Which ASA and CAP Code rules apply to AI-generated ad content?

What to take away

How the CAP Code and ASA regulate AI-generated ad content

The CAP Code is the rulebook the Advertising Standards Authority applies to most UK ads, including online and social media content. It is written by the Committee of Advertising Practice, and the ASA enforces it independently.

The code is not law, but compliance is expected, and persistent breaches can lead to sanctions. The full set of rules, including the broadcast and non-broadcast codes, is published by the ASA Advertising codes - ASA | CAP.

The ASA regulates ads across traditional and digital media. That includes paid search, display, influencer posts, websites and emails, as long as they fall within its remit. Ad copy is judged on its claims, not on the method used to produce it. If an ad appears in a medium the ASA covers, the same rules apply.

Broadcast ads sit under the BCAP Code, which the ASA also administers. Financial promotions are regulated by the Financial Conduct Authority as well as the CAP Code, so a claim in a fintech ad can breach both regimes.

AI content tools are now used for drafting headlines, product descriptions, social posts and even entire campaigns. The speed and scale they offer can make it easier to publish unverified claims. The CAP Code does not ban AI use, but it does require that everything published is accurate and can be substantiated.

A key principle is that the advertiser is responsible for the ad, regardless of how the copy was put together. If you publish it, you are the advertiser in the ASA's eyes. That means you must be able to evidence any claims, and you must not mislead by omission or ambiguity.

The ASA also applies the CAP Code to non-broadcast advertising, which covers most online content. The non-broadcast code sets out the core rules on misleading claims (3.1), misleading omissions (3.3), substantiation (3.7) and recognition of marketing communications (2.1) Non-broadcast Code - ASA | CAP.

For UK marketers, the practical upshot is that AI does not create a regulatory gap. The same standards apply, and the ASA has been clear that it will hold advertisers to them.

If you are building an AI content operation, map your processes to the CAP Code from the start. Our guide to advertising rules covers the wider operational picture advertising rules.

Misleading claims and substantiation in AI-written ad copy

Misleading claims are the most frequent reason for ASA rulings. A claim is misleading if it is likely to deceive the average consumer and affect their decisions. That includes claims that are factually incorrect, that omit important information, or that are ambiguous.

Rule 3.1 requires that marketing communications must not materially mislead or be likely to do so. Drafts can contain plausible-sounding claims about performance, popularity or endorsements that no evidence supports.

Substantiation is the evidence you hold to back up a claim. Rule 3.7 requires that before you publish an ad, you must be able to prove any claim capable of objective substantiation. Evidence must be in hand, not promised for later.

A statistic or testimonial that cannot be traced to a source cannot be substantiated, so the ad should not run.

Common problem areas include comparative claims, such as "better than" or "the best", and claims about environmental benefits. These often need solid evidence. A draft may not reflect what evidence you actually hold, so it can easily overstate your position.

Price claims need care too. A discount misleads if the higher price was never charged, and the Competition and Markets Authority can act alongside the ASA. Check any crossed-out price against your real selling history before you publish.

The table below sets out the claim types that most often trip up AI-assisted copy, and what the CAP Code expects in each case.

Worked example: the skincare page

A UK skincare brand uses an AI content tool to draft a product page. The draft includes the line: "Clinically proven to reduce wrinkles by 80% in two weeks." The brand has no clinical trial to support that figure.

The ASA would ask for the evidence behind the claim. With none available, the claim would be ruled misleading and the ad banned. The fix is to remove the claim, or to replace it with one the brand can prove.

A supported version might read: "92% of users agreed their skin felt smoother after four weeks, based on a consumer study of 120 participants." The study report should be kept on file.

To avoid these problems, you need a process for checking drafts before they go live. That process should verify every factual claim, check that comparisons are fair, and confirm that any endorsements are genuine. A quality checklist can help standardise this step quality checklist.

Rule 3.3 requires that ads do not mislead by omission. If your ad is likely to be understood as making a particular claim, you must not leave out information that would change how consumers see it. A draft may not flag what is material, so human review is essential.

Disclosure rules when AI produces or reviews advertising

Disclosure in advertising is about making sure consumers understand when content is an ad, or when there is a commercial relationship behind it. Rule 2.1 requires that marketing communications be obviously identifiable as such, and it applies to AI-assisted content as much as to any other.

If an ad still reads as an ad, no extra disclosure is usually needed. The rules focus on whether the commercial intent is clear. A paid social media post must be identifiable as an ad, for example with a platform's paid partnership label or a clear hashtag like #ad.

Apply this test. Would a reasonable consumer take the ad differently if they knew how the image, voice or testimonial was produced? If yes, disclose the AI use. If the copy is accurate, substantiated and clearly an ad, no AI label is needed.

The ASA has not issued a blanket rule requiring AI disclosure in every ad.

AI-generated voices, images and influencer content

A paid influencer post that uses an AI-generated voice or face must still be identifiable as an ad, and the synthetic presenter must not be passed off as a real customer. Rule 2.1 covers the ad label; rule 3.1 covers the misleading impression.

