SASIGNAL ATLASCross-industry intelligence / Research desk
SIGNAL ATLAS / RESEARCH DESK

Advertising, marketing & brand

Dossier · Advertising, marketing & brand · Original Phase 1 research

Advertising, marketing & brand

Industry ID: 16 | Slug: advertising-marketing | Researched: 2026-09-15 | Analyst: agent

Research-capability note. The shared session WebSearch budget (200 calls) was exhausted after five searches by this agent. Per the macro brief's standing instruction, discovery switched to WebFetch against primary sources — company IR pages, regulator sites, standards bodies and named-byline trade press. This worked well for holding-company and platform filings and poorly for JavaScript-rendered IR portals (Alphabet, Dentsu, Havas, Magnite, PubMatic, Stagwell) and for registration-walled trade research (IAB, WARC, ANA, WFA, EMARKETER). The specific consequences are recorded in §13 and in data_gaps. This sector should be prioritised for re-run with search capacity.

1. Definition and boundaries

In scope. The buying, selling, measurement and production of paid commercial communication, plus the professional services layer around it:

  • Ad tech and programmatic — demand-side platforms, supply-side platforms, ad servers, exchanges, identity resolution, verification and the auction mechanics that clear it all.
  • Measurement and attribution — audience currencies, accreditation, marketing mix modelling, incrementality testing, clean rooms, brand-lift and outcome measurement.
  • Agencies and holding companies — media planning and buying, creative, production, PR, health, experiential, and the holding-company financial model.
  • Martech stacks — customer data platforms, campaign management, email and CRM marketing, content supply chains, marketing automation.
  • Brand strategy — positioning, brand equity measurement, brand-versus-performance allocation.
  • Search and social advertising economics — the unit economics of demand capture and demand creation on platforms.
  • CTV advertising — streaming ad supply, upfronts, biddable and programmatic transaction paths.
  • AI in creative production and media buying — generative production, agentic buying, AI-assistant advertising surfaces.

Explicitly out of scope, and who owns it.

Excluded Owner
The foundation models and compute underneath generative creative tools Sector 01 (AI & foundation models)
Marketing-adjacent enterprise software sold as general SaaS (CRM platforms as businesses) Sector 02 (Enterprise software)
Retail media as a retailer profit line — network revenue, retailer margin mix, the $200.4bn WARC market size Sector 12 (Retail & e-commerce)
Consumer brand product strategy, wholesale-versus-DTC, fashion trend cycles Sector 23
Media content production economics (studios, streamers as content businesses) Media sector
Venture funding dynamics as a phenomenon in itself Sector 11 (Startups & VC)

The boundary disputes, named.

  1. Retail media. Sector 12 covers it as a retailer P&L line and explicitly left the attribution methodology and agency/buyer economics to this sector. That is the division observed here: this dossier does not restate retail media revenue figures, and covers instead how closed-loop attribution works, what it does to measurement standards, and how buyers and agencies are priced for it. Sector 12's finding that WARC puts Amazon at 78% of US retail media and Walmart at 7.5% is treated as context, not re-derived.
  2. AI-assistant advertising. OpenAI's advertising business is an advertising business, and is covered here. OpenAI's corporate funding, valuation and model economics belong to sector 01. Where the two touch — token costs entering the agency P&L — this dossier covers the agency side only.
  3. Identity and data infrastructure. LiveRamp sits on the line between martech and enterprise data software. It is covered here because its function is advertising identity resolution and its acquirer is an agency holding company.
  4. Antitrust. The Google ad-tech case and the FTC's Amazon advertising case are covered here as sector-determining events. The broader platform-regulation picture (DMA, app stores, search defaults) is not.
  5. Brand as an accounting concept. Brand valuation and intangible-asset accounting are adjacent and not covered; brand strategy as a marketing discipline is.

2. Subcategories

# Subindustry What distinguishes it
1 Platform advertising (Google, Meta, Amazon) Owns the inventory, the auction, the targeting and the measurement; sells on self-reported performance; roughly 56-60% of US/North American ad revenue
2 Independent ad tech (DSPs, SSPs, ad servers) Sells neutrality and cross-platform reach; margin is a take rate on someone else's media; structurally squeezed from both ends
3 Agency holding companies Sells labour, negotiating scale, data assets and orchestration; revenue is fee and, increasingly, principal media margin
4 Independent and challenger agencies Sells specialism and the absence of holdco conflict; the fastest-moving cohort on AI pricing experiments
5 Measurement, verification and currencies Sells trust; accredited by the MRC; the one part of the value chain whose product is the absence of a conflict of interest
6 Martech and customer data infrastructure Sells software seats and data throughput; the layer being absorbed into agency holding companies (Epsilon, LiveRamp)
7 CTV and streaming advertising Premium video supply with a linear-TV pricing heritage and a programmatic delivery mechanism; contested transaction path
8 Creative production and content supply chain The part of the sector most directly exposed to generative AI; historically a low-margin, high-headcount business
9 Search and demand capture Bids against measurable commercial intent; the subcategory whose input (query volume and shape) is being altered by AI assistants
10 AI-assistant advertising New surface, sub-1% of market, disclosed revenue but collapsing unit prices, no accredited measurement
11 Media auditing and marketing procurement Advertiser-side counterweight; small revenue, disproportionate influence on contract structure
12 Out-of-home, audio and print The residual legacy channels; DOOH forecast at -0.4% for 2026 by IAB, the only channel it expects to shrink alongside non-CTV digital video

3. Market structure

Concentration: winner-take-most at the platform layer, oligopoly at the agency layer, fragmented and eroding everywhere in between.

The defining structural fact of 2026 is that the sector's growth is real and its intermediaries are not capturing it.

  • Market size, with modellers named. IAB and PwC put US digital advertising revenue at $294.6bn for 2025, up 13.9% (30th annual report, released 2026-04-16). IAB raised its 2026 US ad spend growth forecast to 12.3% from 9.5% on 2026-09-10, on a survey of more than 200 brand and agency decision-makers. Madison & Wall models global ad spend growth of 11% in 2026 (9.8% excluding US political) to over $1.3tn, with a US market of about $479bn excluding political. WARC models global growth at about 9%. Treat all four as estimate or forecast; none is an audited census.
  • The platform tier. Madison & Wall estimates Google, Meta and Amazon took 56% of US advertising revenue in 2025 (Google 29%, Meta 19%, Amazon 9%), up from 53% in 2024, with everyone else falling from 47% to 43%, and about 60% of North American ad revenue in 2026. Meta's Q2 2026 advertising revenue of $59.363bn (+27%) came with impressions +14% and average price per ad +12% — more volume and higher prices simultaneously.
  • The agency tier. Four global groups plus one recently-listed fifth. Omnicom's acquisition of Interpublic closed 2025-11-26, taking the market from five majors to four. Dispersion inside that oligopoly is now extreme: in H1/Q2 2026, Omnicom +6.1% organic, Publicis +4.7% organic, WPP -4.7% like-for-like.
  • The independent ad tech tier. The Trade Desk, the largest independent DSP, grew 3% in Q2 2026 against a market forecast to grow 12.3%, and guided Q3 to at least $650m against $715m delivered in Q2. This is the clearest available measure of the squeeze.

