Generative AI has been pushed hard by every single ad platform, despite most of the features being crap, and the ads it produces looking even more crap.
The features include writing copy, generating product shots, swapping backgrounds and creating video ads from images. Additionally, outside of what can be created within these platforms, ads with synthetic spokespersons and janky voiceovers are popping up everywhere.
Consumers have started to notice, and mostly, they don’t like what they’ve seen. I wrote about why most consumers don’t trust generative AI a couple of weeks ago, and the ad platforms seem to be arriving at the same conclusion – transparency matters.
This month, Google became the latest to roll out formal AI disclosure tools for advertisers, joining Meta and TikTok in building AI transparency directly into the ad experience.
Here’s what’s changed, how the major platforms compare, and what it means for how you should (and shouldn’t) be using gen AI in your own campaigns.
What Google just launched
On 9 July, Google announced it’s expanding AI transparency in ads, adding a new “How this ad was made” panel across Search, YouTube and Discover. It sits inside the ‘Google Ads Transparency Centre’, accessible via the three-dot menu or info icon on any ad, alongside existing panels like “About this advertiser” and “Why this ad”.
The mechanics work like this:
- Ads made with Google’s own generative AI tools get a disclosure added automatically. No action needed from the advertiser.
- Ads made with AI elsewhere (Midjourney, an in-house tool, an agency’s stack) now have a manual control in Google Ads so advertisers can flag that AI was used.
- Depending on local rules, that disclosure might also surface directly on the ad itself, not just in the panel.
However, according to Google’s My Ad Centre help documentation, disclosure isn’t a blanket requirement everywhere. It depends on the region and the type of content, with the EU, India and New York specifically named as places with their own rules on AI-edited or AI-created content.
Google is clear that its underlying policy hasn’t changed: misleading or deceptive ads are already prohibited, whether AI made them or not. This is about giving people the context to judge that for themselves.
Policies on other ad platforms
Meta got here first. Per Meta’s own Help Centre, ad images created or significantly edited with Meta’s generative AI creative features, or created or edited using third-party AI tools, get an “AI info” label applied so people can see how the image was made. Meta has said it uses IPTC metadata and invisible watermarks on its own AI images, and is building the capability to read the same “AI-generated” signals in the C2PA and IPTC standards from other companies’ tools, so it can label content made with tools like Midjourney or Adobe Firefly, too. Where Meta’s rules get firmer is political and social issue advertising: for those ads, advertisers are already required to disclose when a photorealistic image or video, or realistic-sounding audio, was created or edited with third-party generative AI to depict a real person saying or doing something they didn’t, a realistic person or event that doesn’t exist, or an altered version of real footage. Fail to disclose, and the ad gets rejected; repeated failures can bring penalties against the advertiser. From 1 June 2026, Meta also runs automated detection on these ads, so it can apply the label itself even without advertiser input.
TikTok has a comparably firm setup within its own Advertising Policies. Under its “Edited media and AI-generated content (AIGC)” policy, significantly edited or AI-generated ad content is allowed, provided it carries either TikTok’s AIGC label or a clear disclaimer, caption, watermark or sticker of the advertiser’s own. TikTok draws a clear distinction between “significant” and “insignificant” AI modifications. Significant changes such as fully AI-generated images, video or audio, making someone appear to do something they didn’t do, or cloning a person’s voice to make them say something they never said – require disclosure. By contrast, routine edits such as adjusting lighting, brightness or colour, denoising, or even removing or modifying backgrounds do not. Ads Manager also has a dedicated, TikTok-labelled “AI-generated content” disclaimer advertisers can attach to an ad, separate from the standard and clickable disclaimer types.
YouTube, under the Google umbrella but worth a separate mention, has firm requirements for political and election ads with synthetic content, but no blanket disclosure mandate for standard commercial ads. The general principle there is the same as Google’s wider ad policy: don’t mislead, and disclose when realistic AI content could be mistaken for something real. Despite this, because I am an angry perimenopausal woman, I keep being shown ads on YouTube for a pilates app where two not-real women tell me how much weight they lost in days, and there is no disclosure that it’s AI-generated. I’d call that misleading.
The common thread across Google, Meta, and TikTok is C2PA: the Coalition for Content Provenance and Authenticity standard that embeds machine-readable metadata into files from tools like Adobe Firefly, DALL-E and Midjourney. As more generative tools adopt it by default, platforms are moving from relying on advertisers to self-disclose towards detecting AI content automatically, whether you’ve flagged it or not.
Why is this happening now?
Two forces are pushing this at the same time. One is regulation. The EU AI Act’s transparency provisions require AI-generated content shown to EU users to carry machine-readable disclosure, with wider enforcement obligations still to come during 2026.
The other is trust, and this is the bit that should matter as much to advertisers as the compliance angle. There is a significant gap between how positively marketers assume people feel about AI-generated ads and how people actually feel about them. Pretending something is real only to have an ‘AI-generated’ label slapped on it is not a great look.
What advertisers should actually do about it
Disclosure requirements aside, the more useful question is where gen AI genuinely helps your ads and where it quietly undermines them.
Acceptable use of gen AI in ads:
- Product placement and lifestyle shots. Dropping a real product photo into a generated scene (on a kitchen counter, in a park, styled for a season you haven’t shot yet) is a legitimate, low-risk use. It’s still your real product; you’re just changing the context around it. TikTok’s own ad policy explicitly treats background removal and modification as an “insignificant” AI edit that doesn’t require disclosure, for what it’s worth.
- Iteration and ideation. Generating quick concept variations, headline options or storyboard frames to pressure-test creative direction before you commit budget to production.
- Backgrounds, scenes and non-claim visuals. Anything decorative that doesn’t misrepresent the product itself.
- Copy drafting and structuring, provided a human reviews it for accuracy and tone before it goes anywhere near a live campaign.
- Personalisation at scale, such as generating creative variants for different audience segments, as long as every variant is held to the same accuracy standard as the original.
Where gen AI shouldn’t go near your ads:
- Don’t make the product look better than it is. If AI-generated imagery shows features, finishes, sizes or performance that the real product doesn’t have, that’s not a creative choice, it’s a misrepresentation, and it’s exactly the kind of thing these new disclosure tools exist to surface.
- Don’t fabricate results or outcomes. Synthetic before-and-afters, invented performance stats, or “typical results” that no real customer has achieved.
- Don’t create synthetic testimonials or spokespeople without disclosing them as such. An AI-generated person saying a product changed their life is unlikely to resonate with customers, and under these new policies, the creative might be rejected.
- Don’t use AI to imply endorsement. Synthetic voices or likenesses that resemble real, identifiable people, celebrity or otherwise, without consent.
- Don’t treat the label as a box-ticking exercise. Disclosing that AI was used doesn’t excuse a misleading claim underneath it. The policies are explicit that existing rules against deceptive advertising still apply on top of the disclosure requirement, not instead of it.
While I’ve still seen my fair share of clearly misleading, AI-generated ads across all of these platforms, I’d assume every major platform will require some form of AI disclosure within the next year or two, if it doesn’t already, and that detection will increasingly happen automatically.