The final week of February 2026 landed like a ledger of unpaid debts. Earnings calls exposed cracks across the old guard of holding companies and DSPs. Retail media kept printing money while traditional agency models struggled to articulate what they do. A three-week Google algorithm update quietly completed, leaving publishers no richer in clarity than when it began. Courts shave startedshaping the legal boundaries of what AI can take from the web. And buried inside product changelog announcements, platforms kept expanding their reach into territory that advertisers barely noticed.

WPP's worst year since Covid, a $676 million bet, and a whistleblower in the room

The timing was not subtle. On the same Thursday that WPP unveiled its three-year turnaround plan in London, a parallel story was unspooling in US courtrooms that put internal GroupM trading practices directly on the public record.

On February 26, WPP reported that its 2025 revenue, less pass-through costs, fell 5.4% year-on-year to $13.6 billion, with total reported revenue down 8.1% — the worst annual performance since the pandemic. WPP's shares opened Thursday morning down nearly 10%, extending a decline that began in mid-2025. CEO Cindy Rose, who joined from Microsoft in September 2025, stood before investors and declared the company was "no longer a holding company". The new structure, called Elevate28, collapses the group into four divisions: WPP Media, WPP Production, WPP Enterprise Solutions, and a newly assembled WPP Creative unit that combines VML, Ogilvy, and AKQA under current VML CEO Jon Cook.

The cost savings target is $676 million annually. Rose declined to be specific about how that figure translates to headcount — and given that WPP's workforce had already contracted 8.7% in the 12 months ending December 31, 2025, leaving 98,655 employees, that reticence spoke for itself. When asked what her message to staff was, Rose acknowledged "transformation fatigue" and said the plan was designed to reach completion within 18 months. The "overwhelming sentiment," she insisted, was excitement. She told investors WPP would become "a trusted growth partner for clients in the era of AI" rather than a traditional agency network.

What gave that framing an uncomfortable edge was a simultaneous legal story. On February 27, Digiday published an analysis of court documents from a whistleblower case that WPP had been fighting, noting that in its defence filings, the company had placed its own account of media agency trading practices on the public record. The data in those filings suggested that GroupM's largest clients were, based on internal figures, largely not benefiting from proprietary inventory arrangements that generated close to $1 billion in annual agency income. The lawsuit centres on allegations that GroupM systematically retained client rebates — a pattern that gained additional credibility because Chinese authorities raided GroupM offices and detained employees in October 2023 for precisely that practice.

WPP cannot afford the discovery process if this case goes further. Internal communications about rebate practices could feed both shareholder litigation and client contract reviews simultaneously. For clients currently reviewing agency agreements under the rising wave of principal media buying — where Digiday reported on February 26 that demand is increasing as advertisers want clearer economics — the question of where agency income actually comes from has become impossible to ignore.

Omnicom and Publicis both posted stronger 2025 results. Omnicom came in at $17.5 billion for the year. WPP's structural complexity, which Rose herself cited as the cause of "inconsistent strategic execution," had allowed its competitors to move faster. The four-division model is the most radical structural change WPP has attempted in years. Whether it addresses the underlying economics — or simply reorganises the org chart around the same fundamental questions — will take longer than 18 months to answer.

The Trade Desk: $3 billion in revenue, 16% stock drop, and a fight with its biggest clients

The Trade Desk's Q4 2025 and full-year earnings arrived on February 25, and the headline revenue of nearly $3 billion for 2025 might have landed differently if the Q1 2026 guidance had not disappointed. The company guided Q1 to at least $678 million — representing roughly 10% year-on-year growth — and the stock fell from approximately $25 per share to around $21 in the two hours following the investor call. Several market analysts had called this the most consequential earnings moment in The Trade Desk's decade-long public market history.

Adweek's reporting detailed that CEO Jeff Green attributed Q4 softness partly to reduced spend from automotive and CPG brands, alongside sustained competitive pressure from Amazon. He defended OpenPath — The Trade Desk's initiative to connect directly to premium publisher supply — after Dentsu and WPP had quietly exited the programme over concerns about hidden fees and supply chain transparency, as Adweek had reported exclusively in the prior week. Green's defence was pointed: OpenPath is designed to be competitive, he said, and agencies focused on principal-based buying "are not doing as good a job representing their clients." SSPs, in his framing, are convoluting the ecosystem while trying to exploit it.

