The week ending March 14, 2026, brought a cluster of developments that, taken together, describe an industry at an inflection point. Agentic AI infrastructure is moving from conference slide to live product. An advertising standards body expelled a platform giant. The Winter Olympics became a programmatic stress test. And beneath all of it, Google's search ranking volatility continued burning without a public explanation.

IAB Sweden expels Meta over deceptive ads

The sharpest headline came from Stockholm. On March 11, 2026, IAB Sverige — Sweden's national advertising trade body — voted to expel Meta from its membership, concluding that the company's efforts against fraudulent advertising are insufficient. The expulsion was formalized at an emergency board meeting on March 11, after a first vote taken on March 10 was declared invalid due to a procedural error. The sequence matters: the board had initially moved to retain Meta with conditions, then discovered it lacked the legal authority to attach those conditions, and reconvened to take a clean, unconditional vote.

The result: full expulsion. Daniel Weilar, chairman of IAB Sweden, stated that "Meta's work against deceptive ads is not enough," and that Meta would need to produce evidence of meaningful improvement before it could re-enter as a member. A formal appeal mechanism exists - Meta can challenge the decision at IAB Sweden's annual general meeting on April 15, 2026. Until that date, or unless it presents evidence of progress, the expulsion stands.

The context that preceded the decision is not new to the industry. Internal company documents reviewed by Reuters in November 2025 revealed that Meta internally projected earning approximately 10% of its 2024 annual revenue — roughly $16 billion — from advertisements promoting scams and banned goods. Those same documents indicated that Meta's platforms expose users to an estimated 15 billion higher-risk scam advertisements every single day. The Swedish advertising market, while not outsized globally, is not negligible either: programmatic advertising in Sweden reached 4.7 billion kronor in 2024, a 12% year-on-year increase, making the country an active and growing programmatic market.

The reaction on LinkedIn was immediate. Justin Lebbon, an industry consultant who shared the IAB Sverige announcement on March 12, attributed the Swedish decision partly to pressure from Swedish media organisations, and framed it as a model for other markets: the post drew 50 reactions and named IAB UK and IAB Thinkbox directly. Whether those organisations move is unclear, but the Swedish precedent is the first formal expulsion of a major platform by a national advertising trade body over advertiser safety.

MediaPost's coverage on March 13 noted the Swedish board action among its top stories, alongside Madison and Wall's updated U.S. ad-growth forecast of 10.2% for 2026. Both items — one about platform accountability, one about ad spend momentum — arrived on the same day, which is something of a compression of the industry's contradictions: platforms face governance challenges even as total digital ad spend keeps climbing.

MCP moves from protocol to plumbing

The most technically significant thread of the week was the industry's accelerating adoption of Model Context Protocol (MCP), the open-source standard developed by Anthropic that defines how AI agents connect to external data sources and tools. AdExchanger published an AdExplainer on MCP on March 9, describing it as a "universal adapter" for AI in advertising. The analogy that appeared in the piece: if MCP is a USB standard, the connectivity layer that plugs systems together is the row of ports on the laptop.

Several live implementations emerged this week. Adweek reported on March 11 that The Trade Desk is testing AI campaign creation using Anthropic's Claude, with CEO Jeff Green arguing at a panel that programmatic advertising is structurally suited to AI automation — citing 20 million ad impression opportunities per second, each resolved within 10 milliseconds. The Trade Desk plans to launch an agentic AI framework for partners in 2026, though specifics were withheld. Green's March panel came shortly after the company's Q4 2025 earnings report, where revenue grew 14% but guidance fell below analyst expectations. Green is nonetheless personally betting on the platform's direction: he purchased $148 million of company stock.

Also this week, Adweek reported on March 3 that LiveRamp is now enabling customers to plug in third-party AI agents directly into its platform — a shift from requiring developers to call the LiveRamp API to a direct agent integration model. The change allows AI agents to automate audience planning, segmentation, measurement, and campaign optimization. LiveRamp's customers include major agency groups, large brands such as Uber and Coca-Cola, and publishers including Paramount, Netflix, and Spotify. LiveRamp reported Q4 revenues of $212 million, up 9% year-on-year, though its shares have fallen nearly 7% over the same period.

