodic report, 2026-08-04
“Adjusted fixed operating expenses increased 6%, driven by higher cloud computing costs and software license fees, as well as adverse changes in foreign currency exchange rates”
c50 · Filing, press release, 8-K earnings release, 2026-08-04

By quarter

  • Q1 2026described · our inference · exploratory · $2.4mn to $14mn
  • Q2 2026bounded · implied by management · exploratory · $2.6mn to $7.9mn

engineering

Generative AI build investment

0.11% to 1.2% of the quarter’s revenue

Incremental total: counts at zero at the point and in full at the high end, with no traced baseline.

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

Management says capacity to invest rose with the Transformation Program's savings, lists AI capabilities among the investments, and says the return on AI investment is already positive. None of these statements carries a number, so the channel stays described. The size shown is the ledger's own estimate, , built from the technology share of Personnel expense and an assumed share of that effort on generative AI.

Evidence: 6 quotes, 1 confound, 3 from before coverage

Internal money committed to building generative AI capabilities, which management groups under strategic investments it protects when other costs are cut. Separate from the outside vendor bill.

Why this motive

Management describes measured deployment that is scaled when benefits are clear (claim c1) and investment for long-term growth funded by program savings (claims c2 and c48), the investing-for-the-long-term tell in the methodology's motive table. The CFO's statement that the return is already positive carries no figure (claim c42).

Before LLMs: expanded

At the anchor this was engineering and product payroll inside Personnel expense of for FY2024, with teams of AI experts described as long standing. The 10-K says investment in Gen AI continued in 2024 and gives no amount for it. The size is the whole of an activity that existed before.

“We believe we are well-positioned to leverage this technology, given we have built strong teams of AI experts and gained valuable experience from using AI extensively for many years.”
CEO, prepared remarks, earnings call, 2024-05-02
“We have made significant investments in people, technology, marketing, and added or expanded travel offerings. In 2024, we continued our investments in Gen AI to benefit internal productivity and improve the consumer and partner experience.”
Filing, business, 10-K periodic report, 2025-02-20
“Our Gen AI initiatives will require increased investment in infrastructure and headcount.”
Filing, risk factors, 10-K periodic report, 2025-02-20

What else could explain it

  • bundling: AI is one of several strategic investment areas named together (the Connected Trip, key markets, loyalty); the AI part is not separated.

Quotes

“Our approach remains focused on deploying this technology in measurable ways and scaling when we see clear benefits.”
c1 · CEO, prepared remarks, earnings call, 2026-08-04
“During the quarter, we further increased our capacity to invest for future growth as the expected annual run rate savings enabled by our transformation program continued to increase. These savings provide us with additional flexibility to accelerate innovation, enhance our products, and strengthen our business over the long run.”
c2 · CEO, prepared remarks, earnings call, 2026-08-04
“We are continuing to invest in the Connected Trip, expanding our presence in key markets, advancing our AI capabilities, and strengthening the value we create for both travelers and partners.”
c19 · CFO, prepared remarks, earnings call, 2026-08-04
“As Glenn said, strategically, we are, of course, very much focused on learning ourselves, diversifying our channels, then in the end, focused on keeping the direct traffic to us by investing the AI tools, the customer-facing AI tools in our own environment.”
c27 · CFO, qa, earnings call, 2026-08-04
“We are already seeing an ROI that is positive on our AI investments today as a company.”
c42 · CFO, qa, earnings call, 2026-08-04
“What we are doing in the end is creating capacity to self-fund and invest ultimately in our strategic initiatives that then will help to drive long-term growth for the company.”
c48 · CFO, qa, earnings call, 2026-08-04

By quarter

  • Q1 2026described · our inference · exploratory · $8.0mn to $89mn
  • Q2 2026described · our inference · exploratory · $8.1mn to $90mn

Cost displaced by AI3 channels · $3.7mn to $121mn sized · $3.7mn to $121mn incremental

customer support · cheap to verify

Traveler customer service displaced by AI

0.04% to 1% of the quarter’s revenue

Incremental total: counts in full.

Matched line moved : the whole line, not this channel.

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

The double-digit decline in customer service cost per booking, stated for Agoda alone in the prior quarter, is now stated without a brand qualifier, and voice AI covers the majority of eligible inbound traveler calls. Customer service cost is still not a reported line, so the rate has no base. Sales and other expenses, the line that contains it, came in below what the prior-year share of revenue would give. The filing attributes that primarily to customer service efficiencies, and the CEO attributes the efficiencies to AI. That is the movement of the whole line, not a size for this channel. The size shown is the ledger's own estimate, , built from an assumed customer service share of the line, the stated double-digit decline per booking and an assumed AI share of that decline. The counterparty is mixed: the third-party customer service providers that carry most of the cost, and the company’s own customer service staff.

Evidence: 9 quotes, 3 figures, 6 confounds, 6 from before coverage

Customer service cost for travelers (contacts, third-party service providers, cost per booking) reduced by AI automation and self-service. The cost sits inside Sales and other expenses and is not reported on its own.

Why this motive

A cost per unit falls and management attributes it to AI: customer service cost per booking is down at a double-digit rate (claim c12) and the CEO says the falling cost per interaction is because of AI (claim c39). This is the efficiency tell in the methodology's motive table.

Before LLMs: expanded

At the anchor traveler customer service was call centers, third-party service providers, online self-service and virtual assistants, paid for inside Sales and other expenses of for FY2024, or per quarter on average. Customer service cost was not reported on its own. The size is the change AI made, not the whole line.

