On July 9, 2026 Google logged a policy update — "Updates to AI labelling requirements (July 2026)" — that turns AI disclosure from a label Google shows viewers into an obligation the advertiser has to meet. If an ad's image or video creative was generated or edited with AI, you now have to declare it, and the requirement reaches five of Google's advertising products at once: Google Ads, Display & Video 360, Campaign Manager 360, Merchant Center, and Ads Editor. This is the advertiser-side companion to the consumer-facing "How this ad was made" panel — the panel is what a viewer sees, this is the rule that says you have to put it there. This guide is the practitioner read on the obligation itself: exactly what the policy requires, the five surfaces it touches, the two ways to satisfy it (and the auto-applied labels you cannot override), where the disclosure actually appears and why the EU, India, and New York get an on-ad overlay while everywhere else gets a panel note, why it targets images and video specifically, and the part most write-ups miss — that a per-asset disclosure rule is really a per-asset record-keeping problem, and record-keeping is exactly where a messy, tool-sprawl production process falls apart.
On July 9, 2026 Google logged a change to its Advertising Policies titled "Updates to AI labelling requirements (July 2026)" (Google Advertising Policies Help). The change is small to read and large to comply with: it turns AI disclosure into something the advertiser is responsible for, rather than something Google merely displays. If the image or video creative in your ad was generated or edited with AI, you now have to declare it. The requirement is rolling out gradually through July 2026 rather than flipping on in a single moment.
It helps to hold two things apart, because most coverage blurs them. There is the consumer-facing "How this ad was made" panel — the note a viewer can open to see whether an ad was made with AI — and there is this, the policy that says advertisers have to feed that panel. The panel is the display surface; this is the obligation behind it. This guide is the practitioner read on the obligation: what the policy requires, the five products it reaches, the two ways to satisfy it and the auto-labels you cannot override, where the disclosure shows and why region changes it, why it is aimed at images and video, and the operational reality underneath — that a per-asset disclosure rule is really a per-asset record-keeping problem. For the display-side mechanics and the provenance stack beneath the panel, the companion is Google adds AI-generated ad disclosure; the news write-up of the policy itself is at Google now requires advertisers to declare AI-generated ad creative.
Strip it to the obligation. An ad that contains image or video creative generated or modified with generative AI has to carry a disclosure, and the responsibility for making sure it does now sits with the advertiser. That is the whole substance of the shift — before, Google labeled ads made with its own tools and left the rest alone; now the rule reaches out and asks you to account for creative Google did not make. It is the advertiser-side counterpart to the transparency Google is showing consumers, closing the loop between "here is a place viewers can check" and "here is who is required to fill it in."
The framing that keeps you out of trouble is provenance. The policy is not asking whether your ad is good, or whether AI is allowed — it plainly is — but whether you can state, per creative, that AI was involved in the pixels. That reframes the question from a creative one to a bookkeeping one: for every image and every video you run, do you know how it was made? Advertisers who can answer that per asset satisfy the rule with a checkbox; advertisers who cannot are guessing, and guessing under a disclosure policy is how you end up either over-declaring clean creative or under-declaring synthetic creative, both of which are avoidable.
The scope is what makes this more than a Google Ads footnote. The obligation spans five of Google's advertising products — Google Ads, Display & Video 360, Campaign Manager 360, Merchant Center, and Ads Editor — and the specific control you use differs by product. In Google Ads it lives as an option in the asset library; in Merchant Center it is a three-dot menu on the item; the other surfaces have their own equivalents. For a solo creator running a single Google Ads account this is one setting to learn. For anyone buying across the stack — an agency in DV360 and Campaign Manager 360, an e-commerce brand feeding Merchant Center, a team bulk-editing in Ads Editor — it is the same obligation expressed five different ways, and each surface is a place the declaration can be missed.
The practical takeaway is that the rule is not "check a box in Google Ads" but "handle AI provenance everywhere your creative enters Google." Merchant Center is the one people forget, because product-feed imagery does not feel like "an ad" the way a video campaign does — but AI-generated or edited product images flow through it and fall under the same requirement. If any part of your creative pipeline produces synthetic visuals, the declaration has to follow those assets into every one of the five tools they end up in, which is only manageable if you know their origin before they get there.
There are two ways to satisfy the requirement. You can add the label to the creative asset yourself, or you can turn on a new "AI label" setting and let Google attach the disclosure programmatically once the asset is flagged as AI. Most advertisers running third-party creative will use one of these, because they are the cases where Google has no independent signal about how the file was made and is relying on you to say. Enabling the setting is the low-friction path; it does not, Google is careful to note, "guarantee your compliance with specific regulations," so it is a tool for meeting the platform policy, not a substitute for understanding the law that applies to your campaign.