Where margin actually sits.

  1. Platforms hold the structural margin, because they own supply, demand, the auction and the measurement of their own performance. Meta's operating margin fell to 31% in Q2 2026 — but on AI capex, not on ad-business competition.
  2. Holding companies earn a fee margin in the mid-to-high teens on net revenue (Publicis 17.5% headline in H1 2026, Omnicom 17.8% adjusted EBITA in Q2 2026) and a materially lower one when the business is in decline (WPP 8.4% headline in H1 2026, guided to 12-13% for the full year). Increasingly, margin also comes from principal media — buying inventory as principal and reselling it — which Ebiquity finds 96% of advertisers are aware of, 58% use, and 90% are uncertain whether their agency's recommendations on it serve the advertiser's interests.
  3. Independent ad tech earns a take rate that is being competed to zero at the top: Amazon charges near-zero DSP fees on programmatic guaranteed deals against 15-20% at competitors, bundling proprietary shopping data as the differentiator.
  4. Measurement and auditing earn small absolute revenue with disproportionate influence, because they set the terms on which everyone else is paid.

Barriers to entry. Low at the service layer — anyone can start an agency — and close to insurmountable at the platform layer, where the barrier is a first-party audience with transactional intent plus the ability to measure oneself. The middle layer has the worst of both: high capital intensity, no proprietary demand, and customers who can disintermediate.

Who has pricing power. Platforms, demonstrably: Meta raised average price per ad 12% in a single year while growing impressions. Holding companies have negotiating scale but not pricing power — WPP's CEO conceded on 2026-08-06 that a shift away from time-and-materials "will take a few years" and named exactly one client already on outcome-based pay. The independent middle has none: Amazon's fee structure sets the ceiling.

4. Who matters

Leading companies

Platforms

Notable independents and specialists

Active investors

  • Disclosed venture capital into this sector is thin. The capital that matters in 2026 is strategic M&A (Publicis/LiveRamp at $2.167bn enterprise value; Omnicom/IPG) and private equity in independent marketing services (New Mountain Capital, Carlyle, EQT and Falfurrias have all been reported as backers of named independents, but none of those positions could be verified from primary sources during this research and they are recorded as unverified).

Standards bodies and platforms

Regulators

Research institutions

Trade organisations

Consumer and civil-society groups

  • Electronic Frontier Foundation, Center for Digital Democracy and the Norwegian Consumer Council (Forbrukerrådet) remain the most consequential critics of behavioural advertising practice. None was verified against primary sources during this research and none is recorded as an entity.

5. Products, business models, technologies, customers

Major products. Auction-based advertising inventory (search, social, retail, video, display, CTV, audio, out-of-home); buying and selling software (DSP, SSP, ad server, curation); identity and data collaboration; measurement and verification services; creative and production services; media planning, buying and orchestration; marketing automation and customer data platforms.

How money is actually made today.

  • Platforms take a clearing price per impression or click. The economics are set by auction design, which is precisely why auction design is now the regulatory battleground.
  • Holding companies earn fees (retainers, FTE-based, project) plus commissions plus principal-media margin plus technology licensing. The mix is shifting toward the last two, which is the substance of the transparency debate.
  • Independent ad tech earns a percentage take rate on media flowing through the platform. Amazon is compressing that to near zero on guaranteed deals.
  • Measurement earns subscription and project fees, with accreditation as the moat.

How that is changing.

  1. From labour to output, slowly and with almost no contract evidence. Outcome-based remuneration is the most-discussed replacement for time-and-materials. WPP's CEO named exactly one client (Jaguar Land Rover) on 2026-08-06 and said broad transition "will take a few years." Independent agencies surveyed in 2025 were mostly still billing hourly.
  2. A new variable input cost has appeared in the agency P&L. Model tokens. Dept reported staff "burning through 1.5 million tokens in a day" and deployed an AI gateway; PMG built daily token caps; Brainlabs runs a tiered allowance; Rise uses PubMatic audit logs to catch buying agents drifting from parameters. Omnicom's CFO, asked about client reimbursement of token costs: "It's early days there. There isn't much of that yet." Until that is resolved, AI adoption is margin-dilutive.
  3. Data assets are becoming the differentiator and are going captive. Publicis is acquiring LiveRamp for $2.167bn enterprise value ($2.546bn equity value, $38.50 per share, 29.8% premium, announced 2026-05-17), folding the largest neutral identity spine into a single holding company alongside Epsilon.
  4. Rebundling. The WFA/Ebiquity 2026 Media Budgets Survey finds 75% of global marketers plan deeper integration between media and creative. Named consolidations include Bayer's reported $752m move to IPG, Mars' reported $1.7bn restructuring toward Publicis, and Jaguar Land Rover's reported $500m consolidation with WPP.

Technologies that matter. Real-time bidding and auction mechanics (first-price, unified pricing, reserve pricing); identity resolution and clean rooms; marketing mix modelling and geo-experimentation; generative creative production; agentic media buying; and — newly — AI-assistant answer surfaces as an inventory type.

Customer segments and what they buy on.

  • Large CPG and retail advertisers buy on reach, price per unit of attention, and increasingly on integrated media-and-creative delivery. PepsiCo's 2026 move to Publicis (reported $1.1bn net billings gain) is the canonical example.
  • Performance and DTC advertisers buy on measured incremental return, which is exactly the metric that signal loss has degraded.
  • Mid-market advertisers buy through self-serve platform tools and are the fastest-growing user base of AI-assistant advertising, where self-serve opened across 31 European markets on 2026-08-31.
  • B2B advertisers buy on pipeline influence, and are structurally the most exposed to AI-assistant-mediated research displacing search-driven demand capture.

6. Geography

Demand concentrates in the US. Omnicom took 59% of Q2 2026 core revenue from the United States; Publicis took €4,434m of €7,229m — 61% — from North America in H1 2026. IAB's forecast is a US number; Madison & Wall's concentration series is a US and North American number. Any global claim in this sector that is sourced to IAB or Madison & Wall is a US claim wearing a global label.

Capital concentrates in the US, corporate control in Europe and Japan. Three of the four major holding companies are European (Publicis — France; WPP — United Kingdom; Havas — Netherlands/France) and the fourth-largest globally is Japanese (Dentsu). The largest is American (Omnicom). This matters because the sector's most reliable financial disclosure — the holding-company results this dossier uses as its backbone — is produced under AMF, FCA, Euronext Amsterdam and TSE rules, not only SEC rules.

