Google’s AI Ad Disclosure Labels: An Advertiser Playbook

Google’s AI ad disclosure labels arrived on July 9, 2026: a “How this ad was made” panel inside My Ad Center that tells people whether the ad in front of them was created or edited with AI. The rollout is global — Search, YouTube, and Discover — and it is the first Google ad-disclosure requirement to reach beyond election ads since the 2023 political-ad policy.

The design carries a split that every advertiser needs to internalize. Ads built with Google’s own generative AI tools get the disclosure automatically — no action required. Ads built with third-party AI tools depend on the advertiser ticking a box, and Google has said it will not check whether that box is ticked truthfully. That honor system lands three weeks before the EU AI Act’s Article 50 transparency obligations become enforceable on August 2, 2026, with penalties reaching €15 million or 3% of global annual turnover.

This playbook covers what actually shipped, where the labeling control lives, which creative pipelines auto-label versus require manual disclosure, how the EU AI Act changes the stakes for EU-facing campaigns, how Meta’s stricter system compares, and the audit steps agencies should run this month. Everything below is sourced from Google’s announcement and help documentation, the EU AI Act text, and Meta’s newsroom — with single-source claims flagged as such.

The announcement came from Keerat Sharma, Google’s VP and General Manager for Ads Privacy and Safety, on July 9, 2026. The feature itself is an extension of My Ad Center — the existing surface that already handles ad blocking, “why this ad” explanations, and advertiser identity. Users tap the three-dot menu or info icon on any ad across Google Search, YouTube, or Discover and see whether the ad was created or edited with AI.

Two details separate this from a routine transparency update. First, the rollout is global rather than phased by market — Google is introducing it everywhere at once, though its own language is “introducing” and “rolling out,” so expect availability to settle in over the coming weeks rather than being fully live in every market today. Second, this is the first time Google has extended an AI-disclosure requirement beyond election ads. The 2023 election-ads rule — which required disclosure of synthetic or digitally altered content in political advertising — remains in force alongside the new general system.

The timing also makes it the second significant Google Ads policy move in just over a week: Google’s July 2026 Ads terms update on AI automation authority took effect July 1, rewriting what Google’s AI is allowed to do inside your account by default. July 2026 is shaping up as the month Google formalized both sides of the AI-advertising relationship — what its AI can do for you, and what you must tell people about yours.

The most consequential line in the launch coverage is about what Google will not do. Per TechCrunch’s reporting, when an advertiser self-declares that an ad was made with third-party AI, Google “will not perform its own check to determine if that’s the case.” The self-disclosure flow is genuinely opt-in and genuinely unverified. For anything created outside Google’s own tools, the entire system runs on advertiser honesty.

The contrast with Google’s other properties is instructive. YouTube already auto-labels AI-generated video content using detection-based systems, regardless of whether the creator discloses. On the ads side of the same platform, no equivalent detection applies to third-party creative — a stricter standard for creators than for advertisers, at least for now. Google’s own generative outputs do carry SynthID, an imperceptible watermark, which is what makes the automatic labeling of Google-made creative reliable. Nothing equivalent exists for the Midjourney render an agency uploads into a Performance Max asset group.

One more nuance matters for legal teams: the label’s placement is not always buried in the panel. Google’s announcement states that “based on local requirements, a label may also appear directly on the ad” — automatically or when the advertiser uses the control. In jurisdictions with disclosure mandates, the quiet panel entry can become a visible on-ad badge. Existing policy is otherwise unchanged: misleading ads remain prohibited whether AI-made or not, and advertiser-verification requirements still apply. The new feature adds transparency, not new content rules.

News coverage treats “third-party AI tools” as one generic bucket. In practice, agencies run specific pipelines, and each one lands differently under the new system. The matrix below maps the common creative paths against the disclosure trigger, label placement, and verification level — compiled from Google’s announcement, the Google Ads help documentation, and the tool vendors’ own product pages. No outlet has published this cross-reference; we built it because it is the first question every client asks.

The Canva row deserves emphasis. If your team generates ad creative in Canva Grow 2.0’s agentic ad-creation workflow — unveiled at Cannes Lions on June 25, 2026 — and pushes assets into Performance Max via Canva’s Google Ads app integration, that creative will not auto-label. Canva is a non-Google tool, so the advertiser must flag the assets manually in Asset Studio. This is precisely the scenario the self-disclosure control was built for, and precisely where an unaware team quietly accumulates unlabeled AI creative.

The advertiser-facing control lives in Google Ads under Asset Studio. In the Asset Library, you select an asset, choose “Manage AI label,” and designate it as created or edited with AI — or decline. A parallel flow surfaces the same choice during campaign creation and when generating new assets, via an AI label column with checkbox selection. And if a selected asset already carries a non-Google AI label, Google Ads surfaces a link back to where the asset was originally created so you can review it — the system reads externally set AI metadata even though it does not verify truthfulness.

