The week that opened with a cluster of platform infrastructure changes closed with a ruling from the EU General Court that redraws, at least partly, the boundaries of the Digital Markets Act. Three stories from June 3 together describe a shift in how digital advertising's regulatory and measurement scaffolding is being reconstructed in Europe and beyond.

EU court strikes down Meta's Marketplace gatekeeper status

On June 3, 2026, the EU General Court delivered a split judgment in Case T-1078/23, partially annulling the European Commission's September 2023 decision that designated Meta as a Digital Markets Act gatekeeper. The court struck down the designation of Facebook Marketplace as a core platform service while upholding the classification of Facebook Messenger. The citation is ECLI:EU:T:2026:357, delivered in Luxembourg by the Eighth Chamber sitting with five judges.

The ruling is the first time a full merits examination has tested exactly where the boundaries of gatekeeper classification fall under Regulation (EU) 2022/1925. For the advertising industry, those boundaries matter concretely: gatekeeper status determines what interoperability obligations apply, what data-sharing requirements the Commission can enforce, and how closely rivals can demand access to a dominant platform's commercial ecosystem.

On Messenger, the court rejected Meta's core argument that the messaging service is simply a feature of Facebook rather than a distinct platform. Three facts drove the conclusion. First, Meta distributes Messenger through standalone mobile applications for iOS and Android. Second, users can access Messenger regardless of whether their Facebook account is active or whether they are using the social network at the same time. Third, Meta promotes tools specifically for businesses to engage users on Messenger - lead generation, customer service, and re-engagement - functions the court described as treating Messenger as a self-standing platform. Meta's data showing that 100% of Messenger communications take place between users with Facebook accounts was dismissed as a consequence of Meta's own design choice to require a Facebook ID, not evidence of integration.

The Marketplace outcome was different. Meta had changed Marketplace on July 31, 2023 - six weeks before the Commission's designation decision - limiting users to a maximum of 20 listings per month per category and imposing tighter caps on vehicle, auto parts, and property listings. Meta notified the Commission of these changes on July 3 and August 3, 2023, before the decision was adopted on September 5. The court found the Commission had failed to adequately account for those changes in its assessment of whether Marketplace met the gatekeeper thresholds. The designation for that service was annulled on grounds of insufficient reasoning.

The ruling creates a practical asymmetry. Messenger remains subject to DMA interoperability requirements. Marketplace does not. The Commission can re-examine Marketplace under an updated analysis, but the court's finding that earlier reasoning was inadequate raises the bar for any future designation attempt.

Google adds dedicated AI visibility to Search Console

The same day, June 3, Google launched dedicated Search Generative AI performance reports inside Google Search Console. The announcement, published on the Google Search Central Blog by Hillel Maoz and Moshe Samet, creates a standalone view of how websites appear within AI-powered features - specifically AI Overviews and AI Mode within Search, and generative AI features within Discover.

Five data categories are exposed: impressions, pages, countries, devices (Search only), and dates. The date granularity runs to hourly, daily, weekly, and monthly breakdowns. Publishers can now see which specific URLs surface inside AI features, broken down by geography - a capability that has not existed anywhere in the platform since AI Overviews launched in the US in May 2024.

The rollout is phased. Google is making the reports available to a subset of sites first, with no timeline given for broader access. The data is not a separate silo: AI feature impressions are already counted inside the main performance report's Web Search totals, and the new reports provide a segmented view of that existing data rather than a fresh measurement source. That distinction matters because it means publishers comparing their total impressions before and after this launch will not see any change - only a new way to break out what was already there.

The launch is a response to a measurement problem that has been accumulating since 2024. When Google confirmed in June 2025 that AI Mode clicks and impressions count toward Search Console totals, it folded them into existing Web Search figures without separation. Research published on PPC Land tracked 3,119 informational queries across 25.1 million organic impressions and found a 61% reduction in organic click-through rates where AI Overviews appeared. Without a dedicated report, isolating the effect was impossible. That isolation is now, at least for sites in the initial rollout, possible.

Schibsted's pay-or-okay model draws a GDPR complaint in Norway

On June 3, noyb and Norway's consumer advocacy organisation Forbrukerradet filed a joint GDPR complaint against Schibsted with Datatilsynet, the Norwegian data protection authority. The complaint targets Schibsted's pay-or-okay consent model across its Nordic publications, which include Aftenposten, VG, Bergens Tidende, Aftonbladet, and TV4.

The mechanics are not unusual: readers who visit Schibsted publications are offered a binary choice - consent to personalised advertising tracking, or pay a monthly fee to access content without being tracked. No third path exists. noyb argues that structure violates Article 7 of the GDPR, which requires consent to be freely given. Personalised advertising tracking is not necessary to deliver a news article, the complaint states; it is a commercial arrangement that benefits the publisher. Tying access to it - or charging for its absence - converts consent into a financial transaction and therefore coercion.

The consent rate figures attached to the complaint illustrate why the legal question matters commercially. noyb states that pay-or-okay systems consistently achieve approximately 99% consent rates. Research on genuine preferences shows that only 0.16% to 7% of people actually want to be tracked for personalised advertising. A Schibsted Sweden executive, Fredric Karen, was quoted in an SVT interview stating that "studies show that if you allow this [to reject consent] without demanding any form of payment in return, a great many users will decline." Max Schrems, noyb's honorary chairman, described that statement as an admission that the model is designed to circumvent genuine choice.