Regulation concentrates in the US and the EU. The two live antitrust matters are both American (EDVA and the FTC). The behavioural-remedy precedent practitioners cite is French: the Autorité de la concurrence's 2021 ad-tech decision, which Arnaud Créput characterised in September 2026 as having "not necessarily eliminated the underlying incentives and structural conflicts." The EU AI Act's transparency obligations took effect 2026-08-02 and bear directly on disclosure of AI-generated creative.

Non-US market, with regional sources.

  • France / Publicis Groupe (primary source, in French and English): H1 2026 net revenue €7,229m, +4.7% organic; Europe €1,774m at +4.5%, Latin America €217m at +12.0%, and Middle East & Africa €191m at -6.8% — the only region in decline, consistent with the macro brief's note that Middle East conflict is an active input. FY2026 guidance raised to +4.5% to +5%; free cash flow guidance approximately €2.2bn; net debt €1,215m at 30 June 2026.
  • United Kingdom / WPP plc (primary source): H1 2026 revenue £6,373m, -3.2% LFL; revenue less pass-through costs £4,745m, -4.7% LFL; UK Q3 2025 had been the worst region at -8.9%. Headcount down from 108,044 at end-2024 to 97,388 at 2026-06-30.
  • United Kingdom / John Lewis Partnership: AI agentic searches rose from 0.3% to 2.5% of site visits within one year — one of the few named, dated, advertiser-side datapoints anywhere on agentic traffic, and it comes from a British retailer rather than a US vendor.
  • Japan / Dentsu Group: headcount essentially flat (67,667 at end-2024; 67,454 at end-2025) with an international reduction programme of 3,400 roles, about 3,000 completed by mid-2026. Dentsu received a cease-and-desist order from Japan's Fair Trade Commission, reported on its own news pages in 2025 — a reminder that the Japanese market has its own competition history independent of the US cases.

What is missing geographically. China is entirely absent from this dossier. Alibaba, ByteDance, Tencent and Baidu together run an advertising market of comparable scale to the US one, with a completely different structure (no independent DSP layer, no MRC-equivalent, a different regulatory regime), and no source retrieved during this research covered it. India, Brazil and Southeast Asia are similarly absent except as lines in Publicis' regional table. This is recorded as a first-order data gap.

7. Historical trend patterns

This sector produces more confident false positives per decade than almost any other, because its commentary is produced by the same firms that sell into it. The pattern is consistent enough to be a rule: a deadline-based trend with a vendor category attached and no enforcement mechanism does not resolve as predicted.

What has actually cycled, 2000-2026.

  1. Disintermediation panics, roughly every five years. 2007: Google would sell media directly and agencies would die. 2016-2018: consultancies (Accenture, Deloitte, IBM iX) would absorb agencies. 2018-2020: in-housing would strand the holding companies. None resolved as predicted. What actually happened in 2026 is stranger and more interesting: the holding companies did not die, they diverged — Omnicom +6.1% organic, Publicis +4.7%, WPP -4.7% in the same quarter and the same economy.
  2. Measurement crises followed by standards, roughly every seven years. The 2014 viewability crisis produced MRC viewability standards. The 2016 ANA/K2 media transparency report produced a decade of contract reform, with working media ratios improving from 36% to 43.3% — real progress, far slower than promised, and accompanied by entirely new forms of opacity in walled gardens and offshore trading intermediaries. The 2026 version is the MRC's Digital Advertising Auction Transparency Standards (2026-01-29).
  3. Identity crises driven by platform policy, not regulation. Safari third-party cookie blocking (2017-2020), Apple ATT (2021), and the Chrome deprecation saga (2020-2025).
  4. New surface hype cycles. Second Life brand islands (2007). "The year of mobile" (declared annually 2008-2013). QR codes (failed 2011, succeeded 2020 — the sector's one genuine reversal). Clubhouse and social audio (2021). NFTs and metaverse activations (2021-2022). Each produced a vendor category, a Cannes panel track and an agency practice, and each except QR codes produced no durable revenue line.

Specific past false positives in this sector, named.

  • Third-party cookie deprecation (2020-2025). Announced January 2020, delayed in 2021, 2022, 2023 and 2024, and ultimately abandoned as a forced migration. Six years of product roadmaps, an entire venture-funded "cookieless identity" category, and hundreds of agency transition practices were built on a date that never bound. This is the single best calibration case this sector offers, and it is directly analogous to the macro brief's finding that the EU AI Act's most-cited 2026 deadline was deferred to 2027-12-02. The identity consolidation the deprecation was supposed to force is happening anyway — Publicis is buying LiveRamp — but for entirely different reasons and on a different timetable.
  • "Ad blocking will end the ad-funded web" (2015-2016). Adoption plateaued; the ad-funded web is larger than ever at $294.6bn of US digital revenue in 2025.
  • Attention metrics as the successor currency (2022-2024). A dozen vendors, several standards initiatives, near-zero migration of actual currency.
  • Agency break-up theses. S4 Capital was positioned from 2018 as the digital-native replacement for the holding companies; the holding companies consolidated instead.

What the cycle history tells you about 2026's claims. The two loudest current claims — "AI is collapsing creative production costs" and "AI search is destroying demand capture" — have the exact signature of prior false positives: vendor-originated, deadline-free, quantified only by parties selling the remedy, and not yet visible in any audited financial statement. That does not make them wrong. It makes them unproven, and the sector's own history says the correct posture is to demand the counterfactual.

8. What is changing now (as of 2026-09-15)

The market is growing and the middle is not capturing it. IAB raised its 2026 US ad spend growth forecast to 12.3% from 9.5% on 2026-09-10 — attributing the revision, in its own account via the ANA, principally to a stronger-than-expected first half driven by the Winter Olympics and the FIFA World Cup, not to AI. In that market, The Trade Desk grew 3% and WPP shrank 4.7%. Growth is accruing to Alphabet, Meta, Amazon, and to whichever holding company is currently winning share from the others.

The antitrust question resolved in favour of the incumbent, and is not yet fully public. On 2026-09-02 Judge Leonie Brinkema (EDVA) rejected divestiture of AdX and DFP and declined to order open-sourcing of the DFP auction logic, adopting most of the proposed behavioural remedies as modified by the court. Her reasoning was sealed in a confidential memorandum for 14 days — meaning it becomes public on or about 2026-09-16, one day after this research date — and the parties have 30 days to submit a proposed final judgment. DOJ said it is evaluating next steps and expressed plans to pursue appeals. This is the sector's direct equivalent of the macro brief's unresolved 2026-09-16 FOMC decision: do not assert the content of the opinion.

Enforcement has moved from structure to auction mechanics. On 2026-08-26 the FTC and 22 state attorneys general sued Amazon over an alleged "soft reserve price" feature — described internally, per the complaint, as an "invented auction participant" — introduced in late 2018, alleging that by 2024 about 80% of winning bidders paid nearly their full bid despite a publicly-described second-price auction, and estimating roughly $20bn of revenue since 2019. Amazon says advertisers saved over $8bn from 2021 to 2025. Amazon DSP is explicitly excluded. The MRC had published auction transparency standards seven months earlier.