For agencies managing creative at volume, the more important story is programmatic. According to PPC Land’s reporting on the Google Ads API v24.2 release, the API now exposes SyntheticContentInfo and SyntheticContentAttestation structures on both Asset and Ad resources — carrying AI-generation status plus a human-review dimension that distinguishes fully automated output from human-in-the-loop workflows. In the current release those attestation fields are read-only; PPC Land reports they become writable in v25, scheduled for July 2026, and that the synthetic-content structures were backported to several older API versions so teams can integrate without a full migration. We have not verified these version-level details against Google’s own API changelog, so treat the specifics as PPC Land’s reporting — but the direction is unambiguous.

The practical read: bulk AI-labeling for agencies running custom Google Ads API integrations is weeks away from being writable at scale, and the read-only window is the time to build the internal logic — which asset sources map to which attestation values — so that labeling ships the day mutation opens rather than months after.

Most coverage treated Google’s announcement and the EU AI Act as separate stories. They are one compliance timeline. Article 50 of the EU AI Act — the transparency chapter covering deepfake labeling and AI-generated content disclosure — becomes enforceable on August 2, 2026, twenty-four days after Google’s rollout began. A provisional extension to December 2, 2026 applies to machine-readable marking of generative AI systems already on the market. Penalties reach €15 million or 3% of global annual turnover, whichever is greater, enforced by national market-surveillance authorities in each member state.

The scope distinction is where advertisers get tripped up, so it is worth being precise. Article 50’s disclosure rule for AI-generated text is scoped to content published to inform the public on matters of public interest — most pure ad copy and corporate communication likely falls outside it. The deepfake rule is broader: AI-generated or manipulated image, audio, or video that resembles real people, objects, or events and would falsely appear authentic must be labeled even without any intent to deceive. Photorealistic AI imagery in ad creative can trigger that rule on realism alone. And marketing content styled to look like editorial — native advertising on health or finance topics, for instance — can fall back into scope of the text rule.

This is the trend to read correctly: platform self-regulation and statutory regulation are converging on the same requirement from opposite directions. Google’s system gives advertisers the mechanism; Article 50 supplies the penalty. Google’s help pages already name the EU, India, and New York as regions with disclosure requirements that can force an on-ad label — the company is mapping jurisdictions, not gesturing vaguely. The same dynamic played out when the EU’s transparency rules reshaped Meta’s political-ads policy: the platform feature arrives first, the legal deadline makes it non-optional shortly after.

Google’s announcement contains a deliberately vague phrase — “based on local requirements” — for when a label escalates from the My Ad Center panel to the face of the ad itself. Its help documentation is more specific, naming the EU, India, and New York as regions with disclosure requirements. The tracker below turns that line into something an advertiser can act on. Where Google’s documentation names a region without detailing the underlying rule, we say so rather than guessing — verify specifics with local counsel before relying on any row.

Google is not first to AI ad disclosure. Meta expanded its generative-AI ads transparency policy in February 2025, and the architecture is directly comparable: labels consolidate into an “About this ad” panel alongside “Why am I seeing this ad?” and the existing political-ad advertiser disclosures — structurally the same move Google just made with My Ad Center. Meta’s monetization policy director Pedro Pavón framed the rationale simply: “People want a central place to find relevant information about an ad.”

The differences are what matter for advertisers running both platforms. First, verification: Meta labels ads not only when its own generative tools create or significantly edit an image or video, but also when its detection systems flag third-party AI via industry-standard signals. Meta does not rely purely on advertiser self-report the way Google’s third-party flow does. Second, placement: most AI edits sit behind Meta’s three-dot menu, but when AI generates a photorealistic human, the label moves next to the “Sponsored” tag itself — a visibly stricter tier than Google’s uniform check-the-panel approach. Trade press has reported that undisclosed AI content has become one of the larger ad-rejection categories on Meta; we could not confirm that against Meta’s own published data, but the enforcement direction it suggests is consistent with the detection-based design.

The practical takeaway for cross-platform teams: on Meta, assume the platform can detect what you did not disclose; on Google, assume regulators eventually can. Either way, the era of quietly unlabeled AI creative is closing — the only question is whether the label gets applied by you, the platform, or an enforcement authority.

The right response depends on how your creative is produced and where your campaigns run. Four common situations, and what we would do in each this month.

Two adjacent moves round out the plan. First, if you are now labeling AI creative, measure whether it changes performance — a framework for testing AI-generated ad creative gives you the structure to isolate the disclosure variable rather than guessing. Second, make creative-origin tracking a standing column in your asset library, not a one-off audit: every asset that enters a Google campaign should carry a recorded origin (tool, date, human-review level) that matches what the label attests. If your team wants help wiring disclosure operations into an existing paid-media program — auditing pipelines, writing the SOP, building the API-side attestation logic — our paid media management services handle exactly this kind of platform-policy shift.