TerradsTerrads
← All posts

No, you don't have to disclose AI in your Meta ads

Every guide on this says advertisers must now declare AI-generated creative. Meta's own documentation says otherwise for commercial ads — there is one real requirement, and it probably isn't yours.

ChrisUpdated September 15, 20266 min read

Ordinary commercial Meta ads need no AI disclosure. Meta's mandatory rule covers only ads about social issues, elections or politics. For everything else there is nothing to declare: since 1 June 2026 Meta detects media made or significantly edited with third-party AI tools and attaches an "AI info" label itself, usually on the ad's "About this ad" screen.

Search for Meta's AI disclosure rules and you'll find a wall of confident 2026 posts: new policy, mandatory declaration, tick the box or get disapproved, here's the compliance checklist. Several quote a specific-sounding statistic about what share of ad rejections it now causes.

Almost none of them link to Meta. So we read Meta's own pages instead, and the requirement they're describing does not exist for ordinary commercial advertising.

Here is what's actually true.

The one real disclosure requirement

There is a mandatory self-disclosure rule, and it covers ads about social issues, elections or politics — nothing else. Meta's help centre puts it plainly: "For ads about social issues, elections or politics, advertisers are already required to disclose if the image, video or audio are created or edited with AI."

Note "already." This isn't new. It's the political-advertising regime, and it's narrower than it sounds even within that category — the policy is aimed at photorealistic synthetic media: a real person shown saying something they didn't say, a realistic person who doesn't exist, an event that didn't happen, or real footage altered.

If you sell skincare, software or sofas, this rule is not about you.

Can Meta tell if an ad image was made with AI?

You don't declare anything. Meta detects and labels it for you.

Since 1 June 2026, Meta has run automated detection over ad media to identify creative made or edited with third-party generative AI tools, using industry-standard signals. When it fires, an "AI info" label is attached. No advertiser action is involved — there's no box to tick, and nothing to remember.

That's the entire mechanism for commercial ads. The "compliance checklists" circulating are checklists for a process that doesn't exist.

Where the label actually appears

This is the part worth knowing, because it's where most of the anxiety lives.

The "AI info" label sits on the "About this ad" screen — behind the three-dot menu in the ad's top-right corner. Meta says it can also appear next to the Sponsored label at the top of the ad in some cases.

Read that again: for most labelled ads, seeing the label requires the viewer to open a menu they have no reason to open. This is not a badge stamped across your creative. It is a disclosure in the ad's metadata panel, sitting alongside the "why am I seeing this ad" information that has been there for years.

What triggers it, and what doesn't

Meta draws the line at significant edits, and says so explicitly. The examples it gives of triggering work are things like image generation and background generation. The examples it gives of non-triggering work are minor: resizing and colour correction.

So the boundary is roughly "did AI make or substantially change the picture" versus "did AI tidy an existing photograph." Cropping a product shot to 4:5 and warming the white balance is not what this is aimed at.

One asymmetry worth noting: Meta treats its own in-house generative features slightly differently from third-party tools. Its own tools don't attract a label when the edits aren't significant. Third-party tools are detected via industry signals. In practice both roads lead to the same "AI info" label — but it's a reminder that the detection is signal-based, not a declaration you control.

Does the label hurt performance?

The honest answer: Meta publishes nothing saying it does.

Its documentation on how ad images are identified and labelled makes no statement about the label affecting delivery, reach, or approval. There's no published downranking of labelled ads, no stated review penalty, and no disapproval pathway tied to it.

That is not the same as proof it's neutral. Nobody outside Meta can measure whether a labelled ad performs differently, because you can't run the counterfactual — you don't get to serve the same creative unlabelled to the same audience. Anyone presenting a number for "the cost of the AI label" is inferring it.

What I'd take from the absence: this is framed and built as a transparency feature, not an enforcement one. It's a label, not a penalty. Treat claims to the contrary as unsourced until Meta says otherwise.

Where the panic came from

Two real things got compressed into one imaginary thing.

Meta does have a genuine, mandatory AI-disclosure rule — for political ads. Meta did ship a genuine change on 1 June 2026 — automatic detection of third-party AI in ad media. Blend those and you get "Meta now requires advertisers to disclose AI," which is a sentence that sounds right and is wrong. The mandatory part isn't yours, and the part that is yours isn't mandatory because it isn't yours to do.

It's the same failure mode as the Meta ad size everyone copies wrong: a plausible claim gets written down once, and then it gets cited rather than checked. When a rule appears in fifteen articles and none of them link to the source, that's fifteen people reading each other.

The practical version

  1. Selling a product? There is nothing to disclose. No box, no declaration, no checklist.
  2. Running political or social-issue ads? There is, and it predates all of this. Read the policy properly — the synthetic-media conditions are specific.
  3. Expect the label if AI made or substantially changed the image. Resizing and colour correction don't count.
  4. The label lives behind the three-dot menu, not on your creative.
  5. Don't budget for a performance penalty nobody has measured, and be sceptical of anyone quoting one.
  6. Check the source. Meta's Transparency Center and help centre are public and take a few minutes to read.

The genuinely interesting shift here isn't regulatory, it's that AI-made creative is now common enough to be worth labelling at platform scale rather than policing case by case. That's a real change. It just isn't a paperwork change, and it isn't one you have to act on.

Sources

Sourced from Meta's own help centre and newsroom, checked 10 September 2026; links re-verified 15 September 2026. Meta revises these pages without notice — read the current version before relying on any of it, particularly the political-ads policy, which is the one with actual obligations attached.

Related: how Meta detects AI in the first place — metadata, and what the label costs you — plus the sixteen layouts almost every performance ad uses and why AI image models still can't spell.

Terrads makes ad creative with AI, so the label above can apply to ads built with it — same as every other tool in this category. That's the whole obligation.