Can you use AI-generated images in Facebook ads? Meta’s rules and labels

Can you use AI-generated images in Facebook ads?

Yes, and Meta says so on its own help pages, where the question is not whether AI images are allowed but how they are labelled. The Business Help Center page on AI info opens: “we will automatically label ad content that has been created or edited using some of Meta’s generative AI tools or with third-party generative AI tools offered by other products like Photoshop, Dall-E, or others.”^[1] Labelling is Meta’s job. The page does not ask a shop or a software company to declare anything.

Three things are true at once. AI images are allowed wherever ads are allowed. Meta labels them when it can detect them, and you cannot have a detected label removed.^[2] And the Advertising Standards still decide whether the ad runs: “Automated detection does not change ad eligibility. Your ads must still comply with all applicable ad policies.”^[2] Meta sells the practice itself: Advantage+ creative generates backgrounds, expands images and writes text variations under generative AI terms that took effect on May 6, 2024.^[5]

The self-disclosure duty exists, but it is narrow: ads about social issues, elections or politics, in certain cases.^[4] The rest of this post covers who labels what, where the label shows, and what an AI image still cannot contain.

What does Meta label itself, and when?

Meta labels on two triggers: its own tools, and detected metadata. For its own tools, the February 2025 newsroom post is precise: if the features “do not result in significant edits to the image or video and do not include a photorealistic human, then we will not apply any AI labels.” A significant edit gets a label behind the three-dot menu or next to the Sponsored label, and an “AI-generated photorealistic human” puts it “next to the Sponsored label (not behind the three-dot menu).”^[3] Sponsored has since been renamed Ad.^[1]

For third-party tools, Meta reads the file: “Meta uses industry-standard detection methods, such as C2PA to identify when ad content has been created or edited using third-party generative AI tools.”^[1] It has read C2PA and IPTC markers since 2024, naming Google, OpenAI, Microsoft, Adobe, Midjourney and Shutterstock as tool makers.^[10] The June 1, 2026 update put the resulting label inside the “About this ad” screen.^[3][7] A file with stripped metadata carries no signal; Meta offers an optional self-disclosure for third-party AI media that puts the label at the top of the ad.^[6]

When do you have to disclose AI yourself?

Only when the ad is about social issues, elections or politics, and only for realistic media. Meta’s help page sets the trigger: the ad “contains a photorealistic image, video, or realistic sounding audio that was created or edited using AI tools to” depict a real person saying or doing something they did not, depict “a realistic-looking person that does not exist or a realistic-looking event that did not happen”, or depict a realistic event that is not a true recording.^[2] Edits that are “inconsequential or immaterial”, such as resizing, cropping, color correction and sharpening, need no disclosure.^[2]

“If you do not disclose these specified scenarios, your ad may be removed. Repeated failure to disclose may result in penalties on the account.”

The mechanics are two clicks in Ads Manager: select the Social issues, elections or politics special ad category, then in Ad creative select “Includes media created or edited with AI”. Meta shows “AI info” near the “Paid for by” disclaimer and adds “Media in this ad created or edited with AI” to the Ad Library entry.^[2] Meta’s 2026 midterms post restates the scope: advertisers “are still required to self-disclose when they use AI to create or edit ads about social issues, elections, or politics in certain cases.”^[4] A furniture shop or a dentist is not in that category, so the duty does not reach them.^[7]

Is AI disclosure mandatory for all ads from March 2026?

No. That claim is common on agency blogs and absent from Meta’s pages. AuditSocials writes: “Effective March 2026, Meta requires all advertisers to disclose AI-generated or AI-modified content in their ad creatives”, describes a strike system in which a third violation within 90 days brings “ad account restriction or suspension”, and adds that “there is currently no automated detection for AI-generated content”.^[11] 1ClickReport says any AI-generated or AI-composited creative “must now carry an ‘AI-generated’ label. No exceptions.”^[12]

None of that matches Meta. There is no “AI-generated” label; Meta’s label is “AI info”.^[1] There is no all-advertiser disclosure rule; the duty is limited to social issue, electoral and political ads.^[2][4] Automated detection does exist, through C2PA metadata, from June 1, 2026.^[1][7] The three-strikes-in-90-days schedule appears on no Meta page we could find; Meta’s wording, for political ads, is that repeated failure to disclose “may result in penalties on the account”.^[2] 1ClickReport’s page now carries a September 2026 correction note restating Meta’s narrower rule at the top, while the body still says “No exceptions”.^[12]

What do the Advertising Standards still forbid in an AI image?

Everything they forbid in a photograph. Meta reviews each ad against the Advertising Standards whatever made the picture, and the labelling pages repeat that AI detection “does not change ad eligibility”.^[2] Three rules catch AI images most often.

First, misleading practices. Ads “must not promote products, services, schemes or offers using identified deceptive or misleading practices”.^[8] An AI-generated person presented as a customer, with a name and a quote, is a claim about who uses your product that is not true. Meta’s political-ad rule names “a realistic-looking person that does not exist” as material AI content, and its own tools put the visible label on any photorealistic AI human they produce.^[2][3] Second, health and appearance. For weight-loss products Meta refuses images that “depict a close up on a specific body area with pinched fat” and side-by-side comparisons after the product, and it refuses cure claims for a listed set of conditions and “promises of specific outcomes within a set timeframe”.^[9] A model that draws a flatter stomach on request makes those images easy to produce and just as easy to reject. Third, the claim in the picture: a generated image of a product doing something it cannot do is a misleading claim, and Meta reviews text and images together.^[8]

What does this mean for static ads made with AI?