Example: a supplement brand posts a clip in which an AI-generated "customer" describes her results, with no ad label and no real person behind it. Readers would take the speaker for a genuine user, so the post would breach both rules.

If a voice clone or likeness of a real person is used, you need that person's permission and a written release. Keep the release with the substantiation file.

A separate issue is when AI is used to review or moderate advertising. That is not directly covered by the CAP Code, but it can affect compliance. If an automated tool is used to check ads for regulatory issues, you still need human oversight.

The ASA expects advertisers to take reasonable steps to comply, and an unverified automated check is not enough on its own.

Data protection law also comes into play when AI tools process personal data. The Information Commissioner's Office (ICO) enforces the UK GDPR and the Data Protection Act 2018. If your AI content tool uses personal data, you must have a lawful basis and be transparent about it.

That sits outside the CAP Code, but it is part of the wider compliance picture.

For organisations that want to set clear internal rules, a disclosure policy can help staff understand when and how to reveal AI use. Our article on drafting a disclosure policy explains how to approach it disclosure policy.

How ASA ruling themes apply to AI-assisted campaigns

The ASA publishes its rulings on a regular basis, and each one names the advertiser and the date. You can browse them on the ASA site Rulings - ASA | CAP.

Three themes drive most upheld complaints that touch on AI-assisted copy. First, performance claims: if an online ad states a result and the advertiser cannot produce the evidence at the time of publication, the complaint is upheld. 1.

Third, reviews and testimonials: if AI-generated or copied comments are presented as genuine reviews, the ASA and the Competition and Markets Authority (CMA) can both act.

The evidence standard is the same for every medium. For a before-and-after image, the ASA looks at whether the picture represents a typical result and whether the advertiser can show that. For a testimonial, it looks at whether the person exists, gave permission and used the product.

The ASA judges the overall impression of an ad, not just individual words. If a combination of text and images creates a misleading impression, the ad can still be banned. Human review matters here, because an automated draft will not capture the cumulative effect.

The themes matter because they show where an AI workflow is most exposed. A tool that generates copy at speed will produce claims, comparisons and testimonials faster than a human team can check them. The ruling risk sits in the claims, not in the tool.

For a broader look at the rules and ethics around AI content, see our related article rules and ethics.

Using ASA Copy Advice and AdviceOnline before publication

The ASA offers a free pre-publication service called Copy Advice. It gives advertisers and agencies confidential advice on whether a proposed ad is likely to comply with the CAP Code. That is useful when you are using AI content tools and want to check claims before they go live.

You can find out more on the ASA website Copy Advice - ASA | CAP.

Copy Advice is not a pre-approval scheme, and it does not guarantee that an ad will not be investigated. But it can help you identify potential problems and avoid common pitfalls. The service is available to anyone, and it covers most non-broadcast ads.

AdviceOnline is a separate resource that provides detailed guidance on specific topics. It is not a personalised service, but it can help you understand how the CAP Code applies to different situations. For example, there is guidance on health claims, environmental claims and price claims, all of which are areas where AI-assisted copy can go wrong.

To use these resources effectively, you should build them into your workflow. Before publishing any ad, run it past the relevant AdviceOnline guidance. If you are unsure, submit it to Copy Advice. That step can save you from a costly ruling later.

It is also worth noting that the ASA expects advertisers to take compliance seriously. If you repeatedly publish misleading ads, the ASA can take further action. Using the free advice services shows that you are trying to comply, which can count in your favour if a complaint is made.

For agencies and brands that produce a high volume of AI content, setting up a standard pre-publication check is sensible. That might include a checklist of common issues, a review by a qualified person, and, where needed, a Copy Advice request. Our article on service standards covers how to embed these steps service standards.

A pre-publication checklist

Sanctions, the non-compliant online advertisers list and remediation

If the ASA upholds a complaint, it publishes a ruling. The ruling names the advertiser and describes the ad and why it breached the code. The ASA also asks the advertiser to withdraw or amend the ad. Most advertisers comply.

For those that do not, the ASA has a range of sanctions. These include asking search engines and social media platforms to remove paid ads, and working with partners to block content.

The ASA also publishes a list of non-compliant online advertisers, naming those that have not complied with a ruling. The sanctions page on the ASA site explains the process Sanctions - ASA | CAP.

The non-compliant online advertisers list is a reputational risk. It is used by platforms and can affect ad delivery. For brands that rely on online advertising, appearing on the list can be damaging.

Remediation is about fixing the problem. If an ad is ruled against, you need to take it down or change it. You should also review your processes to prevent similar issues. That might mean retraining staff, adjusting your AI content tool settings, or adding more human checks.

The ASA does not fine advertisers directly, but non-compliance can lead to other consequences. For example, the CMA can take action under consumer protection law if ads are misleading. The ICO can act if personal data is misused. So the overall regulatory risk is broader than just the ASA.

If you use AI content tools, it is worth having a remediation plan. That plan should cover how you will respond to a complaint, who will handle it, and how you will correct the ad. It should also include a review of how the problematic wording was produced, so you can avoid repeating the mistake.

Numbered steps for remediation

Common questions