A genuinely new advertising surface exists, and its prices are falling. OpenAI's ChatGPT advertising business reached a $1bn annualised run rate within 200 days (roughly $83m a month; roughly $330m actually booked through August 2026 on the reporting outlet's own estimate), against internal targets of $2.5bn for 2026 and $100bn by 2030. CPMs fell from $60 at launch to as low as $25 within ten weeks; cost-per-click went live 2026-04-21 at $3-$5 bids. Self-serve opened across 31 European markets on 2026-08-31, and Amazon DSP began selling ChatGPT inventory on 2026-09-10.

Post-merger integration is being paid for in headcount, not in client-facing capacity. Omnicom is taking group headcount from about 120,000 to about 105,000 during 2026 against a $1.5bn gross cost-synergy target, reinvesting 20-30% of first-year synergies in platform and agentic AI. WPP cut about 1,260 roles in H1 2026 and is down 8.1% year over year. Publicis, alone, added people.

Measurement has rebuilt itself and still does not decide budgets. Around eight in ten advertisers run MMM, only about a third continuously, and only 15.4% say effectiveness outputs are the primary driver of budget decisions (Ebiquity/WFA). Meanwhile the input data is degrading from a new direction: Cloudflare reports automated agents and bots exceed half of all web requests, and AI agents defeat conventional bot filters by adding cart items and signing up for newsletters.

Macro context applied. The macro brief's sticky-inflation, hawkish-lean environment is visible here in two specific ways. First, advertising is a discretionary, cyclically-sensitive budget line, and a 12.3% growth print driven by a sporting calendar is not the same as underlying demand strength — 2027 laps the Olympics and the World Cup. Second, the macro brief's warning that aggregate capital figures conceal median experience applies precisely: the sector's headline growth conceals a middle tier in outright decline.

9. The five lists

Five most important current trends

  1. Behavioural-only remedy in the Google ad-tech case, with the opinion still sealed (T-16-01)
  2. Alphabet, Meta and Amazon taking a widening majority of US ad revenue (T-16-02)
  3. Holding-company consolidation paid for in headcount (T-16-03)
  4. Measurement retreating to MMM and still not driving budgets (T-16-05)
  5. The independent DSP stalling: Trade Desk at 3% in a 12.3% market (T-16-07)

Five fastest-growing signals

  1. Advertising inside AI assistants, with disclosed revenue and collapsing prices (T-16-09)
  2. Agentic and bot traffic breaking attribution and audience models (T-16-10)
  3. AI-visibility (GEO/AEO) measurement as a martech category (T-16-11)
  4. AI agents as a media-buying cost centre and the audit tooling against them (T-16-13)
  5. Auction transparency as the new regulatory front — FTC v. Amazon plus the MRC standard (T-16-14)

Five trends most likely to affect businesses

  1. Alphabet/Meta/Amazon concentration (T-16-02) — also in list 1
  2. Agency revenue per head rising on headcount cuts, not AI leverage (T-16-04)
  3. The identity layer going captive at Publicis (T-16-08)
  4. The independent DSP stall (T-16-07) — also in list 1
  5. Auction transparency enforcement (T-16-14) — also in list 2

Five trends most likely to affect consumers

  1. Advertising inside AI assistants (T-16-09) — also in list 2; the assistant many people now treat as a neutral adviser is becoming an ad-funded surface
  2. Search demand capture re-pricing as query shape changes (T-16-12)
  3. CTV ad supply growth — ad minutes per hour up 18% year over year (T-16-06)
  4. Agentic traffic and the erosion of the human-audience assumption (T-16-10) — also in list 2
  5. Generative creative production changing what advertising looks like and who makes it (T-16-20)

Overlaps, stated plainly. T-16-02, T-16-07, T-16-09, T-16-10 and T-16-14 each appear twice. That is not padding: concentration, the independent squeeze and AI-assistant advertising are simultaneously the most important current facts, the fastest-moving signals and the trends with the largest business and consumer consequences. The absence of a trend from the consumer list is meaningful — most of this sector's 2026 change is B2B plumbing.

10. Overhyped / overlooked / cooling / reversing

Most overhyped

1. "AI search is destroying demand capture" as a budget argument (T-16-19). The claim is being used inside marketing organisations to justify larger media budgets and the supporting evidence is thin. The specific evidence that the hype outruns the substance:

  • State Farm, on the record in the same article making the argument, says budgets "have been essentially static for several years."
  • The survey figure being cited — 61% of US consumers used an AI assistant for shopping research in the past three months (Dept) — measures assistant adoption, not traffic or revenue loss. There is no counterfactual in it.
  • The forecast upgrade being attributed to AI-search adaptation is attributed by the forecaster itself to the Winter Olympics and the FIFA World Cup.
  • The two quantitative datasets that do exist both explicitly disclaim causation. Smarter Ecommerce's Mike Ryan on the Shopping-ads data: "The data doesn't prove AI Overviews are causing the change."

2. Vendor-published AI creative "lift" and the collapse of production cost (T-16-20). The strongest independent evidence points to parity, not lift: a Columbia University study of more than 300,000 ads, reported by WARC, found GenAI-made ads perform as well as or better than human-made ones — with the highest click-through rates going to AI ads that were perceived as human-made. On cost, the only concrete public figure found was a single agency-self-reported campaign at 26% of a live-shoot budget, in reporting that noted for other agencies "the exact cost saving is unclear." And the decisive test: if the cost collapse were real at scale it would appear as margin expansion at the companies doing the producing, and it does not. WPP's H1 2026 headline margin was 8.4% (up 0.2pt LFL) and Publicis' 17.5% (up 17bp) — both roughly flat, both while spending heavily on AI. Meanwhile agencies are absorbing a new unbilled cost in model tokens.

3. Honourable mention: AI-visibility metrics (T-16-11). A vendor category has formed around a measurement whose own practitioners concede that "the same prompt can return a different answer nearly every time" and that a model upgrade resets every tracked metric overnight. Budgets are moving toward a number with unquantified variance and no accreditation.

Most overlooked

1. Revenue per head is the wrong way round, and nobody is saying so (T-16-04). The entire AI-agency-economics narrative assumes AI raises output per employee. The disclosed financials say the opposite is happening: the holding company that is shrinking (WPP) shows rising revenue per head because headcount fell 8.1% while revenue fell 5.6%, and the holding company that is growing (Publicis, +4.7% organic) grew headcount from 108,179 to about 114,000 over eighteen months. The fastest-growing agency in the sector is growing the old-fashioned way. Attention has missed this because almost nobody computes the ratio — the commentary compares AI narratives, not denominators.

2. The MRC's auction transparency standard predates the litigation it will be measured against. Published 2026-01-29, seven months before the FTC's Amazon complaint and seven months before the Google remedies ruling. A standards body has quietly written the benchmark that both behavioural remedies and advertiser audit rights will be interpreted against, and it received almost no coverage because standards bodies do not issue press releases that trend.