The holding company exits from OpenPath matter beyond any single product. They signal that the tension between The Trade Desk's desire to control supply chain architecture and the agencies' desire to maintain their own data and margin leverage has not been resolved by the company's improved media quality arguments. Green announced that every engineer at the company now uses AI tools to write or test code, and outlined plans for an agentic AI framework for partners in 2026 — without sharing specifics. CTV remains central, with live-event inventory from NBCU's Winter Olympics cited as a growth driver.

The broader picture of The Trade Desk's position in the ecosystem was simultaneously complicated by a Digiday Ad Tech Briefing published February 24, which examined how AI's promise of transparency in media buying is actually shining a brighter light on the "ad tech tax" — the cumulative fees extracted at each layer of the programmatic supply chain. PPC Land's own February 22 weekly analysis had framed the OpenPath exits alongside ChatGPT's first confirmed brand ads at $60 CPM and Google Demand Gen targeting shifts as collectively describing an industry undergoing a genuine renegotiation of power — not just product updates.

PubMatic goes all-in on agentic AI, reports 250+ campaigns in two months

Where The Trade Desk faced scepticism, PubMatic tried to reframe declining revenue as evidence of an imminent transformation. On February 26, the SSP reported Q4 2025 revenue of $80 million — down 6% year-on-year — and full-year 2025 revenue of $282.9 million, down 3%. These were below prior-year levels, but PubMatic beat its own projections, and the stock rose approximately 6% in after-hours trading on the strength of the forward narrative.

CFO Steve Pantelick attributed part of the downturn to reduced spend from a single unnamed large incumbent DSP — understood in programmatic circles to be The Trade Desk. The company has diversified its DSP base with new mid-tier partners. But the bigger argument came from CEO Rajeev Goel, who predicted that by 2028, 25% of all digital advertising would be "executed autonomously via agentic AI," rising to 50% by 2030. Since announcing its first agent-executed CTV campaign in January 2026 in partnership with agency Butler/Till, PubMatic has run more than 250 agentic ad campaigns. Many involve new advertisers, proving — according to Goel — that agentic solutions bring incremental revenue rather than just shifting existing spend.

AgenticOS, PubMatic's AI layer, generates fees beyond the standard SSP take rate. CTV revenue was up 50% in Q4 year-on-year (excluding 2024 political spend). Mobile app revenue grew 25%. Emerging revenues — covering Activate, commerce media, and new AI solutions — grew 75% in Q4 and now represent nearly 10% of total revenue. Already 40% of new code written at PubMatic in the second half of 2025 was AI-generated. Goel noted that 10% of publishers on the platform are now generating revenue from AI solutions and said that percentage "should be 100% eventually." The company is also participating in the creation of an "Ad Context Protocol" — a standards effort aimed at making programmatic infrastructure interoperable with AI agent workflows.

The contrast between PubMatic and The Trade Desk in the same week was instructive. One is a buy-side platform facing questions about growth trajectory and client relationships; the other is a sell-side platform whose numbers are weak but whose strategic pivot toward agentic AI is attracting investor patience. Whether PubMatic's 250 campaigns are a genuine inflection point or a rounding error in a $600+ billion global advertising market is the question the next 24 months will answer.

Walmart reaches $6.4B in ad revenue as Sparky drives 35% larger baskets

The most striking number of the week came not from ad tech but from Bentonville. On February 19, Walmart reported that its global advertising business grew 37% in fiscal year 2026 to nearly $6.4 billion, with Walmart Connect — the domestic US business — up 41% in Q4 alone. CFO John David Rainey told investors that advertising and membership combined now account for a full third of the company's most recent quarterly profit.

AdExchanger contextualised the figure: VIZIO, the TV hardware company acquired in 2024, posted triple-digit percentage advertising revenue growth during the year. Third-party marketplace sellers — who operate similarly to Amazon's network — are growing faster as an advertising channel than first-party Walmart brand budgets. Rainey acknowledged that the ad business is reaching cumulative totals that make sustaining the same percentage growth rate harder, but said "I really don't see that slowing down."

