AI ads on Meta: the policy watchlist
Meta doesn’t ban AI-made ads, but several of its rules bite harder when the footage is generated. For each risk: what Meta’s standard says as of November 2026, why it hurts, and a safer way to get the same effect. Meta can review an ad again after launch and reject it at any time (Advertising Standards). This isn’t legal advice.
Fake product demos
- Meta’s standard. Ads must follow the prohibited commercial practices standard, which covers products that differ significantly from what was advertised. Meta weighs buyer feedback and can reduce delivery, restrict ad accounts or disable them.
- Risk. Buyers who expected what the demo showed return the product or dispute the charge. The YouTube channel Bluesense Digital calls an AI demo of a product doing what it can’t close to fraud (video).
- Instead. Use AI for the problem, a metaphor or a shot no camera could get, then show the product doing only what it does (video).
AI people presented as customers
- Meta’s standard. Meta doesn’t single out AI testimonials, but its inauthentic behavior standard bars misleading people about who you are or about the origin or popularity of your content. In the US, the FTC’s 2024 rule bans fake testimonials, including AI-generated ones from people who don’t exist.
- Risk. A generated face saying “I’ve used it for three months” is a made-up testimonial. Bluesense Digital also finds AI-made UGC puts off the viewers who notice it (video).
- Instead. A narrator, or characters in an obvious story; real quotes only from real customers who agreed.
Before/after and health claims
- Meta’s standard. The health and wellness standard limits weight-loss, dietary, health and cosmetic ads to people 18 and over. It bans negative statements about someone’s appearance, close-ups of pinched body fat, claims to cure a listed set of diseases such as diabetes or cancer, and sensational promises of results by a date without qualification. Some before/after images are allowed within those limits.
- Risk. A generated “after” shows a result nobody achieved. Shrinking it into a corner to slip past review doesn’t make it compliant.
- Instead. A situational before/after, such as a cluttered drawer then a tidy one; real results with the time they took; no disease claims.
Personal attributes in hooks
- Meta’s standard. Ads can’t assert or imply a viewer’s personal attributes, such as age, health, religion, race, sexual orientation or financial hardship. Meta’s examples reject asking whether you have diabetes and accept announcing a new diabetes treatment.
- Risk. Call-out hooks such as “Over 50 with knee pain?” are a common script pattern, and AI script tools write them readily.
- Instead. Name the situation or the benefit, not the viewer: “Knee-friendly shoes for long runs”.
Impersonating Pages, brands or people
- Meta’s standard. Prohibited commercial practices cover falsely speaking for an established company and deceptive use of a public figure’s name or image. The third-party infringement standard bars ads that infringe copyright or trademarks, music included.
- Risk. Brand-owned “review” or “doctor” Pages posing as independent, a generated celebrity look-alike, a competitor’s packaging. Cloned Pages to get past ad limits backfire too: Meta says separate Pages compete in the same auctions (Meta).
- Instead. Run from your own Page, use partnership ads with real creators who agreed, and license what you didn’t make.
Pricing, scarcity and statistics
- Meta’s standard. The prohibited commercial practices standard also lists deceptive or unreasonable pricing and subscriptions without clear terms; Meta acts on persistent negative buyer feedback.
- Risk. A “last day” offer that runs for weeks, or an invented “96% saw results”: generated ads make both effortless.
- Instead. Real deadlines, and numbers with a source you can show.
AI labels and disclosure
- Meta adds an AI info label to ads made or significantly edited with its own generative AI tools; with an AI-generated photorealistic person, it can sit at the top of the ad (Meta).
- In a 1 June 2026 update, Meta said it is beginning to detect ads made with third-party AI tools through industry-standard signals and to show AI info in About this ad (Meta). Its help page may still describe the older rule while this rolls out.
- Ads about social issues, elections or politics must say themselves when images, video or audio were digitally created or altered, AI tools included.
- Local law can ask for more: check the rules where you sell.
What Dokiri does, and what stays with you
- Every Dokiri file carries an invisible “Made with AI” marking in its metadata (IPTC’s “trained algorithmic media” source type), an industry-standard signal; Meta doesn’t list which signals it reads. Re-exporting from an editor drops it. See your video.
- The Before / after format excludes body, skin, weight and health transformations and shows only what your product description supports (ad formats).
- Scripts claim nothing your description doesn’t say, and you read them before paying.
- Characters are AI-drawn: never caption or present them as customers, experts or reviewers.
- Claims, targeting and disclosure in Ads Manager stay your responsibility.