AI Labels on Google and Meta Ads: What Advertisers Must Declare

Most creative teams now use AI somewhere in an ad: a generated background, an extended product shot, a cleaned-up voiceover, a full video. Google and Meta have both changed how they show that to viewers. Some of it happens automatically, some depends on what the advertiser declares, and some depends on which country the campaign targets.

This guide sets out what each platform says on its own pages and how to build the decision into your brief and approval process. The platform rules below were checked on the Google and Meta pages listed in the references on 31 October 2026. These pages change often, so treat the dates as part of the rule.

The short version: what each platform does today

Political, election and social issue ads have separate, stricter disclosure rules on both platforms. They are covered further down. AI-generated content on your own website is a separate question, covered in our note on Google’s guidance for AI-generated website content.

Google Ads: the AI label setting, step by step

Google announced the “How this ad was made” panel on 9 July 2026 on its Ads & Commerce blog. The panel shows whether AI was used to create or edit an ad and, according to the post, is accessible globally on Search, YouTube and Discover. The working detail is in Google Ads Help, under “Use AI content label settings and disclosures”. The help page states:

• You can designate assets as AI-generated or AI-edited, either with the AI label setting in Google Ads or by adding your own label to the creative.
• Google may label assets itself where the law requires it, when it receives signals from other platforms, or when its fully automated creative features generate the asset. A label Google applies cannot be overwritten by the advertiser.
• Assets generated with Google Ads tools carry non-visible SynthID watermarks and C2PA provenance markup.
• Visible overlays appear on the ad itself for campaigns targeting the European Union, India or New York when assets are designated as AI.
• The setting is not a compliance guarantee. Google says using it does not ensure you meet any specific regulation and advises getting legal guidance.

Where you find it in the account

1. Existing assets: the Asset library and asset reports have an “AI label” status column you can filter by. Start with an audit of what is already running.
2. New campaigns: a “Review assets” prompt appears when ads may need a label.
3. Editing: label or unlabel an asset in Asset studio, then Asset library.
4. Assets generated inside Google Ads: a pop-up applies the label at the point of generation.
5. Video uploaded to YouTube through Google Ads: you are asked whether AI was used to generate or edit the video, to meet YouTube’s disclosure policy for organic videos.

If you add your own label to the image

Google says a self-applied AI label does not breach its text overlay or watermark policies. It also gives two cautions that matter to designers. Keep the label away from corners and edges, because Google cannot guarantee it will not be trimmed when the ad renders. And opt out of image enhancements that can crop it: the help page names Asset Enhancements in Display and Performance Max, and Adaptive Layouts in Demand Gen. Those features help ads fit more placements, so decide per campaign.

Meta: “AI info” and the new “About this ad” menu

Meta’s newsroom post on generative AI transparency in its ads products was first published on 3 February 2025 and updated on 1 June 2026. The update says Meta is rolling out “About this ad”, a single transparency destination in the three-dot menu on every ad, and that the “AI info” label sits there. Two triggers are described:

• Ads created or significantly edited with Meta’s own generative AI creative tools.
• Ads created or edited with third-party AI tools, which Meta says it will begin detecting automatically through “industry-standard signals”. The post describes this as a start, not as finished coverage.

Under the original 2025 approach, still described in the same post, no label was applied when Meta’s tools made no significant edit and no photorealistic person was included. When an AI-generated photorealistic person did appear, the label sat next to “Sponsored” rather than behind the menu. Meta notes that the experience may vary in some regions because of legal requirements.

For context, Meta’s wider policy on AI-generated content renamed the “Made with AI” label to “AI info” on 1 July 2024, after the original label appeared on content with only minor edits such as retouching. For planning, assume a retouched product shot and a fully synthetic spokesperson are not treated the same on Meta.

The stricter rules: political, election and social issue ads

For some categories, disclosure is a requirement and leaving it out is a policy breach.

• Google election ads. Google’s political content policy requires advertisers to tick “Altered or synthetic content” in campaign settings when an election ad has images, video or audio that “inauthentically depict real or realistic-looking people or events”. Edits that do not change the ad’s claims are exempt, such as resizing, cropping, colour or brightness correction and background edits that do not create a realistic depiction of an event. In some formats Google generates the disclosure from the checkbox. In all other formats the advertiser must add its own clear, conspicuous disclosure. Google says it will warn at least seven days before any suspension.
• Meta social issue, electoral and political ads. Since early 2024 (Meta’s November 2023 post said “starting in the new year”), Meta has required advertisers to disclose, in certain cases, when such an ad contains photorealistic images or video, or realistic-sounding audio, that was digitally created or altered to depict a real person, a realistic-looking person who does not exist, or a realistic-looking event that did not happen. The rule was announced in November 2023 and is confirmed in Meta’s 2024 labelling post.
• YouTube organic uploads. If your ad video also lives on your channel, YouTube’s disclosure policy asks creators to disclose realistic content that is AI-generated or meaningfully altered. It exempts clearly unrealistic or animated content and minor edits such as colour filters, background blur, upscaling and AI-written scripts. YouTube may apply a label itself, for example when the file carries C2PA metadata, and in that case the creator cannot remove it.

