UK rules for AI ads and AI influencers: a one-page explainer for small businesses

Last checked: 4 November 2026. We opened every external link on this page on 4 November 2026 (09:45–09:46 GMT) and recorded the HTTP result for each; the log is published below. Links to new pages on this site were rechecked after they went live.

The short version: In the UK you can use AI to make an ad. It's not an offence to show an AI presenter, avatar or character. What you can't do is use one to make people believe something false, and the clearest example is a made-up "customer" who seems to vouch for your product from real experience. Adding an AI label doesn't fix that. If an ad claims or implies a genuine customer experience that never happened, it's still a fabricated testimonial or fake review, with or without a label.

Not legal advice. This is a plain-English summary for small businesses, written by The Independent AI. For a specific campaign, read the official text below or ask a qualified adviser.

---

Two very different things

	(a) An obviously AI or fictional presenter	(b) A fake "real customer"
Example	An AI-animated mascot or an avatar clearly presented as a brand character, or an actor reading a script	An AI person saying "I've used this for three months and my skin cleared up", shown as if they're a real buyer
UK position	Can be fine if the ad doesn't mislead (CAP rule 3.1). Rule 3.47 doesn't require evidence for testimonials that are "obviously fictitious".	Breaks rule 3.47 because you can't show the testimonial is genuine. It may also be a banned fake consumer review under the DMCC Act.
Does an AI label help?	It can help avoid a misleading impression.	No. The ASA says you can't rely on disclosing AI use to undo a misleading impression. The claim of real experience is the problem, and the label doesn't remove it.

What the ASA says, in its own words:

• Disclosure should clarify the message, not contradict it. The ASA suggests asking whether the audience is likely to be misled, and if so, "is the disclosure clarifying the ad's message or contradicting it?" Its example: an AI image showing a cosmetic result that doesn't reflect real-world results is likely to be materially misleading, even if you then rely on "a disclosure that AI was used to rebut the misleading impression created".
  • Source: ASA/CAP: Disclosure of AI in advertising (29 May 2025)
• Saying an influencer is AI-generated "could well help to negate an otherwise misleading impression". The ASA adds that "it will always come down to the specific content and context". This is about honest framing, such as making clear it's a virtual character. It doesn't make a fake customer claim acceptable.
  • Source: the same ASA/CAP article as above.
• "It is a breach of the Code for a marketer to write reviews of their own products whilst posing as a genuine consumer."
  • Source: ASA advice: testimonials and endorsements
• "If an AI-generated celebrity 'endorsement' looks real enough to mislead consumers, it's likely breaching the Code." The advertiser also "remains responsible for any output used in their ads".
  • Source: ASA/CAP: AI and deepfakes, four things advertisers need to know (11 June 2026)

---

Red-flag phrases: if a tool sells itself like this, slow down

The pitch says	Why it's a problem for you
"Undetectable"	You're the advertiser, so you carry the risk. If an ad is designed so that nobody realises the "customer" is fake, the ad is likely to mislead.
"99% realism" / "99.84% realism score"	These are the tool maker's own figures. Realism itself isn't banned. The risk comes when realism is used to make people believe a real person had a real experience.
"No viewer can tell" / "no platform can tell"	Telling viewers is exactly what TikTok's ad policy requires for fully AI-generated ad content. It may also be required in the EU if the content counts as a deep fake (see below).
"Real customer reactions" or "UGC-style testimonials" made by AI	A review that claims to come from someone's genuine experience but doesn't is banned under the UK's DMCC Act and the US FTC rule. A label doesn't change that.

Worked example: Hype or Fact? Genrok's "99% realism, undetectable AI ads" claim

---

United Kingdom

CAP Code rule 3.1: don't mislead

• Official text: CAP Code rule 3.1
• Exact wording: "Marketing communications must not materially mislead or be likely to do so."
• In plain English: If an AI person leads viewers to believe something false that affects their decision, the ad breaks the code, whatever tool made it.

CAP Code rule 3.47: testimonials must be genuine

• Official text: CAP Code rule 3.47
• Exact wording: "Marketers must hold documentary evidence that a testimonial or endorsement used in a marketing communication is genuine, unless it is obviously fictitious, and hold contact details for the person who, or organisation that, gives it."
• In plain English: An AI "customer" has no real experience to back up a testimonial. If the testimonial is presented as genuine, you can't comply with this rule. The only exception is a testimonial that is obviously fictitious, and an ad built to look real isn't.

Digital Markets, Competition and Consumers Act 2024, Schedule 20, paragraph 13: fake reviews

• Official text: DMCC Act 2024, Schedule 20, paragraph 13
• What it bans:
  • submitting, or commissioning someone else to submit or write, "a fake consumer review";
  • publishing consumer reviews "in a misleading way".
• What counts as fake: a review that "purports to be, but is not, based on a person's genuine experience".
• In plain English: An AI person "reviewing" your product as if they'd bought it risks being a banned fake review. These practices are banned outright, and you don't need to show that anyone was harmed.
• How to comply: CMA: Fake reviews guidance for businesses (CMA208)

---

If you sell to the US: FTC rule, 16 CFR Part 465

• 16 CFR 465.1(f): A "consumer testimonial" includes "depictions of the name, signature, likeness, or other identifying personal characteristics of an individual".
  • In plain English: Using an AI face or voice doesn't by itself make an ad a "consumer testimonial". It counts as one when viewers would understand it as a genuine customer's own experience of the product. An obviously fictional or AI presenter is a different thing.
• 16 CFR 465.2(a): A business must not "write, create, or sell" a consumer testimonial that materially misrepresents "that the reviewer or testimonialist exists" or that they used the product.
• 16 CFR 465.2(b): A business must not spread one that it "knew or should have known" was false.
  • In plain English: A made-up AI "happy customer" presented as real is exactly what this rule bans.