3. Token cost is a new, unbilled, variable input in a business that has never had one. Agencies have always had one cost: people. They now have two, and the second is consumed by individual employees choosing unnecessarily powerful models. Omnicom's CFO: "It's early days there. There isn't much of that yet." The first holding company to solve billing for compute gets a structural margin advantage, and nobody is tracking it.

4. Latin America at +12.0% organic. Publicis' single fastest-growing region in H1 2026, larger than its growth in North America or Europe, in a market almost nobody in this sector produces research about.

Trends that appear to be cooling

  • Structural break-up as the expected antitrust outcome (T-16-17). Indicator that turned: the 2026-09-02 ruling rejecting AdX and DFP divestiture outright. Three years of sell-side strategy priced on a forced divestiture now has no catalyst.
  • Multi-touch attribution and the deterministic user graph (T-16-16). Indicator that turned: the acquisition of LiveRamp, the last major neutral identity spine, by an agency holding company. Once the identity layer is owned by a buyer, cross-platform deterministic attribution stops existing as a neutral product category.
  • Third-party cookie deprecation as a forcing deadline (T-16-18). Indicator that turned: the disappearance of "cookieless" as a standalone vendor category, and the replacement of a single deprecation event with continuous, multi-source signal degradation in IAB's own framing.

Trends that may reverse, and the mechanism

  • The Google remedy could reverse via appeal. DOJ has expressed plans to appeal. The mechanism is straightforward: an appellate court that credits the divestiture record could reinstate structural relief, re-pricing every independent SSP in the sector. Falsifiable early indicator: a notice of appeal filed within the statutory window after the final judgment.
  • CTV price deflation could reverse through supply discipline. Ad minutes per hour rose 18% year over year, which is disinflationary — but Disney reports 70% of biddable demand originating from upfront advertisers, which is the mechanism by which sellers recapture pricing. Falsifiable early indicator: upfront volume commitments in the 2027 cycle rising while scatter CPMs hold.
  • The independent DSP stall could reverse if the remedies bite. If the final judgment genuinely exposes AdX bids to rival ad servers and mandates Prebid parity, the independent sell-side gains real inventory access for the first time. Falsifiable early indicator: two consecutive quarters of accelerating revenue at Magnite, PubMatic or The Trade Desk with open-web display, not CTV, as the stated driver.
  • AI-assistant advertising could reverse into irrelevance. A surface whose CPM more than halved in ten weeks is not demonstrating pricing power. Falsifiable early indicator: OpenAI's disclosed run rate failing to roughly double by mid-2027, against a stated $2.5bn 2026 target it is currently well short of.

11. Risks and major uncertainties

Sector-specific risks.

  1. Cyclicality dressed as structural growth. The 2026 US forecast upgrade is attributed by the forecaster to the Winter Olympics and the FIFA World Cup. 2027 laps both. A sector reading 12.3% as a new baseline is setting itself up for a hard 2027 comparison, in a macro environment the brief describes as sticky-inflation with a hawkish-lean Fed.
  2. Post-merger client conflict. Omnicom lost PepsiCo to Publicis; Publicis then had to exit Coca-Cola's reported $805m North American media account on conflict. Consolidation mechanically creates conflicts that mechanically shed revenue, and cost synergies are disclosed while conflict losses are not.
  3. Compute cost entering an unpriced contract base. Agencies are absorbing token costs under master service agreements written for a labour-only cost structure, and are amending with addenda rather than repricing.
  4. Measurement legitimacy. If more than half of web requests are automated and a growing share mimic high-intent behaviour, the audience models underpinning digital advertising are being contaminated in a way that existing invalid-traffic frameworks were not designed to detect. A single credible study quantifying agentic contamination of billable impressions would be a sector-level shock.
  5. Regulatory whiplash in both directions. The Google remedy went softer than expected; the FTC's Amazon case went harder than expected. Neither direction is stable.
  6. Concentration risk for everyone downstream. At 56-60% of North American ad revenue in three companies that measure their own performance, every independent vendor's addressable market is shrinking regardless of execution.

Genuine unknowns — "we don't know" versus "nobody can know."

We don't know (answerable, and the answer exists somewhere):

  • The exact content of the Google remedies opinion. It was sealed and becomes public on or about 2026-09-16. This is a one-day-away known unknown.
  • Alphabet's Q2 2026 advertising segment revenue. It is disclosed in the 10-Q; this research could not retrieve it because the IR site is JavaScript-rendered and EDGAR's browse interface was unavailable.
  • Whether AI has changed agency productivity. The answer is computable from disclosed revenue and headcount at four companies; the obstacle is that revenue definitions differ across them.
  • The real size of AI-assistant advertising. OpenAI knows. It has disclosed a run rate, which is not the same thing.
  • Chinese advertising market structure and scale. Well-documented in Mandarin-language sources this research could not reach.

Nobody can know (genuinely indeterminate):

  • Whether AI assistants substitute for or complement commercial search at scale. The behaviour is two years old, the assistants change monthly, and the measurement instruments are themselves unstable.
  • Whether behavioural remedies change conduct. The French 2021 precedent suggests not; the US order is differently constructed; there is no controlled experiment.
  • The equilibrium price of attention in an environment where creative production is near-costless. Classical models assume creative supply is constrained by production cost.
  • Whether brand-building retains its measured effectiveness when a growing share of "consumers" are agents acting on stated preferences.

12. Scenarios to 2030

Base case — "concentration grinds on, the middle thins." Platform share continues its roughly three-point-a-year climb toward two-thirds of North American ad revenue. Behavioural remedies are implemented, comply on paper, and change marginal publisher economics without changing market shares. Holding companies stabilise around four global groups with Publicis and Omnicom taking share from WPP and Dentsu, and sector-wide headcount continues a slow decline while revenue per head rises on the denominator. AI-assistant advertising reaches a real but unremarkable few billion dollars. Falsifiable early indicator: Madison & Wall's big-three share series printing 62-64% for North America in 2027 while total market growth normalises to mid-single digits post-Olympics.

Upside — "interoperability actually works." The final judgment's real-time-bid exposure and Prebid parity requirements meaningfully open AdX supply to rival ad servers. Independent SSP and DSP revenue re-accelerates; open-web publisher yields improve; the MRC auction standard becomes the de facto audit benchmark and advertisers gain enforceable transparency for the first time since 2016. Falsifiable early indicator: two consecutive quarters of accelerating open-web display revenue at Magnite or PubMatic, with the companies naming remedy compliance as the driver.

Downside — "the 2027 comparison bites." 2027 laps the Winter Olympics and the FIFA World Cup into a sticky-inflation, restrictive-rate economy. Advertising, the most cuttable line in a CFO's budget, contracts. Holding companies that cut to 105,000 and 97,000 heads discover they cut capacity, not overhead. The independent middle consolidates or exits. Falsifiable early indicator: IAB's January 2027 outlook forecasting US growth below 5%, or any major holding company guiding to negative organic growth for 2027.

