OpenAI, Yelp, the French advertising industry, and eight years of European data protection enforcement collided on May 30, 2026, in a week that made clear how rapidly the financial stakes of digital advertising are shifting. Measurement infrastructure, consent mechanics, and AI-citation dynamics all moved at once.

OpenAI flips the switch on performance advertising

The clearest signal of the week came from OpenAI. On May 28, Digiday reported that the company had activated cost-per-action (CPA) ads inside ChatGPT's Ads Manager for select advertisers, letting brands pay only when a user clicks through, signs up, or completes a purchase. Three weeks earlier, OpenAI's head of monetization Asad Awan had said that bidding on outcomes was "in the plan" and "should be done soon." The quiet activation made good on that, without a press release.

What arrives June 5 goes a step further. PPC Land reported on May 30 that OpenAI notified beta advertisers via email that conversion-optimized campaigns will begin rolling out that date, with a hard eligibility cutoff: any account that has at least one conversion event flowing through either the JavaScript Pixel or the Conversions API by June 1 will receive early access. The date is not symbolic. A conversion-optimized campaign requires historical signal to learn from. Without that data flowing in before June 5, the system has nothing to optimize against.

The technical architecture is explicit. The Pixel captures a privacy-preserving identifier called oppref from landing page URLs, stores it in a first-party cookie, and batches events before transmission. The server-side Conversions API sends purchase data directly from an advertiser's backend to OpenAI, bypassing browsers and, by extension, ad blockers. When both run simultaneously, deduplication requires a shared event_id on both sides. The design mirrors how Meta's Conversions API has operated for years, though the field names and matching logic are OpenAI's own implementation.

The broader context makes the June 5 date significant. Search Engine Roundtable noted on May 28 that screenshots from Juozas Kaziukenas had already shown a "Conversions" objective appearing in the Ads Manager interface ahead of the official rollout, suggesting the feature was entering a controlled release. The platform opened to all US businesses without a minimum spend on May 5, had added custom audience targeting on May 14, introduced daily budgets and ZIP-code-level geo-targeting on May 22, and now moves to conversion optimization by June 5. That is six significant product additions in 26 days.

The revenue figures behind this acceleration are substantial. PPC Land documented in March 2026 that the ChatGPT ad pilot crossed $100 million in annualized revenue within six weeks of launching. Criteo, one of a small group of technology partners, reported conversion rates on ChatGPT approaching twice those of traditional search in retail categories by May 5. Those figures come from Criteo's own reporting on its client base; independent third-party verification at scale has not been published.

The measurement pivot also surfaces a tension Digiday identified in a separate May 29 briefing: OpenAI has been telling marketers to treat ChatGPT as a test channel rather than a performance one, yet its partner infrastructure, pixel development, and now CPA activation suggest a platform actively building toward performance at scale. The arrival of Skai as a commerce media integration partner, alongside Criteo, widens the channel's retail reach further.

Eight years of GDPR, and 40% of the fines have vanished

Across the Atlantic, a quieter reckoning landed on the data privacy front. PPC Land published a detailed analysis on May 30 of enforcement data compiled by Alliance Risk, a cyber risk advisory firm, covering the entire eight-year history of GDPR enforcement. The numbers are striking: of the roughly EUR7.1 billion in GDPR fines announced since May 2018, nearly 40% - around EUR2.8 billion - has been either annulled by courts or is actively contested before judges.

Two major annulments arrived in March 2026 alone. Luxembourg's Administrative Court struck down the EUR746 million fine imposed on Amazon by the National Commission for Data Protection in 2021, ruling that the regulator had applied strict liability without assessing negligence as a threshold condition, as required by Court of Justice of the European Union case law. A Rome tribunal annulled the Italian Garante's EUR15 million fine against OpenAI on a different basis: once OpenAI established its Irish subsidiary, the Irish Data Protection Commission became the lead supervisory authority, rendering the Italian regulator without jurisdiction to issue a final sanction. Nine months of Italian enforcement work was voided on that single procedural point.

Four large fines remain under active appeal. Meta's EUR1.2 billion penalty for illegal US data transfers - still the largest individual GDPR fine ever issued - continues through the Irish courts. TikTok is contesting its EUR530 million fine from the Irish DPC, awarded in April 2025 for transfers to China; Ireland's High Court granted a conditional stay in November 2025, meaning the transfer-suspension order is not currently being enforced. Two further Meta fines of EUR265 million and EUR91 million are also under challenge.

