Brands are experimenting with AI-generated personas and AI-made content, and readers often ask whether different rules apply. This article sets out what the UK’s Advertising Standards Authority (ASA), the Competition and Markets Authority (CMA) and the US Federal Trade Commission (FTC) have published, and where the published material is silent.
Do separate rules exist for AI-generated influencers in the UK?
According to the ASA’s CAP News article of 29 May 2025, the CAP and BCAP Codes, which are the UK advertising codes the ASA enforces, contain no AI-specific rules. The same article says the existing rules apply “regardless of how content is generated, edited, or targeted”, and describes the ASA’s position as being that “the same rules apply in the same way, regardless of whether AI is used or not.”
A second ASA article, dated 11 June 2026, makes the same point in different words: the Code is “media-neutral, meaning the rules apply regardless of how the content was created.”
So, as of those dates and for the UK, the starting point is not a special rulebook. It is the ordinary rulebook applied to a new kind of content.
Does a virtual influencer still need an “Ad” label?
The ASA’s guidance on recognising ads says that whenever a brand gives an influencer “a payment or any other incentive (requested or unsolicited)”, the content “will need to be obviously identifiable as advertising”. The same page says its definition of an influencer includes a virtually produced persona, which suggests that labelling expectations are not meant to stop at human creators. That page does not otherwise discuss AI, so the application to virtual personas is a reading of the guidance, not an AI-specific statement.
The ASA’s published labelling points, as of its current advice, include:
- The ASA and the CMA prefer labels that explicitly call out the content as advertising, such as “Ad”.
- The label should be upfront, which usually means at the beginning. Disclosure that appears only after clicking “See more” is not acceptable.
- Labels such as “Gifted”, “Supported by” and “In association with” are described as weak, and the ASA advises against “sponsored” because it is open to varied interpretation.
- A platform tool such as a “Paid partnership” label may be sufficient if it is clear and prominent. Otherwise an “Ad” label should be added.
For a fuller walk-through, see our guide to influencer disclosure rules in the UK, US and EU.
Does saying “this influencer is AI-generated” fix a misleading ad?
Not by itself. The ASA’s May 2025 article says that “disclosure alone is very unlikely to mitigate the harm caused by a fundamentally misleading message.” Its example is an AI image showing cosmetic results that do not reflect real-world outcomes, where an AI disclosure would not cure the misleading impression.
The same article adds a more hopeful note for virtual personas: making clear that an influencer is AI-generated “could well help to negate an otherwise misleading impression.” Whether that works depends on the specific content and how the likely audience would read it.
The article also says there is no blanket legal requirement in the UK to disclose AI use. It suggests two questions advertisers can ask:
- Is the audience likely to be misled if the use of AI is not disclosed?
- Does the disclosure clarify the ad’s message, or contradict it?
It also cites an industry principle from ISBA and the IPA that transparency matters where AI “features prominently in an ad and is unlikely to be obvious to consumers”. The ASA added that it is “ready to change our approach if that becomes the proportionate way” to regulate ads in future, so this is a position as of May 2025, not a fixed rule.
What happens when an AI-made endorsement looks real?
The ASA’s June 2026 article addresses deepfakes directly. It says: “If an AI-generated celebrity ‘endorsement’ looks real enough to mislead consumers, it’s likely breaching the Code.” It also tells advertisers to consider the legal risks of using someone’s likeness without permission and to seek legal advice.
The same article names three ASA rulings as examples of the principle that the technology does not shift responsibility:
| Ruling named by the ASA | Point the ASA drew from it |
|---|---|
| Polyverse Inc | AI-generated images can breach the Code |
| Cosmos Oyun Yazilim | In-game ads for an AI simulator app that objectified women breached the Code |
| Stripe & Stare Ltd | An image made with Google’s automated tools was still the advertiser’s responsibility |
The ASA’s summary is blunt: “If an AI tool produces something problematic, the advertiser remains responsible for any output used in their ads.” For creators, the practical implication, which is this article’s reading rather than the ASA’s wording, is that using an AI tool to produce a sponsored image or voice does not move responsibility elsewhere.
Can a persona that does not exist give a genuine opinion?