“Sales and other expenses are generally variable in nature and consist primarily of: (1) credit card and other payment processing fees associated with merchant transactions; (2) fees paid to third parties that provide call center and other customer services;”
Filing, notes, 10-K periodic report, 2025-02-20
“We strive to provide excellent customer service, including through call centers and online platforms and the use of virtual assistants.”
Filing, business, 10-K periodic report, 2025-02-20
“At each of our OTA brands, our teams are actively exploring ways to leverage Generative AI technology to improve self-service tools, which we believe will reduce live agent contact rates and enable us to answer traveler questions faster.”
CEO, prepared remarks, earnings call, 2024-05-02
“In sum, we believe Gen AI will lower our customer service costs per transaction over time and improve the customer experience.”
CEO, prepared remarks, earnings call, 2024-05-02
“In addition, customer service, which is a critical function that we provide to both our travelers and partners, is an area we believe will be meaningfully enhanced by AI advancements.”
CEO, prepared remarks, earnings call, 2024-05-02
“using Gen AI to drive efficiencies in our operations;”
Filing, business, 10-K periodic report, 2025-02-20

Figures

  • Sales and other expenses as a share of gross bookings · 2026-CQ2
  • Sales and other expenses as a share of gross bookings, prior-year quarter · 2025-CQ2
  • Increase in merchant transaction costs within Sales and other expenses · 2026-CQ2

Reported line it is matched to

Customer service cost sits inside Sales and other expenses. That line fell from to of revenue, which is below what the prior-year share would give. The 10-Q names efficiencies in third-party customer service costs as the primary cause and does not mention AI; the AI attribution is the CEO's on the call. The dollar figure is the movement of the whole line, with the other causes listed.

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

What else could explain it

  • mix shift: Merchant transaction costs in the same line rose with the shift from agency to merchant bookings. This pushes against the movement, so the net movement of the line may understate what customer service did.
  • one time item: The CFO cites a one-time benefit from processing fee reversals in payment expenses, which lowers the same line; no amount is given.
  • transformation program: The Transformation Program's expected run-rate savings rose to , with the added mostly from procurement. Third-party customer service is a purchased service, and the sources do not say whether its savings are counted inside the program. The program is not attributed to AI.
  • fx: The 10-Q names changes in foreign currency exchange rates as a cause of the increase in the line in dollars.
  • other: The CEO credits combined AI and human capabilities for the falling cost per booking, so part of the efficiency is operational and unrelated to AI.
  • other: Revenue grew more slowly than gross bookings because of earlier cancellations, which moves the ratio's denominator. Against gross bookings the line fell from to .

Quotes

“Finally, we're increasingly applying AI across our own business to accelerate software development, enhance customer service, and streamline internal workflows.”
c10 · CEO, prepared remarks, earnings call, 2026-08-04
“Our AI initiatives are reducing customer friction, lowering contact rates, and improving operational efficiency. We have now scaled voice AI support across the majority of eligible inbound traveler calls while continuing to expand digital automation.”
c11 · CEO, prepared remarks, earnings call, 2026-08-04
“As a result of these combined AI and human capabilities, customer service cost per booking continues to decrease at a double-digit rate while overall customer satisfaction remains high.”
c12 · CEO, prepared remarks, earnings call, 2026-08-04
“Adjusted sales and other expenses were 1.9% of gross bookings and provided a leverage despite the higher merchant mix, as higher payment expenses were more than offset by customer service efficiencies. Additionally, payment expenses grew less than merchant gross bookings due to a one-time benefit from processing fee reversals.”
c16 · CFO, prepared remarks, earnings call, 2026-08-04
“As we continue to execute on the transformation program, we identified additional opportunities, increasing our expected annual run rate savings from approximately $550 million to approximately $650 million.”
c18 · CFO, prepared remarks, earnings call, 2026-08-04
“We've been talking about how our cost per interaction with a customer are going down. That's because of AI. That's using it, and we're getting better CSAT numbers.”
c39 · CEO, qa, earnings call, 2026-08-04
“We found additional savings opportunities within existing categories, no new categories. This is all existing categories to go deeper in some of those areas, and the additional $100 million mostly came from the procurement work stream.”
c47 · CFO, qa, earnings call, 2026-08-04
“Sales and other expenses as a percentage of total revenues decreased year-over-year for the three months ended June 30, 2026 primarily due to efficiencies in third-party customer service costs.”
c51 · Filing, mdna, 10-Q periodic report, 2026-08-04
“As we continued to execute on the Transformation Program, we identified additional opportunities, increasing our expected annual run-rate savings to approximately $650 million.”
c58 · Filing, mdna, 10-Q periodic report, 2026-08-04

By quarter

  • Q1 2026direction only · our inference · efficiency · $746k to $63mn
  • Q2 2026direction only · our inference · efficiency · $3.1mn to $74mn

engineering · cheap to verify

Internal workflow and product development productivity

0.01% to 0.55% of the quarter’s revenue

Incremental total: counts in full.

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

This quarter management moved from describing internal AI use to stating a direction: AI cost and technology cost per merge request are coming down, with adoption and productivity metrics in place. The Personnel line was nearly flat while revenue grew, which fits the claim, but the CFO credits cost management actions and headcount rose. Corporate functions named separately this quarter are read in their own channel. The size shown is the ledger's own estimate, , built from the technology share of Personnel expense, a productivity gain taken from public developer studies and an assumed share of the gain that reaches the cost line.

