Meta auto-detects AI-made ads: plan for the AI info label
Meta says its ad system detects ads made or edited with third-party AI tools through industry-standard signals and may show an AI info label. What to prepare.
Meta's page, updated June 1, 2026, says "About this ad" now includes AI info labels, and that Meta automatically detects ads created or edited with third-party AI tools through industry-standard signals. The label sits in the three-dot menu, or next to "Sponsored" when a photorealistic AI human is present. If you make ad creative with an AI generator, assume the label can appear, and decide in advance how the ad reads with it.
What the page says
The page does not list the signals in the text this post relies on, and it does not publish a threshold for when detection triggers. Do not design around a guess at either.
| Item | Detail |
|---|---|
| Where the label shows | Three-dot menu in "About this ad" |
| Stronger placement | Next to "Sponsored" when photorealistic AI humans are present |
| Detection | Automatic, for ads created or edited with third-party AI tools, via industry-standard signals |
| Page date | Updated June 1, 2026 |
Creative choices that change the label
The placement rule hinges on photorealistic AI humans. A product-only shot, a stylised illustration or a motion-graphics layout is a different case from a realistic generated person speaking to camera. That is a creative decision, and neither choice is wrong; the point is to make it knowingly.
If the ad leans on a generated person, test how it performs with the label showing. If it leans on the product, the label matters less. The Sume video API and Image API are neutral on this; you pick the subject in the prompt.
A pre-upload routine
Detection is something Meta does after upload. Your disclosure is something you do before it.
- Mark each asset as product-only, stylised, or photorealistic person.
- For photorealistic person assets, plan for the label beside "Sponsored".
- Keep the job ID, prompt and date for each asset, so you can answer a reviewer.
- Add the disclosure inside the ad flow where Meta offers it, rather than waiting for auto-detection.
What to monitor
Check the ad's details view in the live account for the label after it goes through review. If an ad you consider product-only is labelled, record that and compare against similar assets. Over a few campaigns you will see what your own content triggers, which is more useful than a general rule.
asset,type,label_expected,label_seen_live
mug-hero-9x16.mp4,product-only,no,
founder-clip-9x16.mp4,photorealistic person,yes,Reading the result
A label on an ad is information, not a penalty. The Meta page frames it as transparency in "About this ad". Measure results as you would for any ad: compare cost per result between assets with and without a visible label, over the same period.
Document what you observe. If an asset's label appears only on one placement, note it. If it never appears, note that too. After a few campaigns you will know how your own content behaves.
Treat the page date as a prompt to re-check. The text read for this post was updated on June 1, 2026, and Meta revises these pages.
Sources
Related posts
More in Use cases
- Meta signs the EU AI content transparency code (July 28, 2026)
Meta confirmed on July 28, 2026 that it signed the EU Code of Practice on AI content transparency. What the date means for teams making ad video.
- Midjourney V8.2 edit stills, animated on Sume
Finish a still in Midjourney V8.2, host it at a public HTTPS URL, and use it as the first frame of a Sume video request. Midjourney itself has no video id here.
- H3 mixing camera, character and audio from three inputs
MiniMax H3 can mix camera movement from one video, a character from an image and audio from another. Sume's minimax-h3 takes image, video and audio references.
- Mixed-language explainer: two TTS jobs joined into one track
A voice that handles 23 or 90 languages still needs one language per request. Render each language as its own job, then join the takes with timeline audio.
Written by Sume