The CMA’s guidance for content creators, updated on 3 September 2025, says incentivised posts must be clearly marked as ads and “must reflect your genuine experience”. It covers photos, videos, reels, stories, podcasts and other online posts. The page does not mention AI-generated content or virtual influencers.
Reading the two together is the article’s own interpretation, not a statement from either regulator. A persona that cannot taste, wear or use a product has no first-hand experience to report. Any experience-style claim made through such a persona would therefore need careful thought about whether it could mislead. The ASA’s cosmetic-results example above shows the kind of concern regulators describe.
Edited or generated visuals that overstate what a product does are a related concern for regulators.
Who is responsible if a post breaks the rules?
On labelling, the ASA says that “Both brands and influencers are responsible for ensuring that ads are obviously identifiable.” For affiliate marketing, it says it typically treats the brand as “at least jointly responsible”. The CMA says everyone involved in creating the content shares responsibility for correct labelling. Where a brand has editorial control of the content, the ASA says it can take action under the CAP Code.
The CMA, local Trading Standards and, in Northern Ireland, the Department for the Economy can take action under consumer law, and the ASA can name non-compliant advertisers. Our piece on how the ASA handles complaints about influencer ads explains the complaint process.
The ASA’s influencer guide itself was updated on 23 October 2026. It is now in its fourth edition, and the ASA says it reflects “recent changes in consumer law and new research into how people recognise influencer advertising”. The ASA’s page lists new information on fake consumer reviews, updated guidance following the Digital Markets, Competition and Consumers Act, and information on the CMA’s enforcement powers and penalties. The page content we reviewed did not mention AI.
How does the US FTC treat AI avatars in marketing?
The picture below applies to the United States only, and should not be read as a UK rule or a universal one.
The FTC’s Consumer Reviews and Testimonials Rule went into effect on 21 October 2024. In its questions-and-answers guidance, the FTC says the rule “has no blanket prohibition on the use of AI-generated avatars in marketing”. It also says AI-generated “stock avatars” would not be giving “consumer reviews” as the rule defines that term, though an avatar’s message might be considered a testimonial. That would be prohibited under the rule only if the underlying testimonials were fake or false.
Two further points from the FTC page matter for readers:
- Using a celebrity’s avatar without permission can violate the rule if consumers would believe the celebrity gave a testimonial.
- Using avatars could also be deceptive in violation of the FTC Act, with the FTC pointing to its Endorsement Guides.
The rule’s definition of “fake indicators of social media influence” also includes “purported individual accounts not associated with a real individual”, although the FTC says that concerns follower-type metrics rather than reviews. Our article on fake followers and fake reviews under US and UK rules covers that side.
What do platforms add on top?
Regulators are not the only source of requirements. The ASA’s May 2025 article mentions “platform-based rules and industry initiatives” encouraging responsible and transparent use of AI, without giving specifics. Platform labelling tools for AI content are a separate matter from advertising law, and our explainer on how YouTube and Meta label AI-generated content describes what those two platforms say. A platform label is not necessarily a substitute for an advertising label, and the reverse is also true.
Where the published rules leave AI personas
Across the sources, the common thread is consistency. In the UK, the ASA and CMA apply existing labelling and misleading-advertising rules to AI-made content, with no separate AI regime as of the dates above. Labelling a persona as AI-generated may help avoid a misleading impression but does not rescue a misleading claim. In the US, the FTC says there is no blanket ban on AI avatars, but fake or unauthorised testimonials are a problem however they are produced. Readers working across both markets should check each regulator’s current pages, because the ASA has said it may change its approach and guidance is being updated regularly.
This article gives general information based on published guidance. It is not legal or regulatory advice, and rules and platform policies change and differ by country.
Sources: ASA (Disclosure of AI in Advertising: Striking the Balance Between Creativity and Responsibility); ASA (AI and Deepfakes: Four Things Advertisers Need to Know Before They Hit “Run”); ASA (Recognising ads: Social media and influencer marketing); ASA (Influencers’ guide to making clear that ads are ads); CMA on GOV.UK (Social media endorsements: being transparent with your followers); FTC (The Consumer Reviews and Testimonials Rule: Questions and Answers)
