AI-made images, video and copy are now ordinary ad creative, and Meta has built a labelling system around them. Some ads get an AI info label automatically, some advertisers must disclose AI use themselves, and some AI content breaks policy no matter how it is labelled. This guide covers each case. Every rule comes from Meta's newsroom, Transparency Center or Business Help Center; the closing checklist is our own advice.
How Meta's AI label changed: Made with AI to AI info
Meta's label wording has moved several times, and its newsroom posts give the sequence. In February 2024 Meta said it was building tools to label AI images from other companies, and that photorealistic images made with its own Meta AI feature were labelled Imagined with AI. On April 5, 2024 it announced a Made with AI label for organic content, with labelling planned to start in May 2024.
On July 1, 2024 Meta renamed the label to AI info, saying the old wording had appeared on content with only minor AI retouching. People can click AI info for more information. On September 12, 2024 Meta said content it detects was only modified or edited by AI tools would carry the label in the post's menu, while content it detects as AI-generated keeps the label on the post.
If digitally created or altered content poses a particularly high risk of materially deceiving the public on an important matter, Meta says it may add a more prominent label. Labelled content stays up unless it breaks another policy, such as those on voter interference, bullying and harassment, or violence and incitement.
How Meta detects AI content: C2PA, IPTC and invisible watermarks
Meta's detection relies mostly on information that AI tools write into the file. In its February 2024 announcement Meta said it was building tools to identify the AI generated information in the C2PA and IPTC technical standards, so it could label images from Google, OpenAI, Microsoft, Adobe, Midjourney and Shutterstock as those companies added metadata to their tools. Its own Meta AI images carry IPTC metadata and invisible watermarks, plus visible markers on photorealistic images.
The same post named the limits. Meta said it was developing classifiers to detect AI content that lacks invisible markers, and Stable Signature, a watermark built into the image generation process. For video and audio, Meta asks people to use its disclosure and label tool when they post organic photorealistic video or realistic-sounding audio that was digitally created or altered, and says it may apply penalties if they don't.
For ads, Meta's Help Center says it uses industry-standard detection methods such as C2PA to identify ad content created or edited with third-party generative AI tools, and labels the content when it detects that metadata. Its 2026 US midterms post also names C2PA for organic content. Meta's Transparency Center page on AI labels describes the signals only as industry standard AI image indicators.
AI info labels on ads, and where people see them
Meta began labelling ads made or significantly edited with its own generative AI creative tools in 2024. Its February 3, 2025 post said those labels would sit behind the three-dot menu or next to the Sponsored label, and that an ad with an AI-generated photorealistic human made by Meta's tools would carry the label next to Sponsored. Ads from Meta's tools that neither significantly edit the media nor include a photorealistic human would not be labelled.
On June 1, 2026 Meta updated that post: it would start automatically detecting ads created or edited with third-party AI tools and apply an AI info label, and it was rolling out an About this ad screen in the three-dot menu of every ad. The Help Center now says that in most instances the AI info label is found by opening the three-dot menu, selecting About this ad and tapping the label. The Help Center also notes that ads previously marked Sponsored now show an Ad label.
Two exceptions apply. In regions with specific AI transparency rules, where Meta lists the European region, California, New York, India, Taiwan and South Korea, you may be able to self-disclose generative AI use, and the label then shows on the ad next to the Sponsored label. In Taiwan and South Korea, Meta may also auto-label ads where its own tools produced photorealistic humans. Boosted posts follow the organic approach, so an organic AI label may stay on a boosted post. Meta says you'll see a message during ad creation when an AI info label will appear.
Mandatory disclosure for social issue, electoral and political ads
Meta announced in November 2023 that advertisers would have to disclose digitally created or altered media in certain political and social issue ads, and its April 2024 post says the rule applied from January 2024. The current Help Center article requires disclosure when such an ad contains a photorealistic image, video or realistic-sounding audio that was created or edited with AI tools to depict a real person saying or doing something they did not; a realistic-looking person who does not exist or a realistic-looking event that did not happen, or altered footage of a real event; or a realistic event that allegedly occurred but is not a true recording of it. The Transparency Center policy dated October 6, 2025 said digitally created or altered; its July 1, 2026 version refers to third-party generative AI tools.