Evidence: 11 quotes, 6 figures, 4 confounds, 4 from before coverage

Employee time displaced by AI in product development, decision-making and other internal workflows. The line it would show in is Personnel.

Why this motive

The tells conflict. The CFO reports technology cost per merge request coming down (claim c46), a cost-per-unit efficiency tell, but it is an internal metric with no rate, the CEO calls the results promising and early (claim c13), headcount rose (claim c54), and the CFO credits fixed-cost restraint to cost management actions (claim c17). The less durable motive is used, unchanged from the prior quarter.

Before LLMs: expanded

At the anchor this was employee time inside Personnel expense of for FY2024. The 10-K names internal productivity as a purpose of Gen AI investment and reports no saving, and a cost programme announced in November 2024 covers the same lines without reference to AI. The size is the change AI made, not the whole line.

“We have made significant investments in people, technology, marketing, and added or expanded travel offerings. In 2024, we continued our investments in Gen AI to benefit internal productivity and improve the consumer and partner experience.”
Filing, business, 10-K periodic report, 2025-02-20
“We are seeking to incorporate Gen AI in our business, including for internal productivity purposes and in consumer- and partner-facing initiatives such as AI travel assistants, price comparison tools, and as part of enhancing the development of our Connected Trip vision.”
Filing, risk factors, 10-K periodic report, 2025-02-20
“In November 2024, we announced our intention to implement certain organizational changes, including modernizing processes and systems, initiating an expected workforce reduction, optimizing procurement, and seeking real estate savings”
Filing, mdna, 10-K periodic report, 2025-02-20
“using Gen AI to drive efficiencies in our operations;”
Filing, business, 10-K periodic report, 2025-02-20

Figures

  • Employee headcount · as-of 2026-06-30
  • Adjusted personnel expenses, year-over-year growth, per the CFO · 2026-CQ2
  • Transformation Program expected annual run-rate savings, before this quarter's increase · as-of 2025-12-31
  • Transformation Program expected annual run-rate savings, after this quarter's increase · as-of 2026-08-04
  • Transformation Program run-rate savings added this quarter, mostly from procurement · as-of 2026-08-04
  • Transformation costs · 2026-CQ2

Reported line it is matched to

The claim is that AI is raising developer productivity and lowering technology cost per merge request. Personnel expense was against , growth of while revenue grew . Headcount still rose . The 10-Q explains the line by salaries, currency and lower stock-based compensation and does not mention AI.

2026-CQ2: 2025-CQ2: 2026-CQ2: as-of 2026-06-30: 2026-CQ2:

What else could explain it

  • other: The CFO attributes the fixed-cost leverage, including adjusted personnel growth of , to the targeted cost management actions taken last quarter in response to the macro backdrop.
  • transformation program: The Transformation Program's expected run-rate savings rose from to . Management does not attribute the program to AI.
  • other: The 10-Q says lower stock-based compensation offset higher salaries in the Personnel line.
  • fx: The 10-Q says part of the salary increase came from currency, and revenue growth includes a currency benefit.

Quotes

“Finally, we're increasingly applying AI across our own business to accelerate software development, enhance customer service, and streamline internal workflows.”
c10 · CEO, prepared remarks, earnings call, 2026-08-04
“We're also seeing promising early results in our technology organization, where AI is accelerating software development and improving developer productivity.”
c13 · CEO, prepared remarks, earnings call, 2026-08-04
“Adjusted fixed operating expenses increased 6% year-over-year, including 1% higher adjusted personnel expenses, and were a source of leverage as a percentage of revenue, reflecting the targeted cost management actions we implemented last quarter and our continued focus on managing our fixed expense base while investing in key strategic priorities to drive long-term growth.”
c17 · CFO, prepared remarks, earnings call, 2026-08-04
“As we continue to execute on the transformation program, we identified additional opportunities, increasing our expected annual run rate savings from approximately $550 million to approximately $650 million.”
c18 · CFO, prepared remarks, earnings call, 2026-08-04
“Certainly, everybody talks about the improvements in coding, but of course, you got to talk about the whole software development life cycle. Actually, it's improving all areas of it, and we're working very hard, and we are seeing real results.”
c38 · CEO, qa, earnings call, 2026-08-04
“We have introduced very specific metrics to measure the benefit of this. Think about, for example, in the engineering world, metrics around adoption, metrics around productivity.”
c44 · CFO, qa, earnings call, 2026-08-04
“We're measuring that, for example, by looking at our AI costs over merge requests and our technology cost over merge requests. Actually, those are coming down in a meaningful way.”
c46 · CFO, qa, earnings call, 2026-08-04
“We found additional savings opportunities within existing categories, no new categories. This is all existing categories to go deeper in some of those areas, and the additional $100 million mostly came from the procurement work stream.”
c47 · CFO, qa, earnings call, 2026-08-04
“Employee headcount increased 3% year-over-year to approximately 25,550 as of June 30, 2026.”
c54 · Filing, mdna, 10-Q periodic report, 2026-08-04
“Personnel expenses increased slightly year-over-year for the three months ended June 30, 2026 primarily due to an increase in salary expenses partially driven by changes in foreign currency exchange rates, partially offset by a decrease in stock-based compensation expense.”
c55 · Filing, mdna, 10-Q periodic report, 2026-08-04
“As we continued to execute on the Transformation Program, we identified additional opportunities, increasing our expected annual run-rate savings to approximately $650 million.”
c58 · Filing, mdna, 10-Q periodic report, 2026-08-04

By quarter

  • Q1 2026described · our inference · exploratory · $536k to $40mn
  • Q2 2026direction only · our inference · exploratory · $540k to $41mn

back office · expensive to verify

AI in corporate functions

0% to 0.09% of the quarter’s revenue

Incremental total: counts in full.