The agentic angle is already visible in basket data. Walmart US CEO Dave Guggina disclosed that online shoppers who interacted with Sparky, the company's AI shopping assistant, generated on average a 35% larger order value compared to those who did not. About half of Walmart's app users engaged with Sparky during Q4. When asked how advertising would evolve inside agentic commerce, Guggina offered: "I think we're all learning as we go, and we'll figure that out." For comparison, Amazon's 2025 ad revenue was $68 billion on $830 billion in gross merchandise value — roughly 8% penetration. Walmart's $6.4 billion on equivalent sales of $713 billion represents just over 1%. The arithmetic of that gap, and whether Walmart's trajectory can compress it, is the structural question for retail media in the next five years.

Google's Discover update completes after three weeks — but left publishers guessing

The February 2026 Google Discover core update officially completed on February 27, having started rolling out on February 5. Google had initially indicated the update would take around two weeks; it ran for just over three. This is Google's first core update of 2026, and it applies specifically to the Discover feed — not to general web search rankings.

The stated goals were threefold: show more locally relevant content from publishers based in the user's country, reduce sensational and clickbait content, and surface more in-depth, original, and timely material from sites with demonstrated expertise in a given area. The documentation update also added "provide a great page experience" as an explicit Discover signal — bringing attention to ad-heavy page layouts, auto-playing media, and intrusive interstitials as negative factors.

The reality on the ground was messier. Search Engine Roundtable documented publishers reporting traffic collapses from Discover starting as early as February 2 — several days before the official update announcement on February 5 — with some reporting near-total elimination of Discover traffic. Comments published by Search Engine Roundtable included accounts from sites that had previously appeared in Google News and Discover and found both sources cut simultaneously. The localisation component — deprioritising non-US publishers in US feeds — hit international publishers disproportionately hard, and Google's guidance on how to recover remained sparse.

That Discover update landed simultaneously with a six-day Google AdSense bug, active from February 13, that made anchor and vignette ads unclosable on iOS devices. Users could not dismiss the ad units, which inflated click-through rate data in publisher dashboards while simultaneously damaging the user experience that the Discover update was designed to reward. Publishers trying to interpret why Discover traffic shifted during this period were reading potentially corrupted data.

On February 24, Search Engine Roundtable reported that Google is also separately testing the removal of dates from articles shown in the Discover feed — a limited experiment to measure the impact of deprioritising recency signals. If article dates disappear from Discover, publishers who rely on news-cycle freshness to attract that audience would face a fundamentally different optimisation challenge.

A Similarweb analysis published in the week of February 23, covered by Search Engine Roundtable, compared SERP composition and click share data across several verticals between January 2025 and January 2026. The finding: as AI Overviews expand their presence at the top of Google search results, click share is shifting toward Shopping search ads. The organic result beneath an AI Overview is capturing less attention; the sponsored placement within or alongside the AI layer is capturing more. In short, Google's AI search experience is producing a structural conversion of organic intent into paid ad spend — without requiring a formal announcement.

This connects directly to the Google Ads support form story that PPC Land surfaced on February 22. The contact form for Google Ads support includes a checkbox and disclaimer stating that advertisers authorise "a Google Ads specialist on behalf of your company to make the changes above directly to your company's Google Ads account." The clause is buried in the support flow, not prominently disclosed. Search Engine Roundtable confirmed the language on February 24. The combination — AI search eroding organic while paid grows, and a support form that grants account access by default — adds texture to the dynamic that publishers and advertisers are navigating when dealing with Google.

Meanwhile, call-only ads officially lost the ability to be created in February 2026, following through on a deprecation timeline announced in late 2025. All options to create new call-only ads have been removed. Existing call-only ads will continue receiving impressions until February 2027, when they cease entirely. Google is routing call advertisers toward call assets within responsive search ads, which use Google's AI to test headline and description combinations automatically.

