Meta turns ad signals into AI signals

Meta is extending how it uses data it receives from businesses beyond advertising to user content and AI experiences. Meta isn’t just talking about campaign data; it’s talking about activity shared through pixels, APIs, apps, websites and commerce integrations.

The company said in a release yesterday:

“We’re updating how we use information that other businesses already share with Meta. We already use this data — like games you play or purchases you make on other websites — to make the ads you see more relevant. In the future, we’ll use this information to personalize other parts of your experience, including the content you see in your Feed and AI responses.”

Meta says it is not collecting any new data.

Read: “Better Personalization and Changes to Controls for Your Activity From Other Businesses” (June 9) – Meta

From tipsheet: This is Meta turning ad infrastructure into a broader personalization layer. The signal that once optimized ads is now being routed into content ranking and AI responses. As AI assistants or chatbots become a primary interface, the value of Meta’s advertiser signal network extends beyond advertising and into the personalization of the entire product experience.

Was: Ad platform.

Becoming: Personalization platform.

LLMs & CHATBOTS

Developments

Claude Fable 5 and Claude Mythos 5 (June 9) – Anthropic
Introducing the Third Generation of Apple’s Foundation Models (June 9) – Apple Machine Learning Research
Fluid, natural voice translation with Gemini 3.5 Live Translate (June 9) – Google The Keyword blog

DATA

Audience intelligence and data platform StatSocial has launched a new AI-powered product called Digital Twins that allows marketers to create AI models of audiences rather than recruiting traditional research panels.

StatSocial CEO David Barker tells AdExchanger:

“…the technology is a way to place hundreds of audience members into a virtual room and ask questions at scale. Users can upload creative assets, test messaging or conduct free-form interviews with AI-generated audience members whose responses are informed by behavioral data.

Dean McBeth at audience strategy consultant Shepherd spoke to the use case:

“No one has these panels, but what StatSocial provided [with Digital Twins] was hundreds of millions of people anonymized across tens of thousands of data points. We could get those behaviors and start to understand what people were doing online without asking them.”

Later, McBeth says the audiences he sees in the AI-simulated panels are showing up “in real life.”

Read:

“StatSocial’s New AI Tool Digital Twins Is Helping Shepherd To Pressure-Test Audience Insights” (June 9) – AdExchanger
“StatSocial Launches Digital Twins, Allowing Brands to Instantly Survey Virtually Any Audience in Hours Through AI Models Built on Real People” (June 9) – press release

From tipsheet: Running the (simulated) campaign before it ever really runs appears to be a part of the AI advertising future.

AI is moving upstream from execution into decision-making. If those decisioning models become accurate enough, research could become less about collecting responses from people and more about interrogating models of people.

Related: “Gopuff Knows What Snacks You Want, Thanks To This Composable CDP” (June 9) – AdExchanger

AGENTS

PubMatic builds for agentic media buying

PubMatic CEO Rajeev Goel announced on LinkedIn that his company, Havas, and Telefónica launched what they describe as Spain’s first agentic CTV campaign, using PubMatic’s AgenticOS platform to convert a natural-language campaign brief into a live programmatic campaign.

According to a press release, agents operating across its AgenticOS and Activate platforms ingested the brief, selected inventory, applied frequency and brand-safety controls, optimized bids, and managed campaign execution. Activate serves as the company’s buy-side execution platform.

PubMatic says setup time fell by as much as 80% and “CPM came in 18% below target and impressions exceeded target by 23%.”

Esther López, Head of Media, Telefónica, says in the release, “By working with Havas and PubMatic, we have demonstrated that agentic AI can operate with the level of rigour and performance that brands require. This is not an experiment. It is the foundation of a smarter and more responsible way to invest in media.

Read: “PubMatic and Havas Launch the First Agentic CTV Campaign in Spain for Telefónica, Achieving 18% Lower CPM vs Target” – PubMatic

From tipsheet: The campaign may demonstrate what agentic media buying can look like in practice, but the bigger story may be the protocols underneath it. PubMatic says AgenticOS “connects to partner agents, inventory and data partners through current and new protocols.” If agentic advertising evolves into networks of cooperating agents rather than isolated AI tools, protocols such as ARTF and AdCP could become critical infrastructure, allowing campaign objectives, decisions, and optimization signals to move between agencies, platforms, and publishers.