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

First seen this quarter: the CEO names finance, public relations and public affairs as functions using AI, separately from software development and customer service. Nothing is said about cost or headcount in those functions, and the 10-Q's explanations of Personnel and General and administrative expense do not mention AI. The size shown is the ledger's own estimate, , built from an assumed corporate-function share of Personnel expense, a productivity gain taken from public developer studies and an assumed share of the gain that reaches the cost line. The counterparty is mixed: the company’s own corporate-function staff and the outside vendors those functions pay.

Evidence: 2 quotes, 4 from before coverage

Employee time and outside spend displaced by AI in corporate functions that management names separately from software development and customer service: finance, public relations and public affairs. The lines it would show in are Personnel and General and administrative.

Why this motive

The CEO lists functions where AI is in use (claim c40) and gives no cost effect, price or purpose beyond streamlining workflows (claim c10), and the 10-Q explains the lines without AI. AI named with no measure and no line moving is the exploratory tell for a passing mention; nothing in the sources contradicts it.

Before LLMs: expanded

At the anchor finance, public relations and public affairs work sat inside Personnel expense of and General and administrative expense of for FY2024, with no functional split. The 10-K names internal productivity as a purpose of Gen AI investment and reports no saving. The size is the change AI made, not the whole line.

“We have made significant investments in people, technology, marketing, and added or expanded travel offerings. In 2024, we continued our investments in Gen AI to benefit internal productivity and improve the consumer and partner experience.”
Filing, business, 10-K periodic report, 2025-02-20
“We are seeking to incorporate Gen AI in our business, including for internal productivity purposes and in consumer- and partner-facing initiatives such as AI travel assistants, price comparison tools, and as part of enhancing the development of our Connected Trip vision.”
Filing, risk factors, 10-K periodic report, 2025-02-20
“In November 2024, we announced our intention to implement certain organizational changes, including modernizing processes and systems, initiating an expected workforce reduction, optimizing procurement, and seeking real estate savings”
Filing, mdna, 10-K periodic report, 2025-02-20
“using Gen AI to drive efficiencies in our operations;”
Filing, business, 10-K periodic report, 2025-02-20

Quotes

“Finally, we're increasingly applying AI across our own business to accelerate software development, enhance customer service, and streamline internal workflows.”
c10 · CEO, prepared remarks, earnings call, 2026-08-04
“It's being used in finance. It's being used in PR. It's being used by our public affairs.”
c40 · CEO, qa, earnings call, 2026-08-04

Revenue arriving through AI4 channels · $0 to $147mn sized · $0 to $110mn incremental · 1 not sized

partner support · cheap to verify

AI tools for accommodation and restaurant partners

0% to 0.5% of the quarter’s revenue

Incremental total: counts at zero.

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

AI messaging for guest inquiries is the one live example given; other capabilities are being co-developed with partners. The benefit is described as partner efficiency and a stronger value proposition, with nothing tied to a revenue line. OpenTable's partner tools, described last quarter, were not mentioned. The size shown is the ledger's own estimate, , built as revenue times an assumed adopting-partner share times an assumed incremental demand lift.

Evidence: 2 quotes, 4 from before coverage

AI features delivered to supply partners (guest messaging, service flows, restaurant reservation and insight tools) that management says may bring partners incremental demand and utilization, and through them commission or subscription revenue.

Why this motive

The CEO describes investing in partner tools and co-developing new capabilities with partners (claims c8 and c9), with no price, attach rate or revenue. The live messaging feature at no stated price would read as product-defensive; where the tells conflict the less durable motive is used.

Before LLMs: relabelled

At the anchor the company already offered partners demand, inventory utilization and customer service through its platforms, and OpenTable charged restaurants subscription fees for restaurant management services. The 10-K mentions partner-facing Gen AI initiatives and reports no revenue from them.

“In addition, customer service, which is a critical function that we provide to both our travelers and partners, is an area we believe will be meaningfully enhanced by AI advancements.”
CEO, prepared remarks, earnings call, 2024-05-02
“We believe they benefit from participating in our services by increasing their distribution channels, demand, profile and reputation, and inventory utilization in an efficient and cost-effective manner.”
Filing, business, 10-K periodic report, 2025-02-20
“revenues earned by OpenTable for its restaurant reservation services and subscription fees for restaurant management services”
Filing, business, 10-K periodic report, 2025-02-20
“We are seeking to incorporate Gen AI in our business, including for internal productivity purposes and in consumer- and partner-facing initiatives such as AI travel assistants, price comparison tools, and as part of enhancing the development of our Connected Trip vision.”
Filing, risk factors, 10-K periodic report, 2025-02-20

Quotes

“AI is also strengthening our partner value proposition. We are investing in tools that help partners better engage with guests, improve their property content, and operate more efficiently. For example, our AI-powered messaging capabilities enable accommodation partners to respond to guest inquiries more quickly and consistently, reducing operational friction and improving the traveler experience.”
c8 · CEO, prepared remarks, earnings call, 2026-08-04
“This conference also reinforced our collaborative approach to innovation, including co-developing new AI capabilities using direct partner feedback.”
c9 · CEO, prepared remarks, earnings call, 2026-08-04

By quarter

  • Q1 2026described · our inference · exploratory · $0 to $28mn
  • Q2 2026described · our inference · exploratory · $0 to $37mn

search discovery · cheap to verify

Conversion from the company's own AI assistants and search

0% to 1% of the quarter’s revenue

Incremental total: counts in full.

