What ASCI published
ASCI is the self-regulatory body of India's advertising industry. Founded in 1985, it maintains a code requiring ads to be legal, decent, honest and truthful, and its Consumer Complaints Council rules on complaints about ads in any medium, from television and print to websites and influencer posts. Its guidelines are not legislation, but they set the standard against which complaints in India are judged.
In May 2026 ASCI published a draft for consultation, which closed on 13 June. Its press release said the draft was aligned with amendments to India's Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules made on 10 February 2026, which deal with synthetically generated information online. The final “Guidelines for Responsible Labelling of Synthetically Generated Content in Advertising” were released on 29 September. The three-page document is signed by the chairman of ASCI's board of governors and dated 17 September, and says it takes effect “on the expiration of 3 months from the date of their publication”.
Synthetically generated content is defined broadly: audio, visual or audio-visual material in an ad that is created or materially altered by a computer “to appear real, authentic or true” and depicts a person or event in a way likely to be taken as real. Text is not in the definition, which is why AI-written copy falls outside the labelling rule.
When a label is required
ASCI says the guidelines are about “managing consumer outcomes rather than regulating the technology”, and gives advertisers three questions. Is the synthetic content central to the ad's persuasive message or does it materially influence purchase decisions? Would its use mislead about the product's claims, performance or benefits if unlabelled? Would leaving out a disclosure create a false impression?
Where the answer is yes, a label is mandatory. The guidelines list six examples:
Synthetic influencers and ambassadors. A computer-generated spokesperson must be disclosed.
A real person's likeness or voice, replicated for personalised messages, even with consent. ASCI's example is a brand sending consumers personalised messages featuring a digital replica of a real person's face or voice.
Fabricated events, settings or situations that may affect how consumers understand the product or service.
Products that do not yet exist, such as a rendering of an unbuilt housing development.
Synthetic sound that is the product's point, such as audio in an ad for headphones.
Paid or sponsored AI product suggestions. Where an AI recommends a product because the brand paid, the recommendation must be identified as “Sponsored by [Brand]”.
Where a label is required, brands may use the platform's own label or wording such as “Audio/Video created using AI” or “Audio/Video enhanced using AI”, or any alternative that accurately informs the consumer, following ASCI's existing rules on disclaimers. No label is needed for routine editing and retouching, decorative backgrounds and ambient music, obviously fantastical effects, accessibility uses such as subtitles and translation, or “generating or enhancing advertising copy”.
What a label cannot fix
The guidelines are explicit that a label does not rescue an ad that breaks the code. Four kinds of use are prohibited outright: using AI to fabricate an endorsement or testimonial so that a real person appears to have used or recommended a product when they have not; exaggerating results with synthetic images (ASCI's example is an AI-generated before-and-after body for a weight-loss product); showing invented locations or facilities as real; and using copyrighted work, deepfakes or a person's likeness without consent. ASCI's chief executive, Manisha Kapoor, said in the announcement that the guidelines clarify “where the mere use of an AI label may not make an otherwise misleading advertisement acceptable”.
Number of lettered examples in the guidelines dated 17 September 2026
Source: Advertising Standards Council of India, Guidelines for Responsible Labelling of Synthetically Generated Content in Advertising.
The lists are examples, not exhaustive rules; an ad outside them is judged by ASCI's three assessment questions.
What changed from the draft
We compared the May draft, as set out in ASCI's press release of 12 May, with the final text. The structure survived, but several things changed.
The risk labels went. The draft sorted uses into High Risk (prohibited), Medium Risk (label required) and Low Risk (no label). The final keeps the three groups but drops the risk names.
The fake doctor example went. The draft's prohibited list included “using AI to generate fictional authority figures with identifiable cues, such as an AI-generated fake doctor promoting a supplement”. The final list does not name it. Such an ad could still fall under the ban on fabricated endorsements and misleading claims, but the explicit example is gone.
Product-performance visuals were folded in. The draft required a label on synthetic visuals of product performance unless they replicated how the product actually performs. The final has no separate item; misleading exaggeration of results is in the prohibited list instead.
The events rule narrowed. The draft required a label for “realistic events, settings or situations entirely with AI”. The final requires one only where fabricated events or settings may affect consumers' understanding of the product.
Accessibility was added to the uses needing no label, and the final explicitly lets brands rely on labels supplied by advertising platforms.