On February 20, SerpApi filed a 31-page motion to dismiss Google's DMCA lawsuit in the United States District Court for the Northern District of California. The underlying lawsuit — Case No. 25-cv-10826-YGR — was filed by Google on December 19, 2025, before Judge Yvonne Gonzalez Rogers, alleging that the Austin-based, 42-person search API company circumvented Google's SearchGuard anti-bot system to scrape search results, seeking statutory damages of $200 to $2,500 per act of circumvention.

SerpApi's motion argues that Google lacks statutory standing as a non-copyright holder, that SearchGuard does not qualify as a "copyright access control measure" under the technical definition in the DMCA's Section 1201, and that SerpApi has not circumvented any protected technological measure because what it accesses is publicly visible information — the same results shown to any standard browser. SerpApi's general counsel Chad Anson had published a public statement in January 2026 arguing the DMCA was written to prevent DVD piracy and software hacking, not to allow platforms to restrict access to publicly viewable web content. A hearing on the motion to dismiss is scheduled for May 19, 2026.

The case matters far beyond SerpApi's own business. If Google succeeds in framing SearchGuard as a copyright access control measure, it would effectively make scraping search results legally equivalent to cracking DVD encryption — regardless of whether the underlying content is publicly displayed to any human visitor. SEO platforms, ad monitoring services, competitive intelligence tools, and AI training pipelines all depend on the principle that publicly visible information remains publicly accessible. PPC Land had previously documented the broader irony: Google itself scrapes billions of web pages for AI training without compensating publishers, yet is invoking copyright law to protect its own search results from the same kind of extraction.

The February 23 Search Engine Roundtable recap noted the SerpApi motion filing among the day's top items, alongside lingering Google search ranking volatility that some tracking tools were showing signs of beginning to cool — though cautiously and unevenly across verticals.

Microsoft fights prompt injection while testing shopping carousels and video grids

While Google's legal and algorithmic news dominated the week, Microsoft Advertising made several distinct product moves. On February 24, Search Engine Roundtable reported that Microsoft has implemented mitigations against prompt injection attacks in Copilot. Spammers had been using "Summarize with AI" buttons embedded in websites to insert instructions into AI responses — effectively attempting to manipulate what Copilot tells users. Microsoft publicly labelled these "prompt injection attacks" in a published article, which is itself notable: naming the attack vector in official communications signals both awareness and a degree of counter-pressure on site operators who might unknowingly host such content.

In parallel, Microsoft Advertising tested a shopping carousel within Bing search ads that can show multiple images inside a single card — a carousel within a carousel. This format is designed to present product variations or complementary items without requiring an additional click to a product detail page. On February 26, Microsoft also launched self-service negative keyword lists, allowing advertisers to manage exclusions at the account level rather than campaign by campaign. And by February 27, Bing was testing a two-by-two video grid layout in standard search results — replacing the traditional video list view with a denser visual format that mirrors how social platforms surface video content.

Search Engine Roundtable's February 25 recap also noted that Anthropic updated its crawler documentation, clarifying what ClaudeBot, Claude-User, and Claude SearchBot each do and what blocking them results in. The update is a small but significant signal: as AI companies expand their crawling infrastructure, the documentation layer that tells publishers how to manage these bots is becoming a live, updated resource rather than static boilerplate.

Google, meanwhile, updated Circle to Search on February 26 to analyse the entire image rather than just the portion the user circled, enabling multi-object search from a single photo. Google also published EU Political Ads Regulation changes for the Google Ads API, with the Google Ads API version 23.1 — released February 26 — covering updates to account management, campaigns, conversions, incentives, planning, and reporting. The API now follows a monthly release cycle, a shift from the prior quarterly cadence that Google had signalled in late 2025.

Netflix CTV share doubles by 2027 forecast; Magnite says display is "old and busted"

On February 24, Digiday reported that WARC forecasts Netflix's share of the global CTV advertising market to rise from 3.7% at the end of 2025 to 9.2% by 2027. Four media agency executives attributed the shift to pricing stabilisation, expanding live sports inventory, growing advertiser familiarity with the platform, and programmatic access through the 2025 deal with Amazon's DSP. Digiday's February 25 Future of TV Briefing noted that Netflix's in-house ad platform launch had led some advertisers to double their spending on the streamer, indicating that the consolidation of buying through a single interface is reducing friction in a way that external third-party access did not fully achieve.