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

New this quarter: a test of Booking.com's AI discovery experience, the next generation of Penny with hotel checkout inside it, and Agoda's gallery view. The prior quarter's statement of a conversion uplift was not repeated; management spoke of potential and of satisfaction, and declined to give numbers. The state is therefore read as described, one step below the prior quarter. The size shown is the ledger's own estimate, , built as revenue times an assumed exposed share times an assumed conversion lift.

Evidence: 8 quotes, 1 confound, 3 from before coverage

Bookings gained because travelers use the company's own conversational and natural-language tools: Penny at Priceline, AI search and smart filters at Booking.com, AI Concierge at OpenTable.

Why this motive

Management tags the stage itself: "it's out very early" with no data (claim c32) and "still very small" (claim c20). The CFO frames the tools as keeping direct traffic (claim c27), a product-defensive tell; where the tells conflict the less durable motive is used.

Before LLMs: expanded

At the anchor travelers found and booked through ordinary on-site search and filters, and the AI Trip Planner, Penny and Kayak's generative AI tools were already launched and used by a very small number of travelers. The anchor reports no bookings from them. The size is the change AI made, not the whole line.

“We have also launched consumer-facing Gen AI capabilities, including a trip planner, an AI assistant to answer consumer queries, and a price comparison tool.”
Filing, business, 10-K periodic report, 2025-02-20
“As we have discussed before, our teams continue to work hard to integrate generative AI into our offerings in innovative ways, including Booking.com's AI Trip Planner, Priceline's generative AI travel assistant named Penny, and Kayak's recent release of generative AI-powered features and tools.”
CEO, prepared remarks, earnings call, 2024-05-02
“So it's a very small number of people compared to the number of people who use our services, but we are continuing to advance it.”
CEO, qa, earnings call, 2024-05-02

What else could explain it

  • other: The tools are in limited tests, so any observed effect comes from small and possibly self-selected groups of users.

Quotes

“During the quarter, we continued advancing AI capabilities across our portfolio, including beginning to roll out testing of Booking.com's new AI-powered discovery experience, which helps travelers in the early inspiration phase of planning a trip.”
c5 · CEO, prepared remarks, earnings call, 2026-08-04
“We're also deploying the next generation of Priceline's agentic AI travel assistant called Penny, and we're launching Agoda's new gallery view, which gives travelers a more visual way to browse search results by pairing hotel images with relevant guest reviews.”
c6 · CEO, prepared remarks, earnings call, 2026-08-04
“For example, initial testing of Penny's integrated hotel checkout experience has shown that beginning more of the booking journey into a single seamless experience has the potential to improve traveler engagement while delivering stronger business outcomes.”
c7 · CEO, prepared remarks, earnings call, 2026-08-04
“All these things are good, but, and this is important, it's still very small, and we're not at a stage we're going to start giving away percentages, numbers, increases in CSAT, or anything of that nature.”
c20 · CEO, qa, earnings call, 2026-08-04
“As Glenn said, strategically, we are, of course, very much focused on learning ourselves, diversifying our channels, then in the end, focused on keeping the direct traffic to us by investing the AI tools, the customer-facing AI tools in our own environment.”
c27 · CFO, qa, earnings call, 2026-08-04
“Point of this, summing it all up in that area is that we are working hard with the third-party frontier players to make sure we're properly placed when somebody wants to book if they want to start there. Even more important to me is making sure that we're offering our own opportunities for people to come to us direct.”
c30 · CEO, qa, earnings call, 2026-08-04
“Kevin, it's out very early. I don't have any data to come back to you.”
c32 · CEO, qa, earnings call, 2026-08-04
“We're going to keep on working with that so that people, when they think of travel, they think, "I can go to one of our brands," and they will get everything they could get from any of the large language models, but get even more because we're able to personalize it from what they've done with us in the past”
c59 · CEO, qa, earnings call, 2026-08-04

By quarter

  • Q1 2026direction only · our inference · exploratory · $0 to $55mn
  • Q2 2026described · our inference · exploratory · $0 to $74mn

product ranking · cheap to verify

Conversion from AI personalization and ranking

Not sized

Asked for evidence, the CEO answers only in the future tense (claim c22: with generative AI the company “will be able” to personalize better) and declines to give figures (claim c20); the 10-Q line is a strategy aim (claim c56). The generative AI benefit had not started in the quarter’s words, so the channel is unsized (the methodology rule for money that has not started).

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

Asked for evidence that personalization has improved conversion, the CEO answered in the future tense and gave no numbers. The 10-Q repeats its statement of an AI-powered traveler experience as long-term strategy. The benefit is placed in the future, so the channel is left unsized; the former estimate is no longer used.

Evidence: 5 quotes, 1 confound, 3 from before coverage

Bookings gained because AI applied to the company's proprietary data personalizes what each traveler is shown across the Connected Trip, as distinct from the conversational assistants.

Why this motive

The CEO places the benefit in the future: with generative AI the company "will be able" to personalize better, leading to better completion of transactions (claim c22), and declines to give any current figure (claim c20). This is the pre-revenue tell in the methodology's motive table. The prior quarter read the same way, as a passing mention with no measure.

Before LLMs: relabelled

At the anchor the company already offered personalized travel services and said it had used AI extensively for many years. Ranking and personalization earned commission inside total revenues with no line of their own.

