At Google Marketing Live in June 2026, Google announced Business Agent for Leads — a Gemini-powered AI chat embedded directly inside ad units. Instead of clicking through to a landing page, a user can ask questions, get answers, and hand over personal information without ever leaving the ad. The AI agent draws on the advertiser's website content to hold a conversation that, until now, would have happened on a page the advertiser controlled.

For anyone working in digital marketing or privacy, this isn't an incremental format update. It's a structural change to how ad interactions work, and it breaks fundamental assumptions that consent management has relied on for years.

What Google Actually Announced

Business Agent for Leads is an AI-powered conversational agent that lives inside a Google ad unit. When a user engages with the ad, they don't click through to the advertiser's site. Instead, a Gemini-powered chat opens within the ad itself. The agent can answer product questions, qualify leads, and collect contact information — name, email, phone number — all before a single page view is registered on the advertiser's domain.

Google is piloting the format in three sectors initially: education, automotive, and real estate. These are high-consideration verticals where users typically research extensively before converting, and where the gap between ad click and form submission has historically been wide. The agent is designed to collapse that gap by handling the entire qualification conversation inside the ad.

The agent doesn't improvise. It ingests the advertiser's website content — product pages, FAQ sections, pricing information — and uses that material to generate responses. The advertiser provides the source content; Gemini handles the conversation. Google frames this as reducing friction: fewer clicks, faster lead capture, better conversion rates.

"That's Not an Ad Format. That's the End of the Landing Page."

"That's not an ad format. That's the end of the landing page."
— Simo Ahava

Ahava's observation cuts to the core of what's changing. The landing page has been the foundation of digital advertising measurement for two decades. It's where the analytics tag fires, where the cookie banner appears, where Consent Mode signals get set, where conversion pixels load, where the CMP does its work. Every piece of the consent and measurement stack assumes the user will visit a page that the advertiser controls.

Business Agent for Leads removes that page from the interaction entirely. The user sees an ad, talks to an AI, provides their information, and leaves. The advertiser's website was consulted by the AI during training or retrieval, but the user may never visit it. There's no page load, no tag execution, no banner render.

For consent management, this isn't an edge case. It's the main case — or it will be, if the format scales beyond the initial pilot sectors.

The Consent Problem: Who Manages Consent When There's No Page?

The consent management model — from the ePrivacy Directive to GDPR to Consent Mode — is built on an assumption: the user visits a page, and that page presents a consent choice before data collection begins. The CMP loads early, sets default consent states (ad_storage='denied', analytics_storage='denied'), shows a banner, and updates signals based on the user's choice.

With an in-ad AI agent, that sequence doesn't happen. The user types their name, email, and phone number into a chat inside Google's ad infrastructure. The advertiser's CMP never loads. Consent Mode signals are never set, because there's no tag configuration in this context.

This raises questions without clean answers:

Who is the data controller? The advertiser collects leads, but Google's agent does the collecting. Under GDPR, the entity determining purposes and means of processing is the controller. When Google's AI collects personal data on behalf of an advertiser, inside Google's infrastructure — is this joint controllership?
Where does the consent obligation sit? If an EEA user interacts with Business Agent for Leads, someone must obtain valid consent before processing personal data for advertising. The advertiser can't present their CMP because there's no page load. Google hasn't described an alternative mechanism.
What legal basis covers AI access to advertiser content? If the advertiser's site includes information about real people (team bios, testimonials), the AI is processing personal data to generate responses. Consent for displaying that data on the site may not extend to AI re-processing inside a Google ad.

Consent Mode Without a Page Visit

Google's Consent Mode v2 relies on JavaScript signals set on the advertiser's page. The CMP fires first, sets all consent signals to denied, the banner appears, the user chooses, and the CMP updates signals accordingly.

None of this applies inside an ad unit. There's no advertiser page, no gtag calls, no CMP. The ad_storage, ad_user_data, and ad_personalization signals — which, as of June 15, 2026, are the single authority over advertising data — don't exist in this context.

Google will need to handle consent internally for these interactions. That could mean a consent layer inside the ad unit, relying on Google account consent settings, or a different legal basis entirely. Whatever the mechanism, it's opaque to the advertiser — they lose visibility into whether consent was obtained, what was consented to, and how it's stored.

For advertisers in the EEA, UK, and US states with comprehensive privacy laws, this opacity is a compliance risk. Under GDPR, a controller can't outsource accountability. If the advertiser is the controller for leads collected via Business Agent, it must demonstrate that valid consent was obtained. "Google handled it" isn't a sufficient answer under Article 5(2).

AI Agents Accessing Advertiser Content: A New Privacy Surface

Business Agent for Leads doesn't just collect data from users — it also consumes data from the advertiser's website. The Gemini model reads product pages, FAQ content, and potentially any publicly accessible page on the advertiser's domain to generate responses.