ASCI has not published the consultation responses, so we cannot say who asked for which change.
What it means for PR and sponsored content
The guidelines apply to advertising, and ASCI's code treats paid content broadly: in August 2025 it added a clause requiring media companies to label paid posts on their own social accounts at the start, with words such as “Advertisement”, “Sponsored” or “Partnership”. What follows is our reading of how the new guidelines bear on communications work, not ASCI's.
An AI version of an executive needs a label. A video message from a chief executive's digital twin, or a cloned voice in a personalised campaign, falls in the mandatory list even when the executive has agreed to it. Companies experimenting with AI spokespeople for Indian audiences should plan for the label from the start. Listener reaction to AI versions of real hosts is covered in whether listeners would accept an AI podcast host.
Paid placement in AI answers has a prescribed label. The guidelines spell out the words a paid recommendation inside an AI assistant must carry, rather than leaving it to a general duty to disclose. Whether such placements are on sale is a separate question; we looked at that in whether you can pay to appear in a ChatGPT answer.
AI-written copy is not the issue; who is shown is. A sponsored article drafted with AI needs no AI label under these guidelines, though it still needs to be identified as paid. A sponsored piece illustrated with a synthetic “customer” or “expert” is a different matter: if it implies a real person endorsed the product, it is prohibited, label or not. How the United States approaches the same questions, without an AI-specific rule, is set out in whether there is an AI exemption from advertising law, and the EU's approach in whether the EU AI Act requires labelling AI-written coverage.
Methodology and limitations
This explainer is based on the final guidelines (PDF dated 17 September 2026), ASCI's press release of 29 September 2026 and ASCI's press release of 12 May 2026 announcing the draft, and ASCI's press release of 13 August 2025 on labelling media companies' paid social posts, all downloaded from ascionline.in and read in full on 5 October 2026, and on Business Standard's report of 30 September 2026 and The Hindu BusinessLine's report of 29 September 2026. Our draft-versus-final comparison uses the draft as summarised in ASCI's May press release; we did not compare against any longer draft document. We have not read the February 2026 amendments to India's IT Rules; the alignment is ASCI's statement. The effective date is our calculation from ASCI's “three months from the date of publication”, and depends on whether ASCI counts from the document date or the release date. ASCI's guidelines are self-regulatory; we have not assessed how they interact with Indian law, and nothing here is legal advice. The final section is our reading.
Conclusion
India's advertising self-regulator has chosen a test of effect rather than a rule about tools. AI that tidies an image or drafts copy needs no label; AI that stands in for a person, invents an event or sells a paid recommendation inside a chatbot does; and AI that fabricates an endorsement is banned whatever it says on the screen. The changes from the draft mostly narrowed and simplified the rules, including dropping the explicit fake-doctor example. For companies advertising in India, the practical tasks before late December are to find any synthetic people or voices in current campaigns, label them, and make sure no AI-generated figure implies a real endorsement.
Frequently Asked Questions
Does India require labels on AI-generated ads?
India's advertising self-regulator, ASCI, requires a label when synthetic audio or visual content could influence purchase decisions or mislead about a product without one, under guidelines released on 29 September 2026 and taking effect three months later. Routine edits, decorative elements and AI-written copy do not need a label.
When do the ASCI AI labelling guidelines take effect?
Three months from the date of publication. The document is dated 17 September 2026 and was released on 29 September 2026, so the guidelines take effect in late December 2026.
Do AI influencers need to be labelled in India?
Yes. The guidelines list synthetically generated influencers and ambassadors among the uses where a label is mandatory.
Does a cloned executive voice need a label if the executive agreed?
Yes, where it is used for personalised messaging. ASCI's guidelines require a label for replicating a real person's likeness or voice even with their consent.
What label does ASCI require for sponsored AI recommendations?
“Sponsored by [Brand]”. If an AI system recommends a product because the brand paid for or sponsored that recommendation, the recommendation must be identified that way.
Can a label make an AI-generated testimonial acceptable?
No. Fabricating endorsements or testimonials with AI is prohibited under the guidelines regardless of any label, as are misleading exaggerations, invented locations shown as real, and deepfakes or likenesses used without consent.
Sources
- trade body Advertising Standards Council of India, Guidelines for Responsible Labelling of Synthetically Generated Content in Advertising, dated 17 September 2026. Read 5 October 2026.
- trade body ASCI, ASCI Releases Guidelines for Responsible