Most B2B advertisers in the Gulf and Egypt will never run election ads. Social issue categories can still be wider than expected, so check how each platform classifies your topic before you launch an awareness campaign on a public-interest theme.

Why the targeting country changes the outcome

Google’s on-ad overlay depends on targeting: the help page names the EU, India and New York. That sits alongside new law. The European Commission’s AI Act page says deepfakes should be “clearly and visibly labelled” and that the Act’s transparency rules come into effect in August 2026, supported by guidelines and a voluntary code of practice on marking and labelling AI-generated content. The Commission has already moved other AI Act deadlines through a 2026 omnibus, so check the current text and guidance before relying on any date. Google does not name the laws behind each region, and its July announcement refers only to “local requirements”.

As of 31 October 2026, Google’s help page does not list Saudi Arabia, the UAE, Qatar or Egypt among the overlay regions, and none of the Google, YouTube or Meta pages listed in our references describes an AI ad rule specific to the Gulf or Egypt. That is all we checked: we did not review national media or advertising regulations, so it does not mean there are no local rules. Two practical consequences follow:

• A campaign that targets Riyadh and Dubai and also a diaspora audience in Europe can show a visible AI overlay in one market and not in the other. Approve the creative with the label in place, not just the clean version.
• Check the current local rules for each target market, or ask your legal adviser, before running realistic AI people or scenes. This guide is not legal advice.

Building AI disclosure into creative and approvals

The usual mistake: a designer extends a background with generative fill, the file goes to the media buyer, and nobody records that AI touched it. A few fields in your existing process fix that. If you are setting up that process from scratch, our guide to an AI-assisted content workflow with real approval gates covers the wider design.

1. Log AI use per asset, at the brief stage

Add one line to every creative brief and file name record: no AI / AI-edited / AI-generated, plus the tool used. “AI-edited” should cover generative fill, synthetic voice and face or body changes. Routine retouching and colour correction can be logged separately, because both platforms treat minor edits differently from synthetic content. The same brief is a natural place for the brand rules your AI tools follow.

2. Decide the label before design, not after

• Will the asset use Google’s or Meta’s own AI tools? Expect that the platform may label it.
• Was it made with a third-party tool? On Google, plan to use the AI label setting. On Meta, assume it may be detected.
• Does it show a realistic person who does not exist, or a real person doing something they did not do? Treat it as high-sensitivity. Check the political and social issue rules, and get the person’s consent where a real individual is shown.
• Which countries does the campaign target? If any are overlay regions, design and approve with the overlay in view.

3. Design for the label

If you add your own on-image label, keep it inside a safe area away from edges and test the main aspect ratios.

4. Keep the provenance data

Google embeds SynthID and C2PA in assets made with its tools, and YouTube reads C2PA metadata. Keep the original export from the generating tool in the asset archive, and check whether your editing and compression steps keep or remove that metadata, so your record matches what the platform may detect.

5. Put the disclosure decision on the approval sheet

The person signing off the ad should see the AI status and the label decision next to the copy and visuals. At RX, clients approve content before it goes out and ad accounts stay in the client’s name, so the label decision is part of the same approval rather than a setting the agency changes quietly. If you run paid campaigns on Google and Meta, ask whoever manages them to show you the “AI label” column in your Google Ads Asset library.

6. Re-audit live campaigns

Assets created before the July 2026 rollout may have no label status. Filter the Asset library by the AI label column, review evergreen Performance Max and Demand Gen assets first, and record what you decided.

Does an AI label hurt performance?

We found no Google or Meta source with data on whether an AI disclosure changes click-through or conversion. Google’s July announcement does repeat that deceptive ads are prohibited whether AI made them or not. For a B2B brand, the bigger risk is credibility: a buyer who spots a synthetic “engineer” or an invented site photo starts to doubt the real projects too. A sensible rule is to use AI freely for backgrounds, layouts, variations and motion, and to use real photography and real people for anything that is meant to prove capability. That is also where filmed product and project content earns its cost.