Disruption — "the assistant becomes the shelf." AI assistants become the primary commercial discovery surface for a material share of categories. Click-to-chat formats keep the transaction inside the assistant, publishers lose the referral click entirely, and demand capture re-prices around conversational placement. Search advertising does not collapse but stops growing, and a fourth platform pole forms. Falsifiable early indicator: OpenAI's disclosed advertising run rate exceeding $5bn, or any top-20 US advertiser disclosing a double-digit percentage of its performance budget allocated to assistant surfaces.

Regulatory — "auction transparency becomes enforceable." The FTC prevails against Amazon on the auction-mechanics theory, the MRC standard becomes the compliance benchmark, and mandatory auction disclosure spreads to every major platform. The opacity margin that has funded parts of the value chain since 2010 compresses, and advertiser audit rights become contractually routine. Falsifiable early indicator: a settlement or adverse ruling in FTC v. Amazon that includes auction-disclosure obligations, or a second platform voluntarily adopting the MRC auction standard.

Failure — "measurement legitimacy collapses." Agentic traffic contamination becomes quantified and large. A credible study shows a significant share of billable digital impressions were served to automated agents. Advertisers respond by shifting budget to environments with verifiable humans — retail media with logged-in users, live sports, out-of-home — and the open programmatic web loses its funding model for the second time in a decade. Falsifiable early indicator: an MRC or major verifier publishing an agentic-traffic invalidity rate above 10% for open-web display, or a top-10 advertiser publicly withholding payment on agent-served impressions.

13. Data gaps and limitations

Reduced discovery capability. The shared WebSearch budget was exhausted after five searches by this agent. All subsequent discovery used WebFetch against primary sources. This materially reduced the ability to find sources, though not to verify the ones found. The sector should be prioritised for re-run.

What could not be verified.

  1. Alphabet's Q2 2026 advertising segment revenue. Every route failed: abc.xyz IR pages are JavaScript-rendered and returned navigation shells; SEC's browse-edgar CGI is robots-disallowed; direct PDF path guesses 404'd; CNBC returned 403. The largest single advertising seller's most recent revenue disclosure is therefore absent from this dossier and Google's share is carried only as Madison & Wall's modelled estimate. This is the single largest gap.
  2. The content of the Google remedies opinion. Sealed for 14 days from 2026-09-02, public on or about 2026-09-16 — one day after this research date. Every description of the remedies in this dossier is reconstructed by trade press from the remedies-trial record and a preview filing, and the two most detailed accounts (AdExchanger and Courthouse News) differ in specificity. Recorded as a contradiction on T-16-01.
  3. Dentsu's 2026 financial results. Four separate English IR URL paths returned 404 or redirect loops; the English news page showed nothing after mid-2025. Dentsu figures in this dossier are limited to headcount, sourced from a third-party compilation of filings. The sector's principal non-Western holding company is under-covered here.
  4. Havas H1 2026 financials. IR paths 404'd or redirected to a French newsroom shell. Only headcount is carried.
  5. Magnite, PubMatic and Stagwell Q2 2026 results. IR pages unreachable. The independent sell-side is represented only qualitatively, which is a real weakness given that the Google remedy's principal beneficiaries would be exactly these companies.
  6. IAB, WARC, ANA and WFA full research bodies. All registration- or subscription-walled. Only headline figures were retrievable. Specifically missing: IAB's channel-level Outlook methodology, its State of Data adoption percentages, its Digital Video report's CTV pricing detail, and the WFA/Ebiquity Paid Media Effectiveness Handbook sample. Phase 2 licensing budget should go to WARC first and IAB account access second.
  7. Any audited CPM index for CTV or any channel. None exists publicly. CTV pricing direction in this dossier is inferred from ad-load growth and seller-disclosed upfront share, which is an inference, not a measurement. This is a structural gap in the sector, not a research failure.
  8. China, India, Brazil and Southeast Asia. Entirely absent. No non-English source was reachable for the Chinese advertising market, which is of comparable scale to the US one and structurally different. The regional-source requirement was met with French, UK and Japanese sources (Publicis, WPP, John Lewis, Dentsu), but Asia ex-Japan is a void.
  9. The Columbia University GenAI advertising study. Cited here via WARC's reporting; the original paper, its authors, its sample construction and its controls were not retrieved. It is the most important independent evidence in the dossier and it is second-hand.
  10. Autorité de la concurrence's 2021 decision text. The French precedent is characterised here from practitioner commentary in AdExchanger, not from the regulator's own document.
  11. Agency revenue-per-head comparability. Omnicom reports revenue, WPP reports revenue less pass-through costs, Publicis reports net revenue, and headcount is variously average and period-end. Cross-company revenue-per-head comparison is therefore invalid and is not attempted; only within-company year-over-year movement is used.
  12. OpenAI advertising: everything is single-source (Digiday) and ultimately company-disclosed. No filing, no audit, no independent verification of delivery.

Where the numbers conflict. Recorded formally in the trend records: LiveRamp's price ($2.167bn EV vs $2.546bn equity vs "$2.2bn"/"$2.16bn"/"$2.5bn" in headlines); the big-three share series (56% US 2025 vs ~60% North America 2026, same modeller); WPP headcount (average vs period-end); the FTC's $20bn against Amazon's $8bn; IAB's +12.3% US against WARC's ~9% global; the cause of the 2026 forecast upgrade (AI-search adaptation per buyers, sporting calendar per the forecaster); and Shopping-ad CTR (+20% Smarter Ecommerce vs +17% Optmyzr).

A note on source dependency. This sector's hazard is that almost every widely-cited statistic originates with a party selling into it. The countermeasure applied here is that the dossier's backbone is holding-company and platform financial disclosure — numbers that are audited, filed under securities law, and hard to spin. Where that backbone does not reach (AI-assistant pricing, agentic traffic, AI visibility, creative cost), claims are labelled estimate, opinion or marketing and confidence is set to low or medium accordingly.