Edits that are inconsequential or immaterial to the claim, assertion or issue in the ad don't need disclosure. Meta's examples are adjusting image size, cropping, color correction and sharpening. To disclose, select Social issues, elections or politics as the Special Ad Category, then choose Includes media created or edited with AI in the Ad creative section. An AI info label then appears near the Paid for by disclaimer, and the Ad Library shows Media in this ad created or edited with AI.
If you don't disclose when required, Meta says the ad may be removed or rejected, and repeated failures may bring penalties on the account. Since June 1, 2026, Meta also detects third-party AI media in these ads automatically and shows AI Info under About this ad. Meta says that detection works alongside self-disclosure, so you must still disclose; you cannot ask for a detected label to be removed; it doesn't change ad eligibility; and Meta's own generative AI creative tools remain unavailable for these ads. Our special ad categories guide covers the category itself.
Meta's generative AI tools in Ads Manager, and which get labels
Meta's Help Center lists several generative AI features in Ads Manager, covered since May 6, 2024 by its Ad Creative Generative AI Terms. Meta says some verticals, Special Ad Categories and countries may not have immediate access.
- Text generation, formerly text variations: up to five variations of primary text and headline based on your original text, optionally shaped by personas or a brand tone. Meta's labelling articles describe labels for images and video and don't mention a label for generated text.
- Image generation: variations inspired by your image or video ad, which you can edit with a written instruction. It isn't available on Advantage+ app, sales or catalog campaigns. Meta says images created or significantly edited with certain of its features may carry AI info.
- Background generation: Meta's Help Center names it as a feature that triggers an automatic AI info label when used to create or significantly edit an image.
- Add animation: also named as a label trigger when it creates or significantly edits an image or video.
- Expand image: listed among Meta's generative AI features under the same terms. Meta's labelling articles don't say whether expansion alone triggers a label; the stated test is whether the media was created or significantly edited.
Policy risks: deceptive AI content and public-figure likeness
A label doesn't make content allowed. Meta's April 2024 post says ads containing content debunked by fact-checkers are rejected, and labelled organic content is still removed if it breaks other policies.
The highest-risk area is the use of famous faces. Meta's Prohibited Commercial Practices standard, the new name for its Fraud, Scams and Deceptive Practices standard, defines celeb-bait as falsely using the name, likeness or imagery of a public figure together with misleading tactics to lure people into a scam. Meta's September 2026 notice says these rules now cover ads and organic posts alike, and that it will expand enforcement against celeb-bait in both. Our Prohibited Commercial Practices guide covers that change.
Meta also uses facial recognition here. When its systems suspect a scam ad uses a public figure's image, it compares faces in the ad with that person's Facebook and Instagram profile pictures and blocks the ad if it confirms a match and a scam. Meta says it deletes the facial data from that comparison immediately and that the protection covers nearly 500,000 public figures. AI-made versions of real people, logos or characters also raise rights questions, covered in our copyright and trademark rejections guide.
Practical checklist for AI ad creative (practitioner advice)
This list is our own advice, built on the rules above. It is not Meta guidance. Power Ads supplies Meta agency ad accounts at 4% per top-up; we don't make or review creatives, so these checks stay with you.
- Assume AI use will be visible. Meta detects C2PA metadata and adds labels automatically, and you can't request removal of a detected label on a social issue ad. Plan creative that still works with an AI info label attached.
- Preview every AI-generated variation before publishing. Meta's tools produce versions you haven't written or drawn, and a variation that misstates your offer is still your ad.
- For social issue, electoral or political ads, decide on disclosure before launch. If the media is photorealistic and created or edited with AI to show something that didn't happen, tick Includes media created or edited with AI. Cropping and color correction alone don't need it.