The pay-or-okay model is not unique to Schibsted. It has been adopted across European publishing as a response to tightening consent requirements under GDPR's TCF framework. The Schibsted complaint adds a Nordic regulator to the list of enforcement targets and, if Datatilsynet acts on it, could establish case law applicable to publishers across the region. The EU General Court's Meta ruling earlier the same day reinforced a broader theme: European courts and regulators are still actively contesting the conditions under which digital platforms collect and monetise user data.

AI agents and fraud are converging to the same behaviour

A report published by HUMAN Security on April 9, 2026 - and covered by PPC Land on June 2 - documents a structural problem that has become impossible to ignore. Agentic AI traffic grew 7,851% across 2025. Automated traffic as a whole grew 23.51% year over year, against 3.10% growth in human traffic. The practical consequence is that just half a percentage point now separates the rate of benign AI automation from the rate of malicious fraud across the interactions HUMAN processed through its Human Defense Platform.

An AI shopping agent browsing product pages, adding items to a cart, and completing a checkout behaves mechanically like a fraud bot executing a carding attack or an account takeover sequence. The intent differs. The signal does not. Retail and media verticals drove more than 80% of the AI traffic increase. OpenAI bots - ChatGPT User, OAI-SearchBot, GPTBot, and ChatGPT Agent - accounted for approximately 69% of all observed AI-driven traffic by volume. Meta-ExternalAgent contributed 16%. Anthropic identities, including ClaudeBot and Claude-SearchBot, added roughly 11%. The remaining operators collectively represented under 5%.

The implication for fraud detection is direct. Systems calibrated to flag non-human behavior now face a signal environment where non-human behavior is often legitimate, and where blocking it carries real commercial cost. For attribution and conversion measurement, AI agent transactions introduce a category of activity that current MMP and analytics infrastructure was not designed to classify, count, or value correctly.

Also noted

  • June 3, 2026: Google added dedicated Search Console reports for generative AI impressions across both AI Overviews and AI Mode, rolling out to a subset of sites, with page-level, country, device, and date breakdowns now separable from aggregate Web Search data for the first time. PPC Land
  • June 2, 2026: SiriusXM and Tubi announced a revenue-sharing deal that brings video podcast content from Conan O'Brien, Trevor Noah, and other shows to Tubi's 100 million monthly free-streaming viewers, expanding ad inventory across a combined audio-video audience. PPC Land
  • June 3, 2026: Trump signed an executive order reinstating a voluntary 30-day frontier AI model safety review process - a mechanism his administration had abolished 17 months earlier - amid reports that Anthropic's Mythos model had alarmed Washington and prompted EU access talks. PPC Land
  • June 2, 2026: Spotify launched a Verified by Spotify badge for podcast shows and reaffirmed its ban on AI voice-cloning impersonation, addressing creator identity integrity as AI-generated audio proliferates across podcast platforms. PPC Land
  • June 3, 2026: The AdExchanger June 3 roundup highlighted that Peer39's acquisition of Adloox from Scope3 positions it to challenge DoubleVerify and IAS inside walled gardens, while Scope3 divests to focus on agentic media buying through its Interchange platform and the Ad Context Protocol. AdExchanger

What you should watch closely this week

The stories that generated the most industry conversation this week were the regulatory ones — the DMA ruling, the GDPR complaint, the AI executive order. Those matter. But the changes most likely to affect campaign performance in the next 30 days are quieter, and two of them arrived within 48 hours of each other on the measurement and signal side of the stack.

First: the AdSense full IP control. On June 1, Google launched a publisher-side toggle in AdSense that restores the full IPv4 address — including the fourth octet — in programmatic bid requests to Authorized Buyers, DV360, and Google Ads. It is off by default. Most publishers running AdSense monetization will not have seen a notification. The practical effect of enabling it is concrete: buyers recover city and metro-level geolocation resolution, invalid-traffic detection tools regain host-level granularity for cross-referencing against known data-center and proxy ranges, and IP-to-company firmographic matching — the foundation of B2B account-based targeting — becomes usable again at the impression level. Publishers with B2B audiences, local advertisers, or programmatic floors that depend on accurate location signals have a direct commercial reason to evaluate this setting now. The European caveat applies: GDPR and ePrivacy regulatory environments classify full IP addresses as personal data, and the toggle does not activate for non-personalized ads or TCF-restricted requests regardless of account settings. But for publishers outside or partly outside those constraints, leaving the default in place is a silent concession of signal quality to buyers who already have full-IP access through other supply paths.

Second: the Google Search Console AI visibility reports. The new dedicated generative AI performance reports launched June 3 in a phased rollout. For any site that has been trying to understand whether AI Overviews or AI Mode are affecting traffic — and for most content-dependent businesses, that question has been live since mid-2024 — this is the first time the data has been separable from aggregate Web Search metrics inside the platform itself. The rollout is to a subset of sites only, with no timeline given for the rest. Checking Search Console now to see whether the Generative AI features report section has appeared in the interface is worth doing this week, before the next cycle of performance reviews. The data on impressions by page, country, and device for AI surfaces is already being counted in the platform's aggregate totals — the reports make it visible rather than adding new measurement. Missing that window means missing a baseline from which to measure subsequent changes, including any effects of the May 2026 core update that completed on June 2.