ASCI's three questions: when does an AI ad need a label?
ASCI's September 2026 guidelines ask three questions to decide if synthetic content in an ad needs a label. Here they are, applied to an avatar video ad.
ASCI, India's advertising standards body, tells brands to decide on a label by asking three questions: is the synthetic content central to the ad's persuasion or does it materially influence purchases, does the way it is used mislead about claims or performance if unlabelled, and would leaving out a disclosure create a false impression. A "yes" points toward labelling; this post applies the questions to an avatar video ad.
This is an explanation of a published guideline, not legal advice. The text is dated 17 September 2026 and says it takes effect 3 months after publication. Read 2026-10-01.
What are the three assessment questions?
They are the guideline's own "assessment for labelling" list, which it frames around consumer outcomes rather than the technology.
| ASCI question | Check on an avatar ad |
|---|---|
| Is synthetic content central to the persuasion or purchase decision? | Is the speaker the pitch, as with a synthetic ambassador? |
| Does the manner of use mislead about claims, performance or benefits if unlabelled? | Does the script imply results the product has not shown? |
| Would no disclosure create a false impression of claims or performance? | Would a viewer assume a real person says this? |
Which cases does ASCI already call mandatory?
Section 2 lists labelling as mandatory for synthetically generated influencers and ambassadors, and for replicating a real person's face or voice even with consent when used for personalised messaging. Section 3 lists no label for routine edits such as colour correction, ambient or decorative elements, obvious fantasy, text uses, and accessibility content such as subtitles and captions. A script-driven talking avatar presenting a product sits closer to the first list.
How do I add a label to a Sume avatar video?
ASCI says brands may use platform labels or their own, for example "Audio/Video created using AI". A talking video is generated from a script, so you can burn the label line afterwards as authored cues through Video captions; cues skip speech-to-text. To place a still such as a badge on screen, Timeline compose returns one MP4 from one still and one video.
Does a label settle the question?
No. The guideline says some uses stay misleading or objectionable regardless of labels, and that every ad with synthetic content must comply with the whole ASCI Code. For the prohibited list, see fabricated endorsements. For India's separate rule on routine editing, see the synthetically generated information label.
Sources
Related posts
More in Use cases
- TTS per sentence, then concat: join up to 20 clips in one file
Generate TTS one sentence at a time, then join up to 20 clips with Sume timeline audio concat: sample-domain, no re-synthesis, $0.01 flat per job.
- Blend two videos together API: what Sume can and cannot layer
Sume cannot blend two videos into one frame: video filter works on one clip, and compose puts one still over one video. Two videos join in sequence on Timeline.
- Budget for 500 vertical AI video ads a month: $440 to $1,890
500 ten-second vertical clips cost $439 to $1,890 on Sume by Seedance id and resolution; Runway's credit prices give $800 and $1,800 for two of them.
- SB 1050 takedown orders: what an advertising medium must do
SB 1050 bars an advertising medium from running an ad after a served court order. How ad files and URLs fit in, and what a corrected cut means.
Written by Sume