“We believe we are well-positioned to leverage this technology, given we have built strong teams of AI experts and gained valuable experience from using AI extensively for many years.”
CEO, prepared remarks, earnings call, 2024-05-02
“In order to achieve the easier and more personalized experience of the Connected Trip, we have always envisioned AI technology playing a central role.”
CEO, prepared remarks, earnings call, 2024-05-02
“We endeavor to provide consumers with: (a) personalized and easy-to-use online travel services;”
Filing, business, 10-K periodic report, 2025-02-20

What else could explain it

  • relabel: The CEO says personalization was earlier described as AI in the sense of machine learning prediction models; the sources do not separate what generative AI adds from what already existed.

Quotes

“All these things are good, but, and this is important, it's still very small, and we're not at a stage we're going to start giving away percentages, numbers, increases in CSAT, or anything of that nature.”
c20 · CEO, qa, earnings call, 2026-08-04
“If you recall, when I first started talking about, for example, the Connected Trip, bringing it together, and we were talking about AI in terms of machine learning models, so ending up with different types of mathematical predictions of what we thought they want.”
c21 · CEO, qa, earnings call, 2026-08-04
“Now using Gen AI, we'll be able to do it so much better. I envision in the future we're going to get even more personalized than I ever thought we could, which will lead to what you're asking about in terms of a much better way to actually complete a transaction.”
c22 · CEO, qa, earnings call, 2026-08-04
“We are executing against our long-term strategy to create an ideal AI-powered traveler experience, offering our customers relevant options and suggestions at the times and in the language they want them, making trips booked with us seamless, easy, and valuable.”
c56 · Filing, mdna, 10-Q periodic report, 2026-08-04
“We're going to keep on working with that so that people, when they think of travel, they think, "I can go to one of our brands," and they will get everything they could get from any of the large language models, but get even more because we're able to personalize it from what they've done with us in the past”
c59 · CEO, qa, earnings call, 2026-08-04

By quarter

  • Q1 2026described · our inference · exploratory · $0 to $111mn
  • Q2 2026described · described, no size · exploratory

distribution · expensive to verify

Demand arriving through third-party AI platforms

0% to 0.5% of the quarter’s revenue

Incremental total: counts in full.

implied by managementdisclosure: bounded· motive: channel-defensive· before LLMs: new

The CFO said traffic from large language models, paid and unpaid, is significantly below of room nights and has not moved materially in recent quarters. Applied to revenue of , the methodology's range for the phrase gives . The CEO said the company appears often in AI answers while referrals stay low, and named a start with Google's agentic booking. How much of this demand is incremental is left open.

Evidence: 9 quotes, 2 figures, 3 confounds, 2 from before coverage

Bookings that begin in or are referred by an outside AI platform (OpenAI, Google, Anthropic, Amazon and others) through the partnerships and integrations management describes.

Why this motive

The stated aim is to be placed where a traveler may start (claims c3 and c30), the staying-present tell in the methodology's motive table. The CEO ranks the company's own direct tools as more important, and the volume is described as not moving (claim c24).

Before LLMs: new

The anchor reports no bookings or referrals from AI platforms. The traffic sources it names are direct, search engines, meta-search, affiliates and social media, and AI agents appear only as possible competitors in the risk factors.

“In addition, the widespread adoption of new technologies, such as Gen AI and machine learning, could influence how customers search for and book travel, render our existing technology obsolete, require us to modify or adapt our services or infrastructure, which could adversely affect our results of operations or financial condition.”
Filing, risk factors, 10-K periodic report, 2025-02-20
“companies offering AI agents powered by Gen AI that can perform or facilitate travel-related services, such as virtual assistants.”
Filing, risk factors, 10-K periodic report, 2025-02-20

Figures

  • Share of room nights arriving from large language models, paid and unpaid: the level management says it is significantly below · 2026-CQ2
  • Room nights booked · 2026-CQ2

What else could explain it

  • other: The bound covers paid and unpaid traffic together. The paid part is bought, and its cost belongs to the paid placement channel.
  • other: Travelers arriving through an AI platform may otherwise have come direct or through search, so the revenue need not be incremental.
  • other: The bound is on room nights booked; revenue is recognized at check-in and includes non-accommodation services, so the dollar range is an approximation.

Quotes

“That's why we continue to work closely with leading AI organizations to ensure travelers can engage with our brands wherever their journey begins.”
c3 · CEO, prepared remarks, earnings call, 2026-08-04
“While AI-driven referrals remains a relatively small contributor to our overall business today, we are seeing encouraging momentum that reinforces our belief that the strengths we've built over many years, helping travelers find the right travel options with confidence, positions us well as travelers further gravitate towards AI-powered discovery.”
c4 · CEO, prepared remarks, earnings call, 2026-08-04
“If we look at the traffic we're currently receiving from large language models, so this is not what we are receiving directly in terms of traffic, but from large language models, both on a paid and an unpaid basis, that is still significantly below 1% of our room nights.”
c23 · CFO, qa, earnings call, 2026-08-04
“That hasn't moved so much recently, so no material change over the last few months or quarters.”
c24 · CFO, qa, earnings call, 2026-08-04
“Going a little deeper, the organic traffic, what we are receiving from the LLMs, we believe the vast majority is coming to us according to third-party data sources.”
c25 · CFO, qa, earnings call, 2026-08-04
“On the other hand, he also mentioned we're showing up a lot, but the referrals aren't showing much.”
c28 · CEO, qa, earnings call, 2026-08-04
“In fact, the most recent thing is, I believe maybe today even, I think Google may have come out today, a small percentage of their traffic, seeing their agentic booking, which we are one of the first with them.”
c29 · CEO, qa, earnings call, 2026-08-04
“Point of this, summing it all up in that area is that we are working hard with the third-party frontier players to make sure we're properly placed when somebody wants to book if they want to start there. Even more important to me is making sure that we're offering our own opportunities for people to come to us direct.”
c30 · CEO, qa, earnings call, 2026-08-04
“I will say, though, again, I doubt that anybody is getting a huge amount of business directly from what they're doing with their travel prompts, their discovery.”
c36 · CEO, qa, earnings call, 2026-08-04

