On this page
Last checked: 29 October 2026. This is a plain-English summary for marketers, not legal advice. In 2027, an AI-made ad needs a disclosure when it shows a realistic person or event that did not exist, in most of the markets D2C brands sell in. India has required a visible label on realistic AI-generated content since 20 February 2026, New York requires ads to disclose AI "synthetic performers", the EU AI Act requires deepfakes to be disclosed from 2 August 2026, and California makes large platforms display provenance data from 1 January 2027. Separately, the FTC treats an AI "customer" giving a testimonial as a fake review. Ads that use your real product in a generated scene, with no realistic fake person, are mostly outside these rules.
The rules by market
| Market | What the rule covers | What you have to do | In force |
|---|---|---|---|
| India | "Synthetically generated information": audio or video created or altered by computer that looks real | Platforms must label it prominently and embed provenance metadata; large platforms ask uploaders to declare it | 20 February 2026 |
| EU | AI Act Article 50: deepfakes, meaning AI-made people, places or events that look real | The deployer (the advertiser) must disclose; AI providers must mark outputs in a machine-readable way | 2 August 2026, with marking for older systems by 2 December 2026 |
| New York | Ads containing AI-generated "synthetic performers" that appear to be real people | Whoever produces the ad must conspicuously disclose it | In force (signed December 2025) |
| California | AI provenance data (SB 942, as amended) | AI providers embed hidden disclosures; large platforms must detect and display them | Providers from 2 August 2026; platforms from 1 January 2027 |
| US (federal) | FTC rule on consumer reviews and testimonials | No fake reviews or testimonials, including AI-generated ones that pose as real customers | 21 October 2024 |
Sources: India's IT Rules amendment, summarised by HLC; EU Article 50 timeline, Cloud Security Alliance; New York Governor's office, 9 June 2026; FTC enforcement under the reviews rule, Benesch.
India: what the IT Rules amendment means for ads
The amendment notified on 10 February 2026 defines synthetically generated information as audio, visual or audio-visual content that is "artificially or algorithmically created, generated, modified, or altered" and portrays people or events as indistinguishable from real ones. It must be prominently labelled and carry permanent metadata or provenance markers, and platforms with more than 5 million Indian users must ask users to declare whether content is synthetic and verify those declarations.
Two details matter for advertisers:
- Routine editing is exempt. Good-faith edits, colour correction, compression, translation and accessibility tools do not turn a real ad into synthetic content.
- A stricter draft is pending. A draft published on 30 March 2026 would require the label to be shown continuously for the full duration of a video. Comments closed on 7 May and it has not been notified yet (MediaNama). If it passes, every realistic AI video ad in India needs an on-screen label from the first frame to the last.
India also has the advertising standards body ASCI, whose influencer guidelines already require virtual influencers to disclose that they are not real people. An AI presenter in a UGC-style ad sits very close to that line.
The EU: Article 50
The AI Act's transparency rules did not move when the Digital Omnibus delayed other parts of the Act. Since 2 August 2026, anyone who uses AI to make a deepfake (image, audio or video that resembles real people, places or events and would falsely appear authentic) must disclose that it was artificially generated. AI providers must also mark outputs in a machine-readable way; systems already on the market had until 2 December 2026.
There is an exception for content that is evidently artistic, satirical or fictional, where disclosure can be done in a way that does not spoil the work. A claymation ad is not a deepfake. A photorealistic "customer" talking to camera probably is.
The US: New York, California and the FTC
New York was first to target ads directly: anyone producing an ad must identify whether it includes AI-generated synthetic performers, described as digitally created media that appear to be a real person. It applies to ads shown to New Yorkers, so a brand based elsewhere is not exempt.
California works through the tools: AI providers embed hidden provenance data, and from 1 January 2027 large platforms must detect it and show it to users. Your ad may be labelled whether or not you add anything yourself.
The FTC is the one most AI UGC ads should worry about. Its rule on reviews and testimonials, in force since October 2024, bans fake reviews and testimonials, and the FTC treats AI-generated reviews as covered. A synthetic person saying "I've used this for three months and my skin cleared up" is a testimonial from someone who does not exist.
The UK
There is no AI-specific ad law yet. The Advertising Standards Authority applies the existing rule that ads must not mislead, and it has already ruled against beauty ads that used filters to exaggerate results. An AI-generated before and after that shows results the product cannot deliver is misleading whether or not it is labelled.
