UK AI ad disclosure: no blanket rule, two CAP questions to ask

No: the UK has no blanket legal requirement to say an ad used AI. The Advertising Standards Authority (ASA) states it directly: "There is no blanket legal requirement in the UK to disclose the use of AI in ads." What it gives advertisers instead is two questions to ask before deciding whether to disclose, and the answer to both turns on whether the ad could mislead, not on how the footage was made.

This post reads the ASA's guidance on disclosing AI in advertising (dated 29 May 2025) and its June 2026 piece on AI and deepfakes, both read on 2026-10-03, and Sume's docs for video inspect. It is not legal advice. If your ad is about a real celebrity's likeness, start with ASA on AI deepfake ads.

What are the two questions?

The ASA's guidance asks marketers to consider:

• "Is the audience likely to be misled if the use of AI is not disclosed?"
• "If there is a danger of the audience being misled, is the disclosure clarifying the ad's message or contradicting it?"
• The ASA also says the CAP and BCAP Codes "do not contain AI-specific rules" and that the existing rules apply regardless of how content is generated, edited, or targeted.

When does a disclosure help, and when does it not?

The ASA's examples draw the line. A disclosure can clarify something that would otherwise mislead, but it cannot rescue a message that is misleading to begin with.

How do I run the questions on a Sume video before it goes live?

The questions are about what viewers see and hear, so look at the delivered file, not the prompt. Sume's video inspect reads one clip already stored on media.sume.com and returns probe facts, stills as image artifacts and, if you ask, a transcript. Without a frames field it returns 8 stills; frames: { at: [...] } takes 1 to 24 explicit times. transcribe: true runs speech-to-text at a public rate of $0.01 per audio minute, and a clip with no audio track returns inspect_source_has_no_audio.

Ask for stills at the moments a product claim appears, plus the transcript, then have a person read them against the two questions. Inspect does not judge the claim; it hands you the evidence to judge. Import a file first with POST /v1/media-imports, because inspect refuses off-host URLs.

curl -X POST https://api.sume.com/v1/video-inspect \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: uk-ad-review-001" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/artf_demo/ad.mp4",
    "frames": { "at": [1, 5, 10, 14] },
    "transcribe": true,
    "language_code": "en"
  }'

Who answers for the output?

The advertiser. In its June 2026 piece the ASA says the Code is media-neutral, and that although an ad in its example was produced with Google's automated tools, the advertiser was still primarily responsible for compliance. A generator, an avatar tool or an ad platform's automation does not move that responsibility.

If you decide to disclose, the label is your copy, so you place it. Burning your own AI disclosure with caption cues is one route, and Sume's video captions accept authored cues with text, start and end. A label does not replace a review: the EU's Article 50 label is not an ad disclosure either, and a creator's footage deserves a transcript read before paid reuse.