---

If you sell to the EU: AI Act, Regulation (EU) 2024/1689

We checked the AI Act Service Desk text, which includes the 2026 amendments, and the Commission's transparency pages.

• What counts as a deep fake: Article 3(60)
  • What it says: A deep fake is "AI-generated or manipulated image, audio or video content that resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful".
  • In plain English: Being realistic isn't enough on its own. The content has to resemble something that exists and falsely appear real. A clearly invented cartoon mascot is unlikely to qualify. A lifelike "customer" who looks like a real, existing person and appears authentic may well qualify.
• What deployers must do (the business using the tool): Article 50(4)
  • What it says: You "shall disclose that the content has been artificially generated or manipulated".
  • Exceptions:
    • Where use is authorised by law to detect, prevent, investigate or prosecute criminal offences.
    • For content that's part of an "evidently artistic, creative, satirical, fictional or analogous work or programme", the duty is limited to disclosing the AI content "in an appropriate manner that does not hamper the display or enjoyment of the work".
  • Timing (Article 50(5)): The disclosure must be "clear and distinguishable", given at the latest at first exposure.
• What tool makers must do: Article 50(2)
  • What it says: Providers of generative AI must mark outputs "in a machine-readable format and detectable as artificially generated or manipulated".
  • When it doesn't apply: where the AI performs "an assistive function for standard editing", does "not substantially alter the input data … or the semantics thereof", or is authorised by law for criminal-offence purposes.
  • In plain English: This duty falls on the tool maker, not on you. It's still worth noticing when a tool sells "undetectable" output.
• When it applies: Article 113 and Article 111(4)
  • The Act generally applies from 2 August 2026, and Article 50 isn't among the exceptions with other dates.
  • Under Article 111(4), added in 2026, providers whose generative AI systems were placed on the market before 2 August 2026 have until 2 December 2026 to comply with Article 50(2) (machine-readable marking).
  • We found no similar deferral for the deep-fake disclosure duty in Article 50(4).
• Commission guidance:
  • Guidelines on the Article 50 transparency obligations
  • The voluntary Code of Practice on Transparency of AI-generated Content. The Commission says signing the code is voluntary but the Article 50 duties are legal obligations.

---

Platform rules: company policies, not laws

A platform's AI label, or the lack of one, says nothing about whether your ad is lawful. UK, US and EU rules apply whatever the platform does.

• TikTok: Advertising Policies: Misleading and false content, "Edited media and AI-generated content (AIGC)". The page says "Last updated: April 2026".
  • What it says:
    • Significantly edited media and AIGC "is allowed if" you "Apply the AIGC label, or" add "a clear disclaimer, caption, watermark, or sticker of your own".
    • This covers content with images, video or audio that are "completely AI-generated", and making someone appear to do or say things they didn't.
    • Minor edits such as lighting or background changes don't count.
    • "If we identify AI-generated content that has not been disclosed, your ad will be rejected or restricted."
  • In plain English: On TikTok you must disclose AI content. Disclosing it meets TikTok's rule, but it doesn't fix a fake testimonial under UK law.
• Meta (Facebook, Instagram):
  • What Meta labels: Meta's Help Centre says Meta labels ad images "created or materially edited using our generative AI creative features" with "AI info".
    • The label sits in the three-dot menu, unless the image includes an AI-generated photorealistic human. In that case it appears next to "Sponsored".
    • Meta says "not all ads will have AI info", and lists "when ads are created or modified with third-party creation or AI tools" as one example.
  • What changed on 1 June 2026: Meta's update of that date says it is "beginning to roll out" an "About this ad" area and will "begin automatically detecting ads created or edited using third-party AI tools through industry-standard signals". When it detects one, it will apply an "AI info" label inside About this ad.
    • So third-party AI detection is being rolled out and depends on detectable signals. It isn't guaranteed for every ad, and the Help Centre page above still describes the older position.
  • Advertiser duty: For ads about social issues, elections or politics, Meta says advertisers "are already required to disclose" digitally created or altered images, video or audio, including those made with third-party AI tools.
  • In plain English: Meta may or may not label your ad, and either way the label doesn't make a fake customer lawful.

---

A five-point checklist before you run an AI ad

1. Is anyone shown as a real customer, user or fan? If they're AI, don't have them claim real experience. Rewrite the ad as an obvious brand character or a scripted presenter, and make sure nothing implies they bought or used the product.
2. Could you produce evidence and contact details for every testimonial? (CAP 3.47) If not, cut it. A label isn't a substitute.
3. Does the ad look like a review or a genuine reaction? Check it against DMCC Schedule 20 paragraph 13 and CMA208.
4. Will it run on TikTok, or reach EU or US audiences? Follow TikTok's AIGC rule. In the EU, disclose deep-fake content. In the US, check 16 CFR 465.
5. Does the tool's sales page use the red-flag phrases above? Treat its realism and "undetectable" claims as the company's own marketing, not as protection for you.

---

Related:

• How we judge
• All Hype or Fact? checks
• The AI claims tracker
• All guides

Independence: The Independent AI has no commercial relationship with Genrok, TikTok, Meta, the ASA/CAP, the CMA, the FTC or the European Commission. This page contains no paid or affiliate links.

Licence: The text of this guide is © The Independent AI and licensed under CC BY 4.0. Attribute it as: "The Independent AI, theindependentai.com". The licence covers our own text only. It doesn't cover quoted regulations, platform policies, third-party logos or the linked sources, which belong to their owners.