The geographic concentration of enforcement is remarkable. Ireland accounts for EUR4.04 billion - roughly 66% of the EU total - from just 37 enforcement actions, a product of GDPR's one-stop-shop mechanism that routes enforcement for US technology platforms through the country where those companies hold their European headquarters. France sits second with EUR888.1 million across 81 actions. Spain leads on volume with 1,070 enforcement actions but only EUR137 million in total fines, reflecting a very different enforcement style.

The enforcement numbers arrive exactly as the European Commission attempts to rewrite parts of the regulation. France's advertising trade association, Alliance Digitale, has now formally entered that debate.

France's ad industry draws its lines on the Digital Omnibus

On May 30, PPC Land reported on a formal position paper submitted by Alliance Digitale - the French digital marketing body representing 300 member organisations, spanning agencies, publishers, data providers, and ad tech platforms - to the co-legislators now debating the European Commission's Digital Omnibus legislative package. Published May 21, 2026, the document lays out 17 recommendations, some welcoming proposed changes and others demanding deletions.

The most forceful rejection concerns Article 88b, which would require consent to be managed centrally at browser or operating system level. Alliance Digitale calls for the article to be deleted entirely, on four grounds. Economically, browser-level default refusals would structurally lower acceptance rates for advertising tracking - pushing budgets toward walled gardens, where similar restrictions do not apply. Competitively, concentrating consent signalling at browser level hands de facto standard-setting power to a small number of dominant platform operators. Practically, the association argues users would express preferences at system level without understanding consequences such as receiving a downgraded service or losing access to content that was previously free.

The six-month ban on re-requesting consent after a user declines, introduced by Article 88a(4), also faces a call for deletion. Alliance Digitale's technical argument: reliably identifying the same person across different devices, browsers, and authenticated versus non-authenticated environments for the purpose of enforcing a six-month block is, in the association's words, "technically challenging, if not impossible."

What the association does support is equally specific. It backs the Commission's codification of the CJEU ruling in EDPS v Single Resolution Board, which established that information is not personal data for an entity that lacks the means to re-identify the individual - a provision particularly relevant for pseudonymised data flows through complex advertising value chains. It also supports the creation of an explicit legitimate interest basis for AI system development under Article 88c, though it objects to the unconditional right to opt out of AI model training, warning that fraudsters could exploit the mechanism to escape detection systems.

Yelp dominates AI local discovery - by a margin that defies the competition

A separate study published May 30 added hard numbers to a trend that the marketing industry has discussed without much granular data. A Foundation Marketing analysis of more than 28 million AI-generated responses, conducted in collaboration with AirOps and reported by PPC Land, found Yelp accumulating 512,680 total citations across ChatGPT, Gemini, Perplexity, and Google AI Mode in the fourth quarter of 2025. That puts it 3.4 times ahead of the Better Business Bureau, which ranked second with 149,710. Angi placed third at 145,633.

The platform breakdown is stark. Google AI Mode alone accounted for 340,721 of Yelp's citations - approximately 66% of its total - while Perplexity added 146,196. ChatGPT and Gemini together contributed just over 25,700. The two platforms where Yelp is weakest account for only 5% of local discovery citation volume overall; the platforms where it leads represent the other 95%.

Proximity-intent queries amplify the gap further. Yelp generated approximately 290,000 "near me" citations across AI platforms in Q4 2025, while all five competing platforms combined produced around 87,000. The "near me" segment represents just 17.5% of queries but 34.2% of all Yelp citations - a 2x overrepresentation. In plumbing specifically, 24 unique "plumber near me" questions on Perplexity generated 1,168 Yelp-cited answers, a 49x multiplier. The implication for local advertisers is direct: the platforms where AI sends users for local discovery are not the ones advertisers have historically prioritised.

Measurement gets an MCP bridge

One technical development from May 28 fits alongside these platform stories. AdExchanger reported that media measurement firm Measured launched a Model Context Protocol (MCP) server allowing brands to query their incrementality data directly through ChatGPT, Claude, Gemini, and other AI platforms. The system draws on aggregated and anonymised data from over 30,000 incrementality tests across more than 200 clients, some spending hundreds of millions in paid media annually. The bridge lets an advertiser type a plain-language question into a chatbot and receive an answer grounded in Measured's attribution dataset. CEO Trevor Testwuide framed the design as a direct response to where strategic media decisions are beginning to happen: in AI interfaces rather than traditional dashboards.

Also noted

Share this article
The link has been copied!