It means the AI should draw the setting, not the evidence. A generated background, texture or scene is allowed, is labelled at most behind a menu, and makes no claim.^[1][3] The parts that carry a claim should be real: your product photo, your logo, your headline and any proof such as a review or a price. Keep the metadata on the file; a C2PA label under About this ad does not change eligibility.^[2]

Two habits keep an AI-made static ad inside the Standards. Do not generate a person and present them as a customer; a photorealistic AI human is the case Meta singles out for a visible label, and as a testimonial it is a misleading claim.^[3][8] And do not let the model draw the product or the result. A generated bottle has the wrong label, and a generated body is a before-and-after waiting to be refused.^[9]

Showfolk places your real logo, product photos and headline as layers on the AI-made background, so the ad is yours and nothing in it pretends to be a customer. The AI ad generator page shows how an ad is built, and the FAQ covers what the models are asked to draw. The post on the retired 20% text rule covers what the headline on the image may say.

A checklist before you upload an AI image

1. 1

   Decide what the AI drew

   Background, scene and texture are fine. If the AI drew a person, a body, the product or a result, treat that part as a claim and replace it with a real photo or remove it.^[8][9]
2. 2

   Check the ad category

   If the ad is about social issues, elections or politics and uses photorealistic AI media, select the special ad category and tick Includes media created or edited with AI. Everyone else skips this step.^[2]
3. 3

   Expect a label, not a rejection

   If the file carries C2PA or IPTC metadata, or you used Meta’s generative features, an AI info label appears under About this ad. Ads Manager warns you in the preview.^[1][5]
4. 4

   Read the picture as copy

   Any claim shown in the image, including a result, a figure or an endorsement, is reviewed as if it were written in the primary text. Keep it true and sourced.^[8]
5. 5

   Keep the proof real

   Put your actual product photo, logo and headline on the generated background as layers, so what the ad asserts is something you can stand behind.

Questions people ask

Does Facebook allow AI-generated images in ads?

Yes. Meta allows AI-generated images in Facebook and Instagram ads and labels them itself when it detects AI, through its own generative tools or through C2PA or IPTC metadata in the file. The Advertising Standards apply to the picture as they would to a photo.

Do you have to disclose AI in Facebook ads?

Only for ads about social issues, elections or politics, and only when a photorealistic image, video or realistic-sounding audio was created or altered with AI. Ordinary business ads have no self-disclosure requirement; Meta applies any AI info label on its own.

Where does the AI info label appear on a Facebook ad?

In most cases under About this ad in the three-dot menu. Two cases put it on the ad itself: a photorealistic AI person made with Meta’s tools gets the label next to the Ad label, and a political or social-issue ad with self-disclosed AI media shows AI info near the Paid for by disclaimer.

Does the AI info label hurt ad delivery or review?

Meta says automated detection does not change ad eligibility and that ads must still comply with the Advertising Standards as before. Meta publishes no delivery penalty tied to the label; it is information for the viewer.

Is there a three-strikes rule for undisclosed AI in Facebook ads?

Not on any Meta page we could find. The three-strikes-in-90-days schedule appears on agency blogs. Meta’s own wording, for political and social-issue ads only, is that an undisclosed ad may be removed and repeated failure to disclose may result in penalties on the account.

Can I use an AI-generated person in a Facebook ad?

You can, but not as a customer. An AI person presented with a name and a testimonial is a misleading claim under the Advertising Standards, and a photorealistic AI human made with Meta’s tools gets a visible AI info label. Keep people, products and results real.

Sources

Every figure above points at one of these. Facts were last checked on October 7, 2026.

1. [1]Meta Business Help Center: About AI info on ads created or edited with generative AI tools. Automatic labelling for Meta’s tools and for third-party tools detected through C2PA; the label’s place under About this ad; the Sponsored label renamed Ad
2. [2]Meta Business Help Center: About media created or edited with AI. The self-disclosure duty for social issue, electoral and political ads, the three trigger scenarios, the removal and penalty wording, the immaterial-edit exceptions, and that detection does not change ad eligibility
3. [3]Meta Newsroom, February 3, 2025, updated June 1, 2026: Expanding GenAI transparency for Meta’s ads products. When Meta’s own tools do and do not label, the photorealistic-human rule, and the About this ad screen
4. [4]Meta Newsroom, February 2026: How Meta is preparing for the 2026 US midterm elections. “Advertisers are still required to self-disclose when they use AI to create or edit ads about social issues, elections, or politics in certain cases.”
5. [5]Meta Business Help Center: About Advantage+ creative. Background generation, image generation, visual touch-ups and text generation; the May 6, 2024 generative AI terms; the AI info note
6. [6]Meta Business Help Center: About AI info on ads created or edited with third-party generative AI tools. The optional self-disclosure for third-party AI media and where that label shows
7. [7]Meta Advertising Standards: Ads about social issues, elections or politics. The policy text of the disclosure rule and the June 1, 2026 start of automated detection with no advertiser action required
8. [8]Meta Transparency Center: Advertising Standards, Unacceptable business practices. “Ads must not promote products, services, schemes or offers using identified deceptive or misleading practices”
9. [9]Meta Business Help Center: About Meta’s Health and Wellness advertising policy. Pinched-fat close-ups, weight-loss comparisons, cure claims and timed-outcome promises
10. [10]Meta Newsroom, February 6, 2024: Labeling AI-generated images on Facebook, Instagram and Threads. The C2PA and IPTC standards Meta reads, and the companies whose tools it expected to label
11. [11]AuditSocials, March 26, 2026: Meta AI privacy policy 2026. An example of the “mandatory disclosure for all ads from March 2026” and three-strikes claim
12. [12]1ClickReport: Meta 47 ad policy changes 2026 compliance guide. “No exceptions” claim in the body, with a September 2026 correction note at the top restating Meta’s narrower rule