Google's rollout of AI ad labeling across its advertising products, completed over a five day span from July 9 to July 13, 2026, places every ounce of legal responsibility for the disclosure on the advertiser who builds the ad, while both Google and the publisher who hosts it carry none of it, according to the company's own published documentation.

Two announcements, four days apart, describe two sides of a single mechanism, and reading them together answers a question that neither one poses directly: who is actually on the hook if an AI-generated ad fails to meet a country's disclosure law. Google's July 9 changelog entry, titled "Updates to AI labeling requirements (July 2026)," introduced the AI label setting for advertisers across five products. Google's July 13 notice, published through the AdSense Help Center, confirmed that publishers using AdSense have no equivalent setting, no configuration step, and no compliance duty of their own. Sitting between those two facts is a line Google states about its own product, not about publishers or advertisers specifically, and it settles the liability question more directly than either notice does on its own: neither disclosure route the company offers "guarantees compliance with any specific regulation on its own," and advertisers are told to "seek legal guidance and confirm that their ads and assets meet local obligations."

That sentence is doing most of the work in this story. It is Google declining to certify its own feature. A label applied through the AI label setting, or a custom label an advertiser designs and attaches independently, both satisfy Google's own interface requirements, but neither one is Google's assurance that a specific ad in a specific country has met that country's law. The responsibility to make that determination, according to Google's documentation, sits with the advertiser alone.

What each announcement covers

Reported by PPC Land the following day, the July 9 changelog describes the mechanism from the advertiser's side in detail. It introduced the AI label setting across five products: Google Ads, Display & Video 360, Campaign Manager 360, Merchant Center, and Ads Editor. Inside those interfaces, an advertiser designates whether a given creative asset was created or edited using AI. Two disclosure layers follow from that designation. The first is global: any labeled asset triggers a disclosure inside the "How this ad was made" section of My Ad Center, reachable through the three-dot menu on ads served across Search, YouTube, and Discover, regardless of where the ad runs. The second is regional: for campaigns targeting the European Union, India, or New York specifically, a labeled asset also triggers a visible overlay directly on the ad itself, meaning the identical creative can look different depending on where it serves.

The July 13 notice, addressed to publishers rather than advertisers, is far shorter, running to two paragraphs inside the AdSense Help Center's Announcements section. It states that Google is introducing tools to help advertisers identify and label AI-generated ad creatives beginning July 13, that these labels may appear on creatives served to publisher properties when the underlying demand comes from Google Ads or Display & Video 360, and that "Google's sell-side platforms do not offer AI labeling features." No AdSense-specific setting accompanies the notice, because none is needed. The designation, and everything that follows from it, happens entirely inside the advertiser's own tools before the ad ever reaches a publisher's inventory.

Reading the two together produces a fuller picture than either supplies alone. Advertisers gained a detailed, product-specific interface for managing a disclosure obligation that carries real legal weight in named jurisdictions. Publishers gained confirmation that the same obligation does not extend to them, because they are not the party making the disclosure decision. What neither announcement states outright, but what follows logically from combining them, is that Google itself occupies a third position: it built the mechanism, but its own documentation refuses to stand behind the outcome.

Where liability actually sits

Google's documentation names three distinct parties in this arrangement, and each carries a different degree of exposure. The advertiser bears the compliance obligation directly. Regulations in the European Union, India, and New York require disclosure when an ad contains certain AI-generated or AI-edited assets, and it is the advertiser who selects, or fails to select, the label reflecting that fact. Google's own guidance is explicit that this selection is a compliance decision the advertiser must get right independently, since the tool itself carries no guarantee.

The publisher, per the July 13 notice, bears no obligation at all. A publisher's site is the surface on which a labeled or unlabeled ad happens to render. The label, if one appears, was generated by the advertiser's designation, not by anything the publisher configured, and the AdSense notice draws that boundary explicitly by stating that no sell-side labeling feature exists. A publisher cannot be found non-compliant with a disclosure law over a decision they had no mechanism to make.

Google occupies the middle position, and it is the position hardest to see without placing the two announcements side by side. Google built the interface, publishes the guidance, and, in specific automatic-labeling circumstances, applies labels advertisers never selected. Per Google's own documentation, described in PPC Land's coverage of the July 9 rollout, the company can apply a label without advertiser input under three conditions: when Google itself is legally required to label an ad in a given region, when it receives signals from other platforms indicating AI involvement, or when an advertiser uses one of Google's own fully automated tools to generate and serve creative. In each of those cases, the resulting label cannot be overwritten or removed. Yet even that automatic mechanism does not convert Google into the compliance-bearing party. The documentation's disclaimer, that neither route guarantees compliance with any specific regulation, applies to the feature as a whole, including the instances where Google applies the label itself.