What the platforms label for you
You may not be the only one adding a label. Meta automatically labels ads made or significantly edited with its own generative tools, and ads it detects were made with third-party tools using standards like C2PA; the disclosure appears under "About this ad" (Social Media Today, 7 July 2026). For political and social issue ads, Google (since late 2023) and Meta (since early 2024) have required advertisers to disclose realistic digitally altered content. TikTok asks advertisers to mark ads that contain AI-generated content.
A platform label does not replace a legal disclosure. It sits behind a menu; New York and India's draft both want something the viewer actually sees.
Which AI ad formats need a label
The useful question is not "is this AI?" but "does this show a realistic person or event that is not real?".
| Format | Realistic synthetic person? | Disclosure risk |
|---|---|---|
| Your product photo placed in a generated scene, no people | No | Low |
| Hands-only voiceover (hands open the box, a narrator talks) | No face | Low |
| Claymation, 3D animation, stickman, paper-craft | Obviously not real | Low |
| Found footage or dramatised TV segment | Yes, a realistic actor | Medium: label it as a dramatisation |
| AI talking-head "creator" reviewing the product | Yes | High: label it, and never present it as a real customer |
| AI before and after with a realistic person | Yes | High: label it, and the result must be one the product can deliver |
A realistic AI presenter. This is the format all of the rules above are aimed at: label it, and write the script as a presenter talking about the product, not a customer reporting results they never had.
Hands and a voice, no face. The UGC feel survives without a synthetic person on screen, which keeps it out of most of these rules.
How to build disclosure into the creative
A label added at the end usually looks like a warning. Designed in, it looks like part of the format.
LocalAds' found-footage format renders a "Dramatization, AI-generated" tag on every ad, styled like a broadcast caption so it reads as part of the format.
Five habits that cover most of it:
- Label realistic AI people on screen with a short line such as "AI-generated" or "Dramatisation, AI-generated", visible on a phone screen.
- Never write a synthetic person as a real customer. No "I bought this", no invented results, no star ratings attached to them.
- Keep claims to what the product page supports, especially in health, beauty and finance, where regulators look first.
- Use your real product. A generated scene around a real product photo is a very different thing from a generated product that looks better than yours.
- Keep a record of which live ads contain synthetic performers, so you can answer a platform, a customer or a regulator quickly.
How LocalAds handles it
LocalAds builds every ad from your real product page photos, so the product in the ad is your product. Video ads come in 20 preset formats, and many of them, including claymation, Pixar-style, stylised animation, hands-only UGC voiceover, whiteboard and chalkboard, use no realistic synthetic person at all. Scripts are written from your product page, and you read and edit every line before anything renders. The found-footage format always renders a "Dramatization, AI-generated" tag. For other formats with realistic presenters, add the disclosures your markets require before you publish. See 20 video ad formats for which is which.
More on next year: video ad trends for 2027 and ChatGPT Ads in 2027.
FAQ
Do AI-generated ads need to be labelled? Ads that show realistic people or events that are not real increasingly do. India requires labels on realistic synthetic content, the EU AI Act requires deepfakes to be disclosed, and New York requires ads to disclose AI synthetic performers. Ads that only place your real product in a generated scene are mostly outside these rules.
Does India require AI labels on ads? Yes for realistic synthetic content. Since 20 February 2026, India's IT Rules require synthetically generated content that looks real to carry a prominent label and provenance metadata, and large platforms must ask uploaders to declare it. A draft that would require a label for the full length of a video is pending.
Are AI UGC ads legal? Yes, if they are honest. The risk is presenting an AI person as a real customer. In the US the FTC treats AI-generated testimonials that pose as real consumers as fake reviews. Label realistic AI presenters and keep their script to what the product page supports.
Does Meta label AI-generated ads automatically? Meta labels ads made or significantly edited with its own generative AI tools, and ads it detects were made with third-party AI tools through standards like C2PA. The label appears under "About this ad". It does not replace a disclosure the law requires on screen.
Does a claymation or cartoon ad need an AI label? Generally not under the deepfake rules, because no one would mistake it for real footage. The EU AI Act explicitly allows lighter disclosure for evidently artistic or fictional content. Platform labels may still apply if the platform detects AI provenance data.
When does the EU AI Act labelling rule apply? Article 50's transparency duties apply from 2 August 2026. AI systems already on the market before then had until 2 December 2026 to add machine-readable marking.