The important asterisk is that some labels are not yours to control. When Google applies the disclosure automatically — because it received an AI signal from another platform, because the ad was built with Google's own generative ad tools, or because a local law requires it — that label cannot be overwritten by the advertiser. So the honest mental model is: you get a say only in the one case the rule is really about, third-party AI creative Google cannot otherwise detect, and even there the direction of travel is toward detection rather than self-report. Treating the self-declare control as optional — running synthetic creative and leaving the box untouched — is a bet against detection that is getting worse to make, the same losing bet covered on the Meta side in AI-generated ads disclosure and UGC-style creatives, where automated identification of third-party AI media is already live.
Where the disclosure actually appears to a viewer depends on who the campaign targets. For campaigns aimed at jurisdictions that mandate on-ad transparency — the European Union, India, and New York — Google shows a visible overlay on the ad creative itself. Everywhere Google serves ads, the disclosure surfaces in the "How this ad was made" section of the My Ad Center panel, the per-ad info surface reached from the three-dot menu. So the same creative can wear its AI label openly for a European audience and carry it quietly in the panel for a US audience outside New York, driven entirely by the rules that apply where the impression lands.
That regional split is the reason a global advertiser cannot maintain one labeling posture for regulated markets and none elsewhere without building operational complexity you probably do not want. It is far simpler to know the provenance of every asset up front and declare consistently than to run a per-region matrix of which creative gets flagged where. The timing underlines it: the EU AI Act's Article 50 transparency obligations become enforceable on August 2, 2026, just weeks after this policy landed, and they require AI-generated and manipulated content to be marked in a machine-readable way and disclosed to people who encounter it. Google's rule is the advertising-shaped expression of that regulatory pressure, which is why the covered regions read like a regulator's list rather than a product manager's. A separate, older rule still stands beside it — election advertisers must disclose synthetic or digitally altered content through a dedicated checkbox in campaign settings — a reminder that disclosure started as a narrow carve-out and is generalizing outward. The wider regulatory backdrop for synthetic personas specifically is in regulating humanlike AI.
The rule is written around AI-generated or edited image and video assets, and that focus is deliberate. Synthetic visuals are where the deception risk lives — a fabricated product shot, a person who does not exist reading a script, an edited scene that never happened — so that is where regulators and platforms have concentrated the transparency requirements. Text ad copy written or polished with AI is not the object of this particular rule. The consequence for creators is specific: if your paid creative leans on generated or heavily AI-edited images and video, you are squarely the audience for this policy, and the declaration is per visual asset, not per campaign.
This is also why the move toward provenance you can detect matters more than the label you can type. Google's automatic path leans on SynthID, its imperceptible watermark, precisely because a signal embedded in the pixels does not depend on anyone choosing to be honest — and the industry has been converging on the same rails, C2PA Content Credentials for the signed manifest and SynthID-style watermarking for the embedded signal, throughout 2026. The self-declared checkbox is the weak half of the system by design; the durable half is the machine-readable provenance underneath, explained in depth in the display-side companion guide. For the advertiser, the lesson is not to optimize the checkbox but to make sure the answer it records is true, because the check behind it is getting harder to fool.
A per-asset disclosure rule is, operationally, a per-asset provenance record. The moment you have to attest "this creative was made with AI" or "this one was not," you need to actually know — and that is precisely where a sprawling, improvised production process breaks. If your ad creative comes from four different tools, a freelancer's deliverable, a stock library, and a generator you tried once, reconstructing which asset was AI-touched after the fact is guesswork, and it gets worse under the exact condition paid advertising runs on: volume and iteration. Direct-response campaigns live on testing, which means many variants of the same idea, each with its own creative history. Multiply per-asset attestation by a testing-heavy variant workflow across five products, and the cost of the rule is not the checkbox — it is knowing what to put in it, forty variants deep.
That is the frame that actually decides how much this policy costs you. Advertisers who generate their creative through one governed pipeline, where every asset's origin is known at the moment it is made, treat the disclosure as a lookup. Advertisers whose creative arrives from everywhere treat it as an investigation. The rule is the same for both; the difference is entirely upstream, in whether provenance was recorded at production time or has to be excavated at declaration time. The move that removes the burden is not a better spreadsheet after the fact — it is a production process where the answer already exists. The step-by-step of setting the disclosure controls on the platforms that have them is in how to disclose AI-generated ads; the broader shift of ad creation into the ad platforms themselves is in AI ad generation moves inside the ad platforms.
The specific problem this rule creates is not "AI is restricted" — it is "you now have to attest, per asset, across a variant-heavy paid workflow, how each piece of creative was made." Kompozy answers that at the source, and its angle here is deliberately the record-keeping one rather than volume or organic reach. Because Kompozy generates the creative — it does not hand you an anonymous file to upload and later interrogate — every output has a known origin baked in: which model drafted the copy, which produced the image, which rendered the video, tied to a persona identity you own. When a Google Ads asset-library toggle or a Merchant Center three-dot menu asks whether AI was involved, you are not reconstructing history from four tools; you are reading a fact your pipeline already recorded. That is the exact posture a per-asset attestation rule expects, and it is the difference between the disclosure being a lookup and being an investigation.