The custom label alternative and its own conditions

A detail in the July 9 changelog sharpens the liability picture further. Advertisers who prefer not to rely on Google's platform-generated disclosure can build a custom label using their own design tools and attach it to the creative directly, and Google's documentation confirms this approach will not run afoul of the company's separate policies banning text overlays and watermarks on ad creative, provided the label functions as a disclosure rather than as decorative or misleading text. That option exists precisely because Google is not underwriting compliance through its own interface. An advertiser who wants stronger assurance than Google's setting offers can, and in some cases must, look outside Google's tools entirely, designing their own disclosure and taking on full authorship of whatever legal argument that disclosure supports.

Two technical constraints attach to that custom-label route, and both fall on the advertiser to manage. Google's guidance specifies that a custom label should sit outside a margin of roughly 5.5 percent around an asset's full perimeter, since ad rendering can trim the outer edge of an image and a label placed too close to the border risks being cut off entirely. Separately, Google's own automatic image-enhancement features, which crop and adjust assets to fit different placements, can remove a custom label from frame without warning, meaning an advertiser using a custom label typically needs to disable those enhancements, a control that differs by campaign type. Neither constraint is something Google verifies on an advertiser's behalf; both are disclosed as risks for the advertiser to manage.

Regulatory deadlines make the liability question concrete

The timing gives the liability question a hard edge rather than a theoretical one. The European Union's AI Act, formally Regulation (EU) 2024/1689, imposes transparency obligations under Article 50 that become applicable on August 2, 2026, a date PPC Land has tracked since Google first added the "How this ad was made" panel to My Ad Center. Those obligations require disclosure when AI-generated or AI-edited content is presented to the public, with enforcement carrying penalties reaching 15 million euros or 3 percent of global annual turnover, whichever is higher. Those penalties attach to the entity legally responsible for the disclosure, and Google's documentation, combined with the two announcements, points to the advertiser as that entity in every scenario except the narrow, automatic-labeling circumstances Google controls directly, none of which shift ultimate compliance responsibility away from the advertiser either.

The European Commission has built supporting infrastructure for that same deadline. The Commission published a free set of AI labeling icons through its Shaping Europe's Digital Future portal, with the materials last updated June 10, 2026. The icon set includes a basic disclosure icon, a "Fully AI-Generated" icon, and a "Partially AI-Modified" icon, and the Commission's own guidance states plainly that using the icons is optional while the underlying labeling requirement is not, and that deploying the icon set does not by itself establish legal compliance with Article 50. That caveat mirrors Google's own disclaimer almost exactly: a tool, whether built by Google or by Brussels, does not substitute for a legal judgment the responsible party must still make.

That deadline has held despite legislative pressure to move it. Overnight negotiations in Brussels over the Digital Omnibus package failed to shift the August 2, 2026 date, even as parliamentary committees debated extensions for a separate category of obligations covering high-risk AI systems. A provisional agreement reached between the Council of the EU and the European Parliament on May 7, 2026 pushed compliance for those high-risk systems to December 2027 and August 2028 respectively, but left the Article 50 transparency duty untouched. India and New York, the two additional jurisdictions Google names directly, carry separate disclosure requirements outside the EU framework entirely. In every one of those jurisdictions, the compliance burden, and by extension the legal exposure if it goes unmet, runs to the advertiser making the disclosure.

The watermark that operates regardless of any label

A separate layer, embedded automatically rather than selected by anyone, complicates the liability picture in one narrow respect. Google applies SynthID, a digital watermark developed by DeepMind, alongside metadata conforming to the C2PA standard maintained by the Coalition for Content Provenance and Authenticity, to any image or video generated through Google's own AI tools. That watermarking layer was first introduced by DeepMind in 2024. The system embeds the mark directly into an image's pixel data at the point of generation, a signal that survives cropping, filtering, and compression rather than sitting in metadata that can be stripped away.

Because this watermark persists independently of whatever an advertiser selects in the interface, it functions as a check that operates even if an advertiser declines to apply a label, or later strips one out. It does not, however, change who bears legal responsibility for disclosure. The watermark is a technical signal detectable by systems built to read it. It is not itself a public-facing disclosure that satisfies a transparency law, and Google's documentation does not present it as one. An advertiser relying on the watermark alone, without applying the interface-level label a given jurisdiction requires, would still be the party exposed if a regulator found the disclosure inadequate.

A pattern that recurs across Google's own product lines

The liability structure in this rollout is not without precedent inside Google's broader advertising and content businesses. YouTube introduced mandatory AI content disclosure for creators in May 2025, establishing much of the automatic-detection logic, including non-adjustable labels, that reappears in the July 9 advertiser-facing setting. In that earlier case as in this one, the party uploading the content carries the disclosure obligation, while YouTube's platform-level detection functions as a backstop rather than a substitute for the creator's own compliance judgment.

Advertiser sentiment complicates the picture further, and gives the liability question a practical urgency beyond the regulatory deadline itself. Research the IAB published in January 2026 found a wide gap between how advertisers and consumers feel about AI-generated ads, with far more executives assuming audiences view AI-made ads positively than consumers who actually report feeling that way. A labeling mechanism does not close that sentiment gap by itself, and Google's declining to guarantee compliance means advertisers cannot treat the label as a shortcut past the underlying question of whether disclosure was handled correctly, only as one input into a legal judgment they still have to make themselves.