By quarter

  • Q1 2026described · our inference · channel-defensive · $0 to $28mn
  • Q2 2026bounded · implied by management · channel-defensive · $0 to $37mn

Cost imposed, or revenue lost, by others’ AI2 channels · $0 to $12mn sized · $0 to $12mn incremental · 1 not sized

marketing

Paid placement on AI platforms

0% to 0.16% of the quarter’s revenue

Incremental total: counts in full.

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

The channel moved from prospect to practice: the company is in OpenAI's test group for cost-per-click advertising. No spend is disclosed. All traffic from large language models, paid and unpaid, is significantly below of room nights, so the paid part is smaller still. The 10-Q's list of performance marketing channels still does not include AI platforms. The size shown is the ledger's own estimate, , built as a share of the Marketing line capped by the disclosed share of room nights from large language models.

Evidence: 4 quotes, 1 figure

Performance marketing and other fees paid to AI platforms to appear where travelers start their search. It would sit inside Marketing expenses.

Why this motive

The CFO values the OpenAI cost-per-click test as diversification of paid channels (claim c26) and the CEO likens it to the performance marketing the company already does (claim c31): spending to be present where travelers start, the channel-defensive tell in the methodology's motive table.

Before LLMs: new

The anchor is silent on fees paid to AI platforms. Marketing expenses of for FY2024 went to search engine keyword purchases, affiliate programs, meta-search referrals, social media and brand marketing.

Figures

  • Share of room nights arriving from large language models, paid and unpaid: the level management says it is significantly below · 2026-CQ2

Quotes

“If we look at the traffic we're currently receiving from large language models, so this is not what we are receiving directly in terms of traffic, but from large language models, both on a paid and an unpaid basis, that is still significantly below 1% of our room nights.”
c23 · CFO, qa, earnings call, 2026-08-04
“If I look at the paid traffic, we are part of the test group of OpenAI for CPC. We like that from a diversification perspective in terms of our paid channels, and it's very much playing to our strength, the optimization models that we are running in order to convert, at an optimal way, traffic into bookings.”
c26 · CFO, qa, earnings call, 2026-08-04
“Personally, I like to take a little pride in the fact that I thought that one of them would start going for a performance marketing advertising platform, which is right in our wheelhouse, that we're very, very good at, and they have, and we're working very hard on that, so I'm pleased about that.”
c31 · CEO, qa, earnings call, 2026-08-04
“Our performance marketing expenses, which represent a substantial majority of our marketing expenses, are primarily related to the use of online search engines (primarily Google), affiliate marketing (i.e., business-to-business or "B2B"), meta search, and social media channels to generate bookings through our platforms.”
c57 · Filing, mdna, 10-Q periodic report, 2026-08-04

By quarter

  • Q1 2026described · described, no size · channel-defensive
  • Q2 2026described · our inference · channel-defensive · $0 to $12mn

marketing

Organic search traffic lost and bought back

Not sized

The CEO names Google’s display changes beside the AI overview and hedges (claims c33 and c34: the AI overview “probably” did it), the CFO sees the pressure across much of consumer internet (claim c14), and the line also moved on paid traffic mix and chosen paid investment (claim c15). Nothing separates AI’s part (the methodology rule for AI named beside another cause), so the line is a shape and the channel is unsized; the Marketing line’s movement is the ceiling.

Matched line moved : the whole line, not this channel.

shape matchdisclosure: direction only· motive: imposed· before LLMs: expanded

Two things changed from the prior quarter. The 10-Q now names declines in unpaid search traffic as the primary reason marketing expenses rose faster than gross bookings ( against ), and the CEO, asked whether this is a shift to AI search, said Google's AI overview probably caused it. Marketing expenses came in above what the prior-year share of gross bookings would give. That is the movement of the whole line, with chosen paid-channel investment and mix inside it, not a size for this channel. The share of room nights from unpaid search is still described only as small. AI is named beside display changes with a hedge, so the reading is a shape with no size; the former estimate is no longer used.

Evidence: 8 quotes, 4 figures, 6 confounds, 3 from before coverage

Unpaid search traffic that declines as search engines and AI answers change how travelers find options, and the paid marketing spent to replace it. The cost would show in Marketing expenses as a share of gross bookings.

Why this motive

The CEO says the company is losing unpaid search traffic and does not like it (claim c35), and the filing expects the decline to continue (claim c53). The company did not choose the cost.

Before LLMs: expanded

At the anchor dependence on Google and other search services was a stated risk, including the cost of keeping placement when search result logic changes. Marketing expenses were for FY2024, of gross bookings, and unpaid search traffic was not reported on its own. The size is the change AI made, not the whole line.