The long-term battle may be less about who has the best agent and more about who helps define how agents communicate.

SELL-SIDE

Hearst adds agents to audience intelligence

Two years ago, Hearst launched AURA, which used first-party data, contextual signals, commerce behavior and AI-powered audience modeling to create advertiser audiences.

Now Hearst is adding agents on top with AURA IQ.

Hearst Magazines SVP and Head of Data Solutions & Insights, Mike Nuzzo, tells Adweek, “We’ve been thinking about agentic for the better part of a year, and in the past eight months, we’ve put pen to paper. We’ve built an agentic platform, in that we have a [model context protocol] and agents underneath it that help us do task-based work that used to take a long time.”

Adweek’s Mark Stenberg reports that the agentic system addresses three workflows: content trend analysis, campaign analysis and “bespoke audience generation.”

SEARCH

Study: Visibility is not representation

Results from a study by AI visibility firm Bluefish were released yesterday. Findings included:

“Brands with well-established topic authority are preferred by AI models – Appearing in an AI answer is not the same as being described well, and the gap between them is wider than most teams assume.”

“LLMs highlight different brands depending on the audience – AI is highly personalized, responding to user context and intent when responding to queries.”

“AI citations provide key insights into how to best optimize for AI performance – AI does not weigh all sources equally, and the mix shifts sharply from one vertical to the next.”

Press release: “State of Enterprise Brands in AI: a benchmark of how the world’s largest brands are performing in AI” (June 9) – Bluefish

The report: “State of Enterprise Brands in AI” – Bluefish (email required)

AGENCIES

AI fears create agency opportunity

Analysts at investment research firm Berenberg are seeing an opportunity for investors in Publicis Groupe, WPP and Havas. According to Bloomberg, the analysts said this week that “jitters about the potential disruption caused by artificial-intelligence adoption has created a broad-based buying opportunity for the shares of European advertising agencies.” Read a bit more. (June 9)

From tipsheet: Agencies as AI beneficiaries — write it down.

COMMERCE MEDIA

AI shopping’s value proposition

On LinkedIn yesterday, eMarketer Chief Content Officer Vladimir Hanzlik said:

“In the past, declines in consumer confidence typically steered people away from online shopping and towards physical stores.

In my session at Shoptalk Europe, I argued that this time might be different, because value-seeking shoppers have some great online-first options at their fingertips:

Marketplace shopping is attractive for value-focused shoppers, because it lets them compare different brands and products in one place,

Cross-border shopping is attractive, because it gives shoppers access to better value and lower-cost products from abroad (including from China),

Recommerce is attractive, because it gives shoppers access to pre-owned products at lower prices, and

AI-assisted shopping is attractive, because AI assistants can help you find products at the best possible price…”

TECH

Ownership and infrastructure in the EU

Meta says WhatsApp is company property. The EU says it has become critical infrastructure.

A European Union regulator ruled Monday that Meta must provide AI chatbot competitors with free access to WhatsApp. The case highlights a growing debate over whether owners of dominant digital platforms can use control of distribution to advantage their own AI products.

Access to the API for WhatsApp for Business is at the center of the dispute, according to the European Commission’s statement:

“Meta has been, at first sight, abusing this dominant position by preventing competing general-purpose AI assistants from using the WhatsApp for Business Application Programming Interface (‘API’). On 15 October 2025 Meta introduced a new policy, banning third-party general-purpose AI assistants from the WhatsApp for Business API. As a result, only Meta’s own AI assistant, Meta AI, remained accessible on WhatsApp. The Commission concluded that, at first sight, this constituted a refusal to provide access to an infrastructure developed for and previously open to third parties. After a revision of the policy, on 4 March 2026, Meta accepted third-party general-purpose AI assistants again on WhatsApp, but charged a fee which, at first sight, is in practice equivalent to the previous access ban.”

EU antitrust chief Teresa Ribera said: “These interim measures will safeguard competition in the growing market for AI assistants, by preserving a key entry point to reach consumers in Europe — WhatsApp — and allowing AI companies to innovate, scale up and reach their full potential.”

Meta said the ruling would allow competitors to use WhatsApp Business “for free” and plans to appeal, according to Reuters.