The underlying infrastructure question — whether ad tech plumbing is actually ready for agentic workflows — is what MadConnect addressed at AdExchanger on March 10. The company made its Intelligent Connectivity Layer (ICL) generally available to enterprise marketers and agencies this week. ICL is built to natively support MCP, serving as the switchboard that connects CRMs, CDPs, clean rooms, DSPs, and SSPs so AI agents can operate across them rather than being trapped inside a single platform. Rio Longacre, a partner at Credera (Omnicom's consulting division), noted that even moving a single audience from a CRM into paid media still takes weeks at many brands. The gap between what agentic AI promises and what infrastructure can actually deliver remains, he suggested, the central constraint.

Dstillery, the audience platform, also launched DS-1 this week — an MCP-based agentic interface allowing clients to develop custom audience segments in natural language and push them directly to The Trade Desk for activation. Keynes, a CTV targeting platform, is listed as an early adopter. Keynes CEO Dan Larkman described his company as "90% tech and 10% humans," indicating how far the automation model has already shifted in specialist firms.

NBCU's Winter Olympics: seven DSPs, faster data, a stress test

While the MCP discussion remained largely speculative in nature, NBCU's handling of the Milan Cortina 2026 Winter Olympics provided a live-scale test of programmatic infrastructure. AdExchanger reported on March 10 that NBCU worked with seven buy-side platforms during the Games: The Trade Desk, Google's DV360, FreeWheel's Buyer Cloud (formerly Beeswax), Amazon DSP, Yahoo DSP, StackAdapt, and Viant. By comparison, Paris 2024 had only offered programmatic access through The Trade Desk. The expansion meant broader competition for programmatic inventory and a wider advertiser base.

Ryan McConville, NBCU's chief product officer and EVP of ad products and solutions, told AdExchanger that the company "processed the largest volume of data in Olympics history" and did so "two times faster than we did in Paris." Reporting turnaround — previously requiring "multiple hours or even days" — was compressed to within one to two hours during the Winter Games, enabling near-real-time optimization. NBCU also made Olympics inventory available through Comcast's Universal Ads self-serve platform for the first time, aimed at attracting smaller, digital-native advertisers.

The advertiser headroom was measurable: 200 net-new brands that had never previously advertised during the Olympics bought programmatically during these Games. That figure is notable. It suggests the self-serve and programmatic channel is opening live sports inventory to advertisers who previously lacked the direct relationships or minimum budget requirements to participate. Category exclusivity, however, created specific technical challenges — brands that are title sponsors may demand separation from any creative that features competitive products or even tangentially related imagery. McConville described the need to build "guardrails to look at bids and make sure [ads] actually qualify," combining automated and human review.

ChatGPT ads: Criteo plugs in, retail media watches

The broader structural story of AI platforms entering the advertising market continued to develop. Criteo (NASDAQ: CRTO) announced on March 2 that it is the first advertising technology company to integrate with OpenAI's ChatGPT advertising pilot, connecting approximately 17,000 advertisers to ChatGPT's Free and Go subscription tiers in the United States. The announcement, also covered by Adweek, positions Criteo as a supply-side intermediary between OpenAI's nascent ad inventory and the programmatic ecosystem.

OpenAI requires brands to pay a minimum of $200,000 to test placements. Ads first appeared in ChatGPT on February 9, 2026, and the Criteo announcement came less than a month later. One figure in Criteo's announcement stands out: users referred to retail websites from LLM platforms converted at approximately 1.5 times the rate of other referral channels, based on observations from 500 U.S. retailers in February 2026. Sample size and duration are limited, and the claim should be read as directional rather than conclusive. But it is the first data point the industry has seen on LLM referral quality.