- Never put a real public figure's face, voice or name into AI creative without permission. Under the celeb-bait rules and facial recognition checks this is a direct route to rejection and account risk.
- Don't present AI scenes as real testimonials, real events or real results.
- If you target the European region, California, New York, India, Taiwan or South Korea, check whether Ads Manager offers a self-disclosure option.
- Mix AI and non-AI creative. Our Andromeda creative diversification guide explains why similar creatives compete.
Key takeaways
- Meta's label went from Made with AI (April 2024) to AI info (July 2024), and AI-edited organic content moved the label into the post menu from September 2024.
- Meta reads C2PA and IPTC metadata and its own invisible watermarks, and uses C2PA to detect third-party AI tools in ads.
- On most ads the AI info label sits under About this ad in the three-dot menu; in some regions a self-disclosed label sits next to the Sponsored label.
- Social issue, electoral and political ads must disclose photorealistic AI media that depicts things that didn't happen; failing to do so can get the ad removed and repeated failures can bring account penalties.
- A label never cures a policy breach: celeb-bait and debunked content are enforced regardless.
FAQ
Does every AI-assisted ad get an AI info label?
No. Meta labels ads created or significantly edited with some of its generative AI tools, such as background generation, image generation and add animation, and ads where it detects third-party AI metadata such as C2PA. Its February 2025 post said ads from its tools that neither significantly edit the media nor include a photorealistic human would not be labelled.
Where does the AI info label appear on an ad?
In most cases people find it by opening the ad's three-dot menu, selecting About this ad and tapping AI info. In the European region, California, New York, India, Taiwan and South Korea you may be able to self-disclose, and then the label shows next to the Sponsored label. On self-disclosed political ads it appears near the Paid for by disclaimer.
How does Meta know an image was made with AI?
Mainly from information in the file. Meta says it reads the AI generated information in the C2PA and IPTC standards that other companies' tools add, and its own AI images carry IPTC metadata and invisible watermarks. It has also said it is developing classifiers for content without those markers.
Do I have to disclose AI in a political or social issue ad?
Yes, when photorealistic AI-made or AI-edited media shows a real person saying or doing something they didn't, a person or event that isn't real, altered footage of a real event, or a fake recording of an alleged event. Resizing, cropping, color correction or sharpening alone don't need disclosure.
Can I use Meta's generative AI tools for political ads?
No. Meta's Help Center says its first-party generative AI creative tools remain unavailable for ads about social issues, elections or politics. If you use third-party AI tools for such ads, you must still disclose where the rules require it, and Meta may also detect and label the media automatically.
Can an AI-made celebrity endorsement run if it's labelled?
Not as a scam. Meta's Prohibited Commercial Practices standard bans falsely using a public figure's name, likeness or imagery with misleading tactics, and Meta uses facial recognition to block suspected celeb-bait ads. A label does not make a false endorsement acceptable.
Sources
- about.fb.com/news/2024/02/labeling-ai-generated-images-on-facebook-instagram-and-threads
- about.fb.com/news/2024/04/metas-approach-to-labeling-ai-generated-content-and-manipulated-media
- about.fb.com/news/2023/11/how-meta-is-planning-for-elections-in-2024
- about.fb.com/news/2025/02/gen-ai-transparency-metas-ads-products
- about.fb.com/news/2026/02/meta-prepares-for-2026-us-midterms
- about.fb.com/news/2024/10/testing-combat-scams-restore-compromised-accounts
- transparency.meta.com/governance/tracking-impact/labeling-ai-content
- transparency.meta.com/policies/ad-standards/SIEP-advertising/SIEP
- transparency.meta.com/policies/community-standards/prohibited-commercial-practices
- transparency.meta.com/policies/other-policies/our-prohibited-commercial-practices-policy
- facebook.com/business/help/1010479435004531
- facebook.com/business/help/1486382031937045
- facebook.com/business/help/180641596861873
- facebook.com/business/help/1684513971952814