US CTV advertising is expected to grow 13.8% in 2026. AdExchanger's February 25 coverage of Magnite's earnings, written by Victoria McNally, summarised the company's results with the observation that display advertising is "old and busted" and streaming CTV is the new hotness — an editorial characterisation of what the numbers showed, not Barrett's own words. Magnite's Q4 results gave that framing its weight: CTV was the clear growth engine while display lagged. The direction of travel was not subtle. Publishers and agencies building 2026 media plans are responding to that dynamic — Nielsen's addition of more than 200 advanced audience segments to Nielsen ONE on February 26, timed ahead of upfront season, is part of the infrastructure being assembled to make cross-screen planning operational rather than aspirational.

Viaplay also announced during this period that it secured all broadcast rights to the EFL Championship across every Nordic market in a multi-year deal running until 2028, covering all matches live. The rights deal represents continued investment in live football as premium advertising inventory, at a time when live sports remains the category most resistant to attention fragmentation across streaming services.

Amazon rewrites the rules for sellers and AI agents — three changes before March 4

Amazon introduced three distinct changes to its marketplace framework in the week of February 22, each of which addresses a different part of the seller and advertiser lifecycle.

Effective March 4, Amazon's updated Business Solutions Agreement introduces a formal Agent Policy. AI agents acting on behalf of sellers must explicitly identify themselves and comply with existing seller obligations. This is the first time Amazon has formally acknowledged and regulated the category of autonomous software operating as seller proxies. The practical implications extend to any advertiser or technology vendor using AI-powered tools to manage bids, listings, or inventory on Amazon without direct human authorisation for each action.

A second change, effective February 28, narrows On-Time Delivery Rate enforcement so that only the worst-performing individual seller-fulfilled listings face deactivation — replacing a blanket account-level penalty. For multi-SKU merchants who previously faced account suspension over a small number of underperforming listings, this is a significant narrowing of enforcement scope. The change reflects Amazon's ongoing adjustment of its enforcement posture to be more proportionate and listing-specific rather than account-wide.

Third, Amazon's Brand Stores management API reached general availability in February 2026, exiting beta and giving brands programmatic control over storefront content at scale. Brands managing large catalogues can now automate updates across store pages and modules — a development that matters as the number of SKUs on Amazon continues to expand and manual storefront management becomes operationally unsustainable. Amazon's A+ Content Manager also introduced two simultaneous changes: the Shoppable Collections module ended on February 27, with Brand Story replacing it, and Content Quality Analysis launched in beta — a scoring tool that evaluates content quality across A+ modules. The combination of these changes reflects Amazon's push to automate and standardise content quality signals in ways that reduce reliance on advertiser self-reporting.

Reddit's dual reality: shopping carousels in search, 20% of brands' social budgets

Reddit is operating as two distinct advertising propositions simultaneously. On the product side, PPC Land reported on February 22 that Reddit is testing interactive product carousels inside its search results for a small group of US users, pulling from Dynamic Product Ads partner catalogues and weaving in community discussion context. The technical integration — using DPA partner data to surface buy links alongside forum threads — attempts to capture the intent signal that Reddit's search represents while keeping the shopping experience native to the platform.

On the advertiser behaviour side, Digiday reported on February 27 that since Reddit's 2024 IPO, some brands have allocated up to 20% of their paid social budgets to the platform. Shatesha Scales, supervisor of paid social at media agency Rain the Growth, described 2025 as the year brands "unlocked" Reddit budgets, with 2026 being the ramp-up. The dominant use case remains upper-funnel: niche reach and awareness rather than direct conversion. Taji Zaminasli of Ars X Machina said from a conversions perspective, Reddit does not match what Meta and TikTok deliver. The platform's value proposition — brand-safe communities of genuine interest — makes it a top-of-funnel asset, and brands treating it as a performance channel are likely measuring it against the wrong metric.

