ASCI: an AI testimonial ad stays prohibited even with a label

Yes, it stays prohibited. ASCI's guidelines say advertisements that make misleading claims or violate the ASCI Code violate it even if an AI label is used, and the first example is using AI to create, alter or attribute an endorsement or testimonial that falsely says a real person used, approved or recommended a product.

This summarises a published guideline and is not legal advice. The document is dated 17 September 2026 and comes into effect 3 months after publication. Read 2026-10-01.

What does ASCI list as prohibited even with a label?

Section 1 of the guideline gives four examples.

Where do Sume's avatar and face-swap routes fit?

Avatar videos are script-driven talking videos from a ready avatar, so the speaker is the avatar identity you supply. Face swap is a Beta endpoint that applies a ready avatar face onto a public HTTPS source video. Neither route makes a statement about whether a script or a source clip is truthful; that decision stays with the advertiser. A testimonial script that attributes words to a real person who never said them is example A, whichever route renders it. See which endpoint to use.

What does the label do then?

ASCI's section 4 says brands may use platform labels or their own, such as "Audio/Video created using AI". Authored cues through Video captions carry text, start and end, so a label line is a cue. The guideline treats that as disclosure for content that is otherwise acceptable; it does not turn a prohibited ad into an allowed one.

What should a team check before publishing?

Check each testimonial against a real person who actually gave it, check any before-after visual against substantiated results, and check consent for any likeness or voice. For the labelling side, use ASCI's three questions in when an AI ad needs a label. A comparable UK rule is covered in the ASA and CAP deepfake endorsement post.