ASCI releases draft guidelines on labelling AI-generated content in advertising

MUMBAI: Adland has a new rulebook in the making, and it is not out to ban the machines. The Advertising Standards Council of India (ASCI) today released draft guidelines for the responsible labelling of synthetically generated content in advertising, and the message is refreshingly practical: judge the harm, not the tool.

The announcement came from Mumbai on May 12, 2026. The guidelines are aligned with the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules (2026), amended on February 10. The aim is to ensure transparency while avoiding consumer label fatigue around synthetically generated information. The draft is open for stakeholder consultation till June 13.

The timing makes sense. As brands increasingly turn to artificial intelligence (AI) to power their campaigns, the draft advocates a risk-based approach that focuses on consumer outcomes rather than regulating the technology itself. Under it, AI use in advertising is considered misleading or harmful only when it creates unfulfillable expectations, exploits vulnerable populations, depicts unsafe situations, or replicates a real person’s likeness without consent. In other words, using AI is not the offence. Misleading the consumer is.

The requirement to label AI-generated content rests on the risk it poses to consumers. The guidelines sort AI-generated advertising content into three categories.

1. High risk (prohibited content). These advertisements are illegal, infringe on rights, make misleading claims or violate the ASCI Code. They will violate the code even if an AI label is used, so a label is no shield. Examples include:

• fabricating endorsements or testimonials
• exaggerating product results or features through claims or visual representations to create a misleading impression
• creating fake locations that appear realistic to the consumer
• using deepfakes, copyrighted work or a person’s likeness without consent
• using AI to generate fictional authority figures with identifiable cues, such as an AI-generated fake doctor promoting a supplement, implying medical endorsement or expertise

2. Medium risk (labelling required). Here AI use materially influences consumer decisions, and the lack of disclosure would mislead consumers. Labelling is mandatory to help consumers understand the nature of the representation. Examples include:

• using virtual or synthetically generated influencers and ambassadors
• replicating a real person’s likeness or voice, even with their consent, for personalised messaging
• using synthetically generated visuals for product performance, unless the visuals replicate how the product actually performs
• creating realistic events, settings or situations entirely with AI
• demonstrating a product that does not currently exist
• creating AI-generated exaggerated sound effects that are highly relevant to the product’s core features
• using AI for paid or sponsored product suggestions, which must specifically be labelled as “sponsored by”

3. Low risk (no labelling required). These advertisements feature minor modifications, or use AI in ways that have no material impact on a consumer’s ability to make an informed choice. No label is required for:

• Minor enhancements: routine editing, colour correction, noise reduction, standard blemish removal, and minor lighting tweaks that do not alter the substance or core claims of the ad
• Background and ambient elements: purely decorative AI-generated backgrounds, abstract skylines, ambient music, jingles, or background sound effects (like crowd cheers) that are unrelated to the product’s actual capabilities or promise
• Fantastical elements: obvious, unrealistic effects that audiences recognise as not depicting reality, such as dragons or fairies
• Administrative and text uses: generating or enhancing advertising copy, creating audio descriptions for the visually impaired, or preparing documents in good faith without creating false records

Where disclosures are required, brands may use standard labels such as “Audio/Video created using AI” or “Audio/Video enhanced using AI”, or any other alternative labels that accurately inform the consumer. Disclaimers must follow the ASCI Code on disclaimer guidelines where applicable.

What does this add up to? Three things stand out. First, the draft draws its line at the consumer, not the code. A tidy-up of skin tone or a dragon in a soft-drink ad needs no label, while a virtual ambassador or an AI-built demonstration of a product that does not yet exist does. Second, the rules are sharp where it matters. Even a real person’s likeness or voice, used with consent for personalised messaging, needs a label, and a fake doctor selling a supplement is prohibited outright, label or no label. Third, the stress on avoiding label fatigue is smart. If every ad carries a tag, nobody reads any of them. By reserving labels for cases where AI materially shapes a buying decision, ASCI is trying to keep disclosure meaningful. For brands, agencies and creators, the framework offers clarity on where the lines sit.

The draft guidelines are open for public consultation and can be viewed through the link in the press release. ASCI invites feedback from industry, consumer groups and other stakeholders by June 13, 2026, after which the guideline finalisation process will begin. Stakeholders can submit their feedback at contact@ascionline.in.