Publishers want stability, not disruption — and vertical video as an escape hatch

AdExchanger's February 24 analysis, drawing on conversations with nearly 20 sell-side executives, found a striking degree of uniformity in what publishers want from 2026: nothing new. After cookie deprecation in 2024 and AI search disruption in 2025, the dominant mood is not ambition — it is fatigue. No one in those conversations talked about CPM recovery or breakout programmatic growth. They talked about traffic volatility, automation replacing headcount, curation becoming mandatory rather than optional, retail media as a growth lever, and AI that shows up in actual products rather than slide decks.

Against that backdrop, JWX launched Vertical Video on February 19 — a single-line JavaScript implementation that lets any publisher create a swipeable, TikTok-style video feed directly on an owned property. The product is a direct response to the structural dynamics eroding publisher traffic: audiences are spending more time in social video feeds; AI search is compressing pageview counts; and programmatic display CPMs have not meaningfully recovered from the disruptions of the past 18 months. JWX's argument is that publishers do not need to give up audience to social platforms in order to serve the format those audiences expect.

The broader vertical video push among publishers is documented extensively. Time, CNN, The New York Times, People, TMB, Hearst, and Recurrent Ventures are all adding swipeable video to owned properties. IAB data shows creator economy ad spend more than doubling since 2021 to $29.5 billion in 2024. Publishers building vertical video feeds on owned domains are attempting to claim a portion of that spend on their own terms — a format bet that requires both editorial investment and advertiser education before the economics fully resolve.

Data signals: marketer burnout, mobile ATT opt-ins, and CRM-powered ROAS

Three data stories from the week of February 22 add quantitative texture to the broader trends.

Adobe's study of 1,106 US marketing workers found that 84% work overtime, with the average marketer losing 91 business days per year to low-impact tasks. 32% cited poor work management as a reason they had left a previous job. These figures do not merely describe workload conditions — they explain why AI-assisted workflow automation has adoption momentum that goes beyond cost-cutting. When 91 days of productive capacity per person is consumed by low-value tasks, the operational case for automation does not require a strategic vision statement.

Adjust's Mobile App Trends 2026 report showed finance app sessions up 21%, gaming cost-per-install rising 30%, and Apple's ATT opt-in rate reaching 38% globally. The ATT figure carries the most structural weight. When Apple introduced its App Tracking Transparency framework in 2021, opt-in rates were in the low teens. The climb to 38% suggests that users are becoming more willing to consent — either because prompts are being framed more effectively, because the category of apps requesting consent (finance, health, lifestyle) is one where users see direct value exchange, or because the cultural moment of peak privacy anxiety is beginning to moderate. For mobile attribution models built on probabilistic matching, higher opt-in rates shift the measurement problem from inference toward more direct signals.

LiveRamp and Scowtt announced on February 22 a partnership integrating predictive AI models that use first-party CRM data as live campaign optimisation signals across major ad platforms and programmatic destinations. The claimed ROAS improvement is 40%+. The technical distinction that matters is that CRM data here is not being used as a batch audience upload — it is being used as a continuous signal to update bid and targeting decisions in near-real-time. The architecture difference between uploading a customer list and feeding a live predictive model to an ad platform's bidder is significant; the former operates on historical segments, the latter on current propensity scores that update as customer behaviour changes.

Germany's BVDW calls for education over bans on child safety online

Germany's Bundesverband Digitale Wirtschaft (BVDW) called on February 22 for media literacy programmes and risk-based age verification as its preferred framework for protecting minors online, backed by a Civey survey of 2,500 people conducted February 3–4, 2026. The position directly counters the approach taken by Australia — which in 2025 banned users under 16 from major social platforms — and represents a distinct European regulatory philosophy: address capability rather than restrict access.

The BVDW's stance is relevant to advertisers because the regulatory landscape around youth-targeted advertising is shifting across multiple jurisdictions simultaneously. AdExchanger's January 2026 privacy roundup had highlighted COPPA 2.0 and state-level youth privacy laws in the US as the regulatory development most likely to reshape advertiser targeting practices in 2026. Germany's debate about the right policy instrument — education versus hard age gates — reflects the same tension playing out in different regulatory frameworks. Advertisers who have built targeting structures on broad demographic assumptions about under-17 audiences face different exposure depending on which jurisdiction's model prevails.

Timeline

Share this article
The link has been copied!