“We rely upon Google and other search and meta-search services to generate a significant portion of traffic to our platforms, principally through pay-per-click marketing campaigns. The pricing and operating dynamics on these platforms can change rapidly. If the logic determining placement and display of results of a consumer's search changes, the placement of links to our platforms can be negatively affected and our costs to improve or maintain our placement can increase.”
Filing, risk factors, 10-K periodic report, 2025-02-20
“Our performance marketing expenses, which represent a substantial majority of our marketing expenses, are primarily related to the use of online search engines (primarily Google), affiliate marketing, meta-search, and social media channels to generate traffic to our platforms.”
Filing, mdna, 10-K periodic report, 2025-02-20
“In addition, the widespread adoption of new technologies, such as Gen AI and machine learning, could influence how customers search for and book travel, render our existing technology obsolete, require us to modify or adapt our services or infrastructure, which could adversely affect our results of operations or financial condition.”
Filing, risk factors, 10-K periodic report, 2025-02-20

Figures

  • Marketing expenses, year-over-year growth as printed in the 10-Q · 2026-CQ2
  • Total gross bookings, year-over-year growth as printed in the 10-Q · 2026-CQ2
  • Total gross bookings · 2026-CQ2
  • Total gross bookings, prior-year quarter · 2025-CQ2

Reported line it is matched to

On the call the CEO says Google's AI overview probably caused the pressure on unpaid search traffic. Marketing expenses rose from to of gross bookings, a rise of percentage points, which is above what the prior-year share would give. The 10-Q names declines in unpaid search traffic as the primary cause and does not mention AI. The dollar figure is the movement of the whole line, with the other causes listed.

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

What else could explain it

  • mix shift: The 10-Q names changes in paid traffic mix as a second cause of the higher ratio, and the CFO cites changes in traffic mix.
  • other: The 10-Q and the CFO name added investment in paid marketing channels at attractive returns, which is chosen spending.
  • relabel: The CFO cites a shift of merchandising spend to performance marketing, which moves existing cost into the Marketing line.
  • other: The AI attribution is hedged: the CEO names changes in Google's display and says the AI overview "probably" did it, and the CFO says the pressure is seen across much of consumer internet. Display changes other than AI answers could produce the same decline.
  • fx: The 10-Q says currency raised marketing expenses in dollars; the effect largely cancels in the ratio to gross bookings.
  • other: Display changes at Google are named beside the AI overview as the cause of the pressure on unpaid search, with no split.

Quotes

“This performance came despite the continued pressure on SEO, which we are seeing across much of consumer internet. SEO remains a small component of our overall room nights.”
c14 · CFO, prepared remarks, earnings call, 2026-08-04
“Marketing expense increased 11% year-over-year, modestly faster than gross bookings, driven by changes in traffic mix, incremental investments in paid marketing at attractive ROIs, and a shift of merchandising spend to performance marketing.”
c15 · CFO, prepared remarks, earnings call, 2026-08-04
“I do believe that some of the changes that were made in the display at Google definitely put some pressure on SEO”
c33 · CEO, qa, earnings call, 2026-08-04
“Putting in that AI overview probably has done it.”
c34 · CEO, qa, earnings call, 2026-08-04
“That says that we are doing a good job of making sure that people come to us, think of us, and yes, losing SEO, don't like it. It was never a big part of the business, small part of the business, but obviously I don't like that and I wonder, somebody, they didn't come to us from SEO, which would have been in that direct category.”
c35 · CEO, qa, earnings call, 2026-08-04
“Again, just to emphasize, it's a very small component of our overall direct mix, and we see some slight headwinds, but the overall direct channel is growing in absolute numbers.”
c37 · CFO, qa, earnings call, 2026-08-04
“Marketing expenses as a percentage of total gross bookings in the three months ended June 30, 2026 were slightly higher than the three months ended June 30, 2025, driven primarily by declines in SEO, which remains a small component of our overall distribution mix, as well as changes in paid traffic mix and investments in paid marketing channels at attractive ROIs.”
c52 · Filing, mdna, 10-Q periodic report, 2026-08-04
“While we seek to adapt to evolving search engine dynamics, we expect SEO traffic to decline in the short to medium term, which may lead to increased spend in paid marketing channels.”
c53 · Filing, mdna, 10-Q periodic report, 2026-08-04

By quarter

  • Q1 2026direction only · shape match · imposed
  • Q2 2026direction only · shape match · imposed

Reported lines, year-over-year growth

Revenue +8.1%

Q2 2026. Growing slower than revenue: personnel (+0.4%), total operating expenses (+6.7%). 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. Not drawn: general and administrative, which one-time items move by more than 60% in a quarter; the values are in the table below.

Marketing expensesPersonnelInformation technologyTotal operating expensesRevenue
-20%-10%0%10%20%30%Q1 2025Q2 2025Q3 2025Q4 2025Q1 2026Q2 2026Information technologyMarketing expensesRevenueTotal operating expensesPersonnel
Reported values and filings
LineQ1 2025Q2 2025Q3 2025Q4 2025Q1 2026Q2 2026
Total revenues$4.76bn$6.80bn$9.01bn$6.35bn$5.53bn$7.35bn
Marketing expenses$1.78bn$2.14bn$2.34bn$1.93bn$2.07bn$2.37bn
Personnel$693mn$896mn$945mn$869mn$893mn$900mn
Information technology$200mn$219mn$242mn$247mn$240mn$263mn
General and administrative$142mn$199mn$254mn$262mn$100mn$217mn
Total operating expenses$3.70bn$4.55bn$5.53bn$4.32bn$4.26bn$4.85bn