It scales the way the paid workflow actually runs, which is where the burden really bites. Direct response demands variants — different hooks, different first frames, different framings of the same product — and under this rule every one of those variants is a separate asset you have to be able to declare. Kompozy produces variants as regenerations rather than reshoots: a Persona Short with three alternate hooks, a Carousel in two layouts, a set of Photo Posts and Persona Photos for a product feed, each coming out of the same governed pipeline with the same clean provenance. So the forty-variants-deep problem — the one where after-the-fact attestation collapses — never forms, because the origin of the fortieth variant is recorded identically to the first. A Persona Brief holds all of them to one voice and banned-word list, and a per-post review pipeline with quality gates means a human signs off before anything ships, so what you declare and what you publish stay in lockstep.
The honest boundary matters too, and it is worth stating plainly: this rule governs paid ad creative, and Kompozy is not an ad-buying tool — it does not run your Google Ads account or flip the AI-label settings for you. What it does is make the thing you are declaring, cleanly enough that declaring it is trivial, and then it keeps generating past the edge of the rule. The policy covers image and video ad assets; it says nothing about the organic surface, and Kompozy turns one concept into that full surface — Clipped Shorts, Blog Articles, Email Newsletters, Quote Graphics, Text Posts — reframed to 9:16, 1:1, and 16:9 and fanned across nine social platforms plus blog and email from one queue with autopilot and scheduling. So the division of labor is clean: Google supplies the disclosure controls, you set them truthfully, Kompozy makes creative whose provenance is never in question and produces the owned distribution the ad rule never reaches.
Google's July 2026 update makes AI disclosure an advertiser obligation: if your ad's image or video creative was generated or edited with AI, you have to declare it — either by labeling the asset yourself or by enabling an "AI label" setting for Google to attach it — across Google Ads, Display & Video 360, Campaign Manager 360, Merchant Center, and Ads Editor. Auto-applied labels cannot be overwritten, on-ad overlays show for EU, India, and New York campaigns while everywhere else gets the My Ad Center note, and the rule aligns with the EU AI Act's Article 50 becoming enforceable on August 2, 2026. The mistake is treating it as a checkbox problem. It is a provenance problem: a per-asset rule is only easy for advertisers who already know how each asset was made, and hard for everyone reconstructing it after the fact. Fix that upstream — generate your creative through a pipeline where the answer already exists — and the disclosure stops being a compliance tax and becomes a line you can fill in without thinking. Treat the per-product controls, covered regions, and rollout timing as a launch-window snapshot and confirm the current details in Google's Advertising Policies help pages.
It requires advertisers to declare ads whose image or video creative was generated or modified with AI. You satisfy it one of two ways: add the label to the creative asset yourself, or turn on a new "AI label" setting and let Google attach the disclosure programmatically. Google logged the update as "Updates to AI labelling requirements (July 2026)" on July 9, 2026, and it is rolling out gradually through July 2026 rather than switching on in a single moment.
Five advertising products at once: Google Ads, Display & Video 360, Campaign Manager 360, Merchant Center, and Ads Editor. The control differs by product — for example an asset-library option in Google Ads and a three-dot menu in Merchant Center — so an advertiser buying across the paid stack has to handle the same obligation in several tools, not one place.
Not when Google applies it automatically. If Google received an AI signal from another platform, the ad was built with Google's own generative tools, or a local law requires the label, the disclosure sticks and the advertiser cannot overwrite it. You control the label only in the case where Google has no signal of its own — creative you made with a third-party tool — which is exactly the case the rule is asking you to self-declare.
It depends on the audience. Campaigns targeting jurisdictions that mandate it — the European Union, India, and New York — get a visible overlay on the ad itself, while everywhere Google serves ads the disclosure surfaces in the "How this ad was made" section of the My Ad Center panel, reached from the ad's three-dot menu. Google is explicit that turning on its label setting "doesn't guarantee your compliance with specific regulations," so the switch is a tool, not a legal shield.
No. The requirement is written around AI-generated or edited image and video assets in paid ads, so text-only ad copy is not the focus and organic content you publish outside of advertising is out of scope entirely. That boundary is why the practical response is two-part: track provenance cleanly on the paid creative you have to declare, and keep building the organic distribution the ad rule never touches.
In July 2026 Google made AI disclosure an advertiser obligation, not just a viewer label. Ads whose image or video creative was generated or edited with AI must be declared across five products — Google Ads, Display & Video 360, Campaign Manager 360, Merchant Center, and Ads Editor. Advertisers either label the asset themselves or enable an "AI label" setting for Google to attach it; auto-applied labels can't be removed. The rule aligns with the EU AI Act's Article 50, enforceable August 2, 2026.
Get started → · ← All guides · Compare Kompozy vs other tools