For teams managing creative across more than one Google product, the practical complication compounds the liability question rather than simplifying it. An asset labeled inside Google Ads does not inherit that status inside Display & Video 360 or Campaign Manager 360 if uploaded separately into each, since every product keeps its own independent AI label field. An advertiser who correctly labels an asset in one product but forgets to repeat that step in another has, in Google's framework, created exactly the kind of gap the underlying regulations are meant to close, and the responsibility for closing it remains theirs across every product where the asset appears.

Why this matters for the advertising and publishing community

For an industry trade publication like PPC Land, the value of reading these two announcements together lies in what they jointly clarify: a rollout that could be mistaken for shared responsibility across Google's advertising ecosystem is, on the company's own terms, concentrated almost entirely on one party. Advertisers received a five-product interface, a detailed set of workflows, and an explicit warning that none of it guarantees anything. Publishers received two paragraphs confirming they are outside the compliance chain altogether. Google itself, the party that designed both the automatic labeling logic and the disclaimer accompanying it, avoids taking on the underlying legal risk even in the narrow cases where its systems apply a label no human selected.

That concentration has a direct bearing on how advertising teams should read the August 2, 2026 deadline. It is not a shared industry milestone in the way a platform-wide feature launch might be. It is a legal deadline whose consequences, according to Google's own documentation, land on whichever advertiser's name is on the ad, regardless of which Google product served it, which publisher's site hosted it, or which of Google's own automatic detection systems flagged it first. Publishers can treat the July 13 notice as confirmation that no new task awaits them. Advertisers do not have an equivalent confirmation available. The only sentence in Google's own documentation resembling reassurance is the warning that reassurance is not on offer.

Timeline

  • May 16, 2024 - DeepMind introduces SynthID, a digital watermarking system for AI-generated images and video.
  • August 1, 2024 - The EU AI Act, Regulation (EU) 2024/1689, enters into force.
  • May 21, 2025 - YouTube's mandatory AI content disclosure policy for creators takes effect.
  • September 4, 2025 - The European Commission opens a stakeholder consultation on Article 50 transparency guidelines, running until October 2, 2025.
  • January 2026 - The IAB publishes research documenting a gap between advertiser and consumer sentiment toward AI-generated ads.
  • May 7, 2026 - The Council of the EU and European Parliament reach a provisional agreement on high-risk AI system deadlines, leaving the Article 50 transparency date unchanged.
  • June 10, 2026 - The European Commission's AI labeling icon set is last updated on the Shaping Europe's Digital Future portal.
  • June 24, 2026 - Google releases Google Ads API v24.2, adding synthetic content labeling structures ahead of the Article 50 deadline.
  • July 9, 2026 - Google publishes "Updates to AI labeling requirements (July 2026)," introducing the AI label setting for Google Ads, Display & Video 360, Campaign Manager 360, Merchant Center, and Ads Editor.
  • July 12, 2026 - PPC Land publishes its detailed report on the July 9 rollout.
  • July 13, 2026 - Google publishes AI labeling clarification for AdSense, confirming publishers take on no configuration task or compliance duty.
  • August 2, 2026 - EU AI Act Article 50 transparency obligations become applicable.

Summary

Who: Google, through two separate publications, its July 9, 2026 advertiser-facing changelog and its July 13, 2026 AdSense Help Center notice, addressed advertisers running campaigns on Google Ads, Display & Video 360, Campaign Manager 360, Merchant Center, and Ads Editor, and publishers monetizing websites through AdSense.

What: Google introduced an AI label setting letting advertisers disclose AI-generated or AI-edited creative, then confirmed that publishers have no equivalent tool and no compliance duty of their own. Google's documentation states that neither the automatic labeling route nor an advertiser-built custom label guarantees compliance with any specific regulation, placing the legal responsibility for accurate disclosure on the advertiser in every case, including the narrow circumstances where Google applies a label automatically.

When: The two announcements were published four days apart, on July 9 and July 13, 2026, with PPC Land's detailed report on the first following on July 12, 2026, twenty days before the EU AI Act's Article 50 transparency obligations become applicable on August 2, 2026.

Where: The advertiser-facing setting applies across five Google products globally, with a visible on-ad overlay restricted to campaigns targeting the European Union, India, and New York. The publisher-facing clarification applies to AdSense accounts globally that receive demand from Google Ads or Display & Video 360.

Why: Regulations in the European Union, India, and New York require disclosure when ads contain certain AI-generated or AI-edited assets, with the EU's penalties reaching 15 million euros or 3 percent of global annual turnover for non-compliance. Google built the disclosure mechanism but explicitly declined to guarantee that using it satisfies any specific law, meaning the party legally exposed if a disclosure is wrong or missing is the advertiser, not Google and not the publisher whose site happened to display the ad.