14. Ranking scorecard

Criterion Score Justification
speed_of_change 4 Auction rules, holdco share, and an entirely new ad surface all moved inside twelve months; the underlying cast (three platforms, four holdcos) is stable, so not a 5.
economic_importance 4 Over $1.3tn of global spend (Madison & Wall) and $294.6bn of US digital revenue alone, but it is derived demand on other sectors' P&Ls rather than primary output.
capital_invested 2 Disclosed venture capital into ad tech and martech is negligible in 2026; the capital that moved was strategic M&A ($2.167bn LiveRamp) and platform capex belonging to other sectors.
company_product_density 5 Thousands of distinct martech and ad tech vendors plus hundreds of agencies; a 40-entity register barely samples it.
regulatory_impact 4 Two landmark live US matters, an MRC standards programme and EU AI Act transparency — but the September ruling showed regulation here changes conduct, not structure.
consumer_impact 4 Advertising funds most consumer media and is now funding AI assistants; the effect is pervasive but indirect and largely invisible to the person experiencing it.
strategic_importance 2 No national-security, infrastructure or supply-chain criticality. Commercially consequential, strategically peripheral.
intelligence_demand 5 The sector runs on purchased trend intelligence; forecasts are news events in their own right, as the IAB September revision demonstrated.
paid_research_opportunity 5 An unusually large incumbent research market — WARC, EMARKETER, Ebiquity, Kantar, Nielsen, MRC, Forrester — with demonstrated enterprise willingness to pay.
data_availability 3 Excellent audited holdco and platform filings; but no public CPM index, heavy paywalling at every trade body, JavaScript-only IR portals, and most quoted statistics produced by sellers.
cross_industry_influence 4 Sets the monetisation model for media, retail, consumer internet and now AI assistants; the transmission channel through which AI-surface economics reach the rest of the consumer economy.