The longer-term structural question around what these AI advertising entrants mean for retail media specifically was addressed by Digiday on March 9. Traditional search engine query volume is expected to drop 25% in 2026 as users shift to AI chatbots, per Gartner projections. Retail media's U.S. ad business stands at $69.33 billion. Advertisers are not yet moving dollars out of retail media into AI chatbot placements — Ross Walker, Acadia's director of retail media, stated that of all the brands he works with, none is taking money away from retail media to put into AI ads — but the threat is framed in terms of platform substitution, not immediate budget reallocation.

Google Performance Max: voice-over opt-out window closes March 20

Inside the platforms, a time-sensitive product change from Google is affecting Performance Max advertisers. PPC Land reported on March 10 that Google began notifying advertisers that Performance Max campaigns will automatically receive AI-generated voice-over on existing video ads starting March 20, 2026 — unless advertisers actively disable video enhancement controls in their accounts. The notification was surfaced publicly via a LinkedIn post by Georgi Zayakov, a PPC practitioner, who shared an email originally circulated by Jordan Fry.

The mechanism: Google's system selects headlines and descriptions from a campaign's existing asset groups, generates spoken audio using Google AI voice models, and layers that audio on top of the base video, saving the result as a new video asset. The feature applies only when the original video does not already contain a voice track. The opt-out is a campaign-level toggle, not an ad-level control, meaning advertisers with large Performance Max portfolios need to audit each campaign individually before the March 20 deadline. Advertisers who had already disabled video enhancement control in earlier rollouts are not affected.

The pattern this follows is visible across Google's product roadmap. January 2026 saw Google expand asset-level A/B testing to all Performance Max campaigns; the voice-over announcement moves in a different direction, introducing a new asset variant without a prior testing decision from the advertiser. Google Ads also sent notices to some advertisers this week asking them to confirm by March 31, 2026, whether their campaigns contain European Union political content, a data hygiene exercise required under EU political advertising rules.

Separately, Google Customer Match data upload procedures are changing after April 1, 2026: uploads will no longer function through the Google Ads API and must be migrated to the Data Manager API.

Google search ranking volatility: no end in sight

Google's search ranking volatility continued into the week of March 7–14, 2026, a pattern that has persisted without formal acknowledgment from Google throughout early 2026. SEMrush sensors reached readings of 9.5, which is high. Barry Schwartz at Search Engine Roundtable has been tracking the signal weekly, noting that daily updates are visible but Google has declined to comment on any of them. The February 2026 Google Discover Core Update, which ran from February 5 to February 27 and took an extra week longer than expected, is confirmed complete. The ongoing volatility in standard search results is separate.

The March 2026 Google Webmaster Report documented several additional developments over the month: the Google Search Console AI-powered configuration tool went live; the AI Mode Canvas feature became available to all U.S. users for organizing plans and projects; Google launched new link styles in AI Mode and AI Overviews; and Google Business Profile review policies were updated, which may explain a wave of review disappearances reported by site owners.

One patent surfaced during the month warrants particular attention. A Google patent titled "AI-generated content page tailored to a specific user" describes a mechanism where, rather than directing a search user to the matching page on a company's website, Google instead serves an AI-generated page customized for that specific user. The patent is exploratory, not an announced feature, but it describes a scenario where Google's own AI layer could intercept organic traffic that currently flows to publisher and advertiser pages — a continuation of the Zero-Click trajectory that has concerned the publishing sector since AI Overviews launched.

PPC Land covered on March 8 the publication of Google's "Things to know about Google's web crawling" document, a nine-point overview published on March 3, 2026, covering how Googlebot discovers, renders, and manages site access. The document was flagged by SEO professionals Glenn Gabe and Barry Schwartz on LinkedIn. It does not directly address the reduction in Googlebot's file size limit from 15MB to 2MB per resource — an 86.7% decrease documented in separate crawling infrastructure documentation — but sits within the same documentation family.