This creates a new data processing relationship. The advertiser's website content becomes an input to Google's AI model, which generates novel outputs served in a context the advertiser doesn't control:

What content does the agent access? Is it a one-time crawl, continuous retrieval-augmented generation (RAG), or something in between? The answer determines the scope of the processing.
Can the AI hallucinate? If the agent generates an inaccurate response — a wrong price, an outdated policy — who bears responsibility? The advertiser didn't write the response, but it's their brand.
Does the privacy policy cover this? Most policies describe data collection on the website itself. They don't anticipate a third-party AI ingesting site content and re-presenting it inside an ad unit. Advertisers may need to update their processing records.

From Clickthrough Rates to Content Architecture

If the AI agent is the new front door — replacing the landing page as the first substantive brand interaction — then what you optimize changes fundamentally.

Traditional search advertising optimizes for clickthrough rate: get the user to the page. With Business Agent for Leads, the ad itself does the persuading, qualifying, and converting. The landing page becomes a reference corpus that the AI consults, not a destination the user visits.

This means the structure and quality of your website content becomes a direct input to ad performance — not because Google crawls it for Quality Score (that's been true for years) but because the AI literally reads it to construct conversations. Well-organized FAQ pages, clear product specs, accurate pricing — these become competitive advantages in a way they weren't when the user was reading them directly.

For digital marketers: you're no longer optimizing pages for humans to read. You're optimizing content for an AI agent to retrieve and synthesize. That's closer to knowledge management than copywriting.

Impact on the Initial Test Sectors

Education: Universities handle sensitive data categories — academic interests, career aspirations, sometimes health or disability information. An AI agent collecting this inside a Google ad, with no institutional CMP in the loop, creates GDPR special-category risks.

Automotive: Vehicle inquiries involve trade-in valuations, location data, and financing discussions — personal and potentially sensitive data collected in a context where the dealer's consent banner never fires.

Real estate: Property searches reveal budget ranges, household composition, and timeline. Agencies are typically data controllers under GDPR, accustomed to managing consent on their own platforms. An in-ad AI agent shifts that collection outside their control.

All three sectors share long consideration cycles with significant information exchange before conversion. That's why Google chose them — the AI agent adds the most value where the pre-conversion conversation is longest. But it's also where the consent gap is widest, because the data is the richest.

What This Means for CMP Providers

Consent management platforms, including CookieBeam, are built on a clear assumption: the CMP loads on the advertiser's page and mediates between the user and tracking technologies. Every feature — banner display, consent storage, Consent Mode signal management, audit logging — depends on the CMP being present at the point of data collection.

Business Agent for Leads introduces a collection point where no CMP is present. CMPs aren't irrelevant — the advertiser's website still needs consent management for direct traffic, organic search, and every channel that results in a page visit. But a growing share of ad-driven data collection could happen outside the CMP's reach.

The industry response needs to happen on multiple fronts:

Advocacy: Push Google to expose consent signals or integrate CMP preferences into ad-level consent mechanisms. If the ad unit needs its own consent layer, it should interoperate with the advertiser's existing configuration.
Audit tooling: Help advertisers track where data is collected across all channels, including those outside the CMP's direct control. If leads arrive from Business Agent without consent records, the advertiser needs to know.
Guidance: Most advertisers won't think about these implications until something breaks. CMPs are well-positioned to educate customers about consent gaps in new ad formats before they opt in.

What You Should Do Now

Google Marketing Live 2026 announced a direction, not a finished product. Business Agent for Leads is in pilot, limited to three verticals, and the full data processing details haven't been published. But the direction is clear enough to act on:

Audit your lead sources. If you're in education, automotive, or real estate — or expect to be offered the pilot — map where your leads come from today and identify which ones flow through your CMP-protected pages versus external channels.
Review your data processing agreements. Check your Google Ads data processing terms. Do they cover personal data collected by an AI agent inside an ad unit? If not, clarify with Google before opting in.
Check your privacy policy. Ensure it accounts for the possibility that an AI agent may access your website content to generate responses. If your site content includes personal data (employee bios, customer testimonials), assess whether this new processing activity is covered by your existing lawful basis.
Strengthen your consent infrastructure for what you can control. Your CMP still governs every direct page visit. Make sure your Consent Mode v2 implementation is solid — properly defaulting to denied, correctly updating on consent, covering ad_storage, ad_user_data, and ad_personalization. When data does flow through your site, it should be impeccable.
Ask Google the hard questions. Before enrolling in any in-ad AI agent program, ask: What consent mechanism does the ad unit use for EEA users? What legal basis does Google rely on? Will consent records be shared with the advertiser? Can the advertiser's consent preferences (configured via their CMP) be applied to in-ad interactions?

The landing page isn't dead yet, but Google just showed us what its replacement looks like. Consent management needs to follow the data — and right now, the data is starting to flow in places where no consent banner can reach.