15. Sources

  1. Omnicom Reports Second Quarter 2026 Results — Omnicom Group (Investor Relations) — https://s201.q4cdn.com/282904488/files/doc_financials/2026/q2/OMC-2026-Q2-Earnings-Release-07-28-26.pdf — 2026-07-28 — A
  2. Omnicom at Goldman Sachs Communacopia + Technology Conference 2026 (transcript) — Investing.com, transcribing Omnicom management — https://www.investing.com/news/transcripts/omnicom-at-goldman-sachs-communacopia--technology-conference-2026-integration-gains-93CH-4896507 — 2026-09-10 — B
  3. Omnicom Completes Acquisition of Interpublic — Omnicom Group via PR Newswire — https://www.prnewswire.com/news-releases/omnicom-completes-acquisition-of-interpublic-forming-the-worlds-leading-marketing-and-sales-company-built-for-intelligent-growth-in-the-next-era-302627141.html — 2025-11-26 — A
  4. 2026 Interim Results — WPP plc — https://www.wpp.com/en/news/2026-interim-results — 2026-08-06 — A
  5. Strategy Update and 2025 Preliminary Results (Elevate28) — WPP plc — https://www.wpp.com/en/news/2026/02/strategy-update-and-2025-preliminary-results — 2026-02-26 — A
  6. Publicis Groupe: First Half 2026 Results — Publicis Groupe via GlobeNewswire — https://www.globenewswire.com/news-release/2026/07/16/3328221/0/en/publicis-groupe-first-half-2026-results.html — 2026-07-16 — A
  7. Publicis to acquire LiveRamp to accelerate data co-creation for smarter agents — Publicis Groupe — https://www.publicisgroupe.com/en/news/press-releases/publicis-to-acquire-liveramp-to-accelerate-data-co-creation-for-smarter-agents — 2026-05-17 — A
  8. Meta Reports Second Quarter 2026 Results — Meta Platforms, Inc. — https://investor.atmeta.com/investor-news/press-release-details/2026/Meta-Reports-Second-Quarter-2026-Results/default.aspx — 2026-07-29 — A
  9. The Trade Desk Reports Second Quarter 2026 Financial Results — The Trade Desk, Inc. — https://www.thetradedesk.com/press-room/the-trade-desk-reports-second-quarter-2026-financial-results — 2026-08-06 — A
  10. MRC News — Digital Advertising Auction Transparency Standards (2026-01-29); AI Measurement Guidance (2026-07-08); Out-of-Home Standards (2025-12-04); accreditation status reports — Media Rating Council — https://mediaratingcouncil.org/news — 2026-07-08 — A
  11. IAB/PwC Internet Advertising Revenue Report: Full Year 2025 — IAB and PwC — https://www.iab.com/insights/internet-advertising-revenue-report-full-year-2025/ — 2026-04-16 — A (headline figures only; full report registration-walled)
  12. 2026 Outlook Study: September Update — IAB — https://www.iab.com/insights/2026-outlook-study-september-update/ — 2026-09-10 — A (registration-walled)
  13. 2026 IAB Digital Video Ad Spend & Strategy Report — IAB — https://www.iab.com/insights/video-ad-spend-report-2026/ — 2026-07-14 — A (registration-walled)
  14. State of Data 2026: The AI-Powered Measurement Transformation — IAB — https://www.iab.com/insights/2026-state-of-data-report/ — 2026-02-02 — A (registration-walled)
  15. The Next Phase of AI: From Adoption to Accountability (AI Insights Engine) — IAB — https://www.iab.com/insights/ai-insights-engine-july-2026/ — 2026-07-16 — A (registration-walled)
  16. Google Won't Have To Break Up Its Ad Tech Business, Judge Brinkema Rules — AdExchanger — https://www.adexchanger.com/antitrust/google-wont-have-to-break-up-its-ad-tech-business-judge-brinkema-rules/ — 2026-09-02 — B
  17. Google dodges antitrust breakup of ad tech business — Courthouse News Service — https://www.courthousenews.com/google-dodges-antitrust-breakup-of-ad-tech-business/ — 2026-09-02 — B
  18. What The FTC Can Learn From Google's Ad Tech Case As It Pursues Amazon — AdExchanger (Allison Schiff) — https://www.adexchanger.com/platforms/what-the-ftc-can-learn-from-googles-ad-tech-case-as-it-pursues-amazon/ — 2026-09-14 — B
  19. The FTC's Amazon Lawsuit Is Ad Tech's History Of Opacity Repeating Itself — AdExchanger (Anthony Vargas) — https://www.adexchanger.com/platforms/the-ftcs-amazon-lawsuit-is-ad-techs-history-of-opacity-repeating-itself/ — 2026-09-03 — B
  20. Google Had Its Day In Court. Now, It's Amazon's Turn — AdExchanger, The Big Story — https://www.adexchanger.com/the-big-story/google-had-its-day-in-court-now-its-amazons-turn/ — 2026-09-04 — B
  21. Ad spend forecasts revised upward as more ad dollars are handled by AI tools — Digiday (Sam Bradley) — https://digiday.com/media-buying/ad-spend-forecasts-revised-upward-as-more-ad-dollars-are-handled-by-ai-tools/ — 2026-09-10 — B
  22. Ad Tech Briefing: AI is reinforcing Big Tech's grip on advertising growth — Digiday — https://digiday.com/media-buying/ad-tech-briefing-ai-is-reinforcing-big-techs-grip-on-advertising-growth/ — 2026-09-15 — B
  23. OpenAI's ChatGPT ads business hits $1 billion run rate as Europe gets self-serve access — Digiday (Seb Joseph, Krystal Scanlon) — https://digiday.com/media-buying/openais-chatgpt-ads-business-hits-1-billion-run-rate-as-europe-gets-self-serve-access/ — 2026-08-31 — B
  24. OpenAI turns on cost-per-click ads inside ChatGPT — Digiday (Krystal Scanlon, Seb Joseph) — https://digiday.com/marketing/openai-turns-on-cost-per-click-ads-inside-chatgpt/ — 2026-04-21 — B
  25. Amazon Ads brings advertisers to ChatGPT, extending its supply-chasing streak — Digiday (Seb Joseph) — https://digiday.com/media-buying/amazon-brings-its-dsp-to-openais-chatgpt-ads-extending-its-supply-chasing-streak/ — 2026-09-10 — B
  26. Marketers face dilemma around rising bot and AI web traffic — Digiday (Sam Bradley) — https://digiday.com/media-buying/marketers-face-dilemma-around-rising-bot-and-ai-web-traffic/ — 2026-09-08 — B
  27. In Graphic Detail: Inside the scramble to measure a brand's AI visibility — Digiday (Seb Joseph) — https://digiday.com/marketing/in-graphic-detail-inside-the-scramble-to-measure-a-brands-ai-visibility/ — 2026-09-15 — B
  28. Marketers are using AI search concerns to argue for higher media budgets — Digiday (Sam Bradley) — https://digiday.com/media-buying/marketers-are-using-ai-search-concerns-to-argue-for-higher-media-budgets/ — 2026-09-15 — B
  29. Media agencies build audit tools to prevent AI agents from overcharging — Digiday (Sam Bradley) — https://digiday.com/media-buying/media-agencies-build-audit-tools-to-prevent-ai-agents-from-overcharging/ — 2026-09-04 — B
  30. Future of TV Briefing: Where have all the cord-cutters gone? — Digiday (Tim Peterson) — https://digiday.com/future-of-tv/future-of-tv-briefing-where-have-all-the-cord-cutters-gone/ — 2026-09-09 — B
  31. Future of TV Briefing: The upfront is overtaking streaming's programmatic marketplace — Digiday (Tim Peterson) — https://digiday.com/future-of-tv/future-of-tv-briefing-the-upfront-is-overtaking-streamings-programmatic-marketplace/ — 2026-05-20 — B
  32. As WPP's Cindy Rose marks a year as CEO, turnaround is underway but far from certain — Digiday (Sam Bradley) — https://digiday.com/media-buying/as-wpps-cindy-rose-marks-a-year-as-ceo-turnaround-is-underway-but-far-from-certain/ — 2026-08-31 — B
  33. 'It will take a few years': WPP CEO Cindy Rose says outcome-based pay is still years away — Digiday (Sam Bradley) — https://digiday.com/media-buying/it-will-take-a-few-years-wpp-ceo-cindy-rose-says-outcome-based-pay-is-still-years-away/ — 2026-08-06 — B
  34. AI reshapes agency economics, but their contracts are still scrambling to catch up — Digiday (Kimeko McCoy) — https://digiday.com/marketing/ai-reshapes-agency-economics-but-their-contracts-are-still-scrambling-to-catch-up/ — 2026-09-10 — B
  35. Media Buying Briefing: How Publicis keeps winning clients without pitches — Digiday (Sam Bradley) — https://digiday.com/media-buying/media-buying-briefing-how-publicis-keeps-winning-clients-without-pitches/ — 2026-09-14 — B
  36. As AI alters cost of creative, indie agencies review how they charge clients — Digiday (Sam Bradley) — https://digiday.com/marketing/as-ai-alters-cost-of-creative-indie-agencies-review-how-they-charge-clients/ — 2025-08-08 — B
  37. Cindy Rose says WPP performance 'unacceptable' as growth forecast downgraded again — Campaign (Shauna Lewis) — https://www.campaignlive.com/article/cindy-rose-says-wpp-performance-unacceptable-growth-forecast-downgraded-again/1937893 — 2025-10-30 — B
  38. Ten Years Later, Media Transparency Is Still Broken. Here's What Marketers Need to Do About It. — Ebiquity (Chantal Cooper) — https://ebiquity.com/news-insights/blog/ten-years-later-media-transparency-is-still-broken-heres-what-marketers-need-to-do-about-it/ — 2026-06-08 — B (media auditor; declared commercial interest)
  39. From measurement to management: the emerging role of MMM — Ebiquity (Chantal Cooper) — https://ebiquity.com/news-insights/blog/from-measurement-to-management-the-emerging-role-of-mmm/ — 2026-06-11 — B
  40. The rebundling era: How holding companies are reuniting media and creative — Ebiquity (Chantal Cooper) — https://ebiquity.com/news-insights/the-rebundling-era-how-holding-companies-are-reuniting-media-and-creative-to-meet-the-demands-of-modern-marketing/ — 2026-02-26 — B
  41. WARC News & Opinion — global ad spend outlook, Columbia University GenAI advertising study, MMA/BCG CMO survey — WARC — https://www.warc.com/newsandopinion — 2026-09-15 — B (subscription-gated; underlying academic paper not retrieved)
  42. IAB Raises 2026 U.S. Ad Spending Forecast to 12.3 Percent — ANA (Matthew Schwartz) — https://www.ana.net/magazines/show/id/news-2026-09-14-iab-ad-spend-forecast — 2026-09-14 — B (member-gated beyond summary)
  43. Google Ads data shows query length shift post-AI Mode — Search Engine Land (Jason Tabeling) — https://searchengineland.com/google-ads-data-shows-query-length-shift-post-ai-mode-458162 — 2026-09-11 — C (no disclosed sample, source or methodology)
  44. AI Overviews may be affecting Shopping ad CTR and impressions — Search Engine Land (Anu Adegbola), reporting Smarter Ecommerce and Optmyzr data — https://searchengineland.com/ai-overviews-may-be-affecting-shopping-ad-ctr-and-impressions-487265 — 2026-09-07 — B (authors explicitly disclaim causation)
  45. WPP, Omnicom, IPG and Dentsu cut thousands of jobs; Publicis moves the other way — BestMediaInfo, compiling company filings — https://bestmediainfo.com/mediainfo/advertising/wpp-omnicom-ipg-and-dentsu-cut-thousands-of-jobs-publicis-moves-the-other-way-12482264 — 2026-09-02 — C (aggregator; underlying figures are company filings)
  46. FAQ on the Google Ad Tech Antitrust Decision — EMARKETER — https://www.emarketer.com/content/faq-on-google-ad-tech-antitrust-decision — 2026 — B (paywalled; only the April 2025 liability date was retrievable)

Tier summary: 14 Tier A (company IR, trade-body primary research, standards body), 30 Tier B (named-byline trade press, court reporting, auditor research), 2 Tier C (unsourced agency panels and an aggregator). No claim in this dossier rests on Tier C alone.

Research provenance
Source artifact
02-dossiers/16-advertising-marketing.md
Corpus date
15 September 2026
Prepared for this site
16 September 2026
Site publication
18 September 2026
Verification
Inherited; not fully rechecked

More research in this sector · Explore trend records