W3C pushes DIDs v1.1 to implementation stage

A development with longer-term implications for identity infrastructure arrived on March 5, 2026, when the World Wide Web Consortium published Decentralized Identifiers (DIDs) v1.1 as a Candidate Recommendation. The specification, available at https://www.w3.org/TR/2026/CR-did-1.1-20260305/, formally invites software developers and DID method authors to produce experimental implementations. The public comment period runs until April 5, 2026, via GitHub issues.

DIDs are cryptographic identifiers that don't rely on a centralised registry — unlike cookies, email addresses, or device IDs. The relevance to ad tech is indirect but real: organisations building identity infrastructure for age verification, audience authentication, or cross-platform identity resolution can shape the final normative specification during this comment window. Funding for portions of the specification work came from the U.S. Department of Homeland Security's Science and Technology Directorate and the European Union's StandICT.eu program.

Amazon's Health AI expands to main site; ad stack opens to MCP

Amazon expanded its Health AI agent to Amazon.com and the Amazon Shopping app on March 10, 2026, moving the system beyond the One Medical app where it launched in January. Eligible Prime members receive up to five free direct-message care visits for more than 30 conditions. The agent is built on Amazon Bedrock as a multi-agent architecture. The announcement was authored by Prakash Bulusu, CTO of Amazon Health Services, and Andrew Diamond, Chief Medical Officer of Amazon One Medical. The expansion deepens the integration between Amazon's healthcare services and its retail infrastructure — the same digital surface where advertising runs.

The ad infrastructure side of Amazon is also moving. Adweek's ad tech coverage noted Amazon opening parts of its ad stack to AI agents through MCP, extending the same pattern visible across LiveRamp, The Trade Desk, and Dstillery. Amazon's reach — the Amazon DSP participated in the NBCU Olympics programmatic setup, alongside DV360 and seven other buy-side platforms — makes its infrastructure decisions material to the wider ecosystem.

TV ad revenue: linear slides, streaming slows, CTV still growing

The picture on television advertising spending was filled in this week by Madison and Wall projections published by MediaPost on March 13, 2026. Total TV advertising in 2026 (excluding political) is estimated to decline 2.4%, an acceleration from 2025's 1.4% pullback. National TV advertising (excluding political) is projected to sink 2% to $27.9 billion, down from $28.4 billion in 2025. Core local TV is expected to fall 3% to $13.5 billion. Digital TV (streaming and CTV combined) is estimated up 3% to $21.0 billion — but that figure represents a dramatic slowdown from the 12% growth seen in 2025.

The nuance in those numbers is that streaming growth is decelerating, not reversing. CTV continues to attract budgets migrating from both national linear and local TV — including geo-targeted streaming campaigns that used to be local TV buys. Meanwhile Madison and Wall raised their full 2026 U.S. ad-growth forecast to 10.2%, or 8.9% excluding political, driven in part by the FIFA World Cup and midterm election spending. That macro figure sits alongside declining linear TV in a way that describes the redistribution happening inside the total number.

WPP restructures; The Trade Desk faces OpenPath scrutiny

Holding company news this week centred on WPP's $676 million annual cost-cutting plan, covered in depth by Adweek's agencies section. The plan restructures WPP into four divisions: WPP Media, WPP Production, WPP Enterprise Solutions, and WPP Creative. The company is targeting a return to growth by 2028. The restructuring follows a reported exit by both Dentsu and WPP from The Trade Desk's OpenPath supply chain initiative over transparency concerns — Adweek had the exclusive on those exits in late February. Jeff Green declined to comment directly on those departures when asked at the March 11 panel with Marketecture Media's Ari Paparo.

The week's other agency-adjacent development: NewsGuard launched an AI content farm detection tool on March 13, covered by Adweek, developed in collaboration with AI detection startup Pangram. The system evaluates broad swaths of entire domains rather than individual pages, flagging sites where AI-generated content appears to constitute a significant portion of output. NewsGuard says it has already found 3,000 AI content farm sites — more than double what it detected the previous year using manual methods. The tool is positioned as a brand-safety resource for advertisers placing programmatic buys across the open web.

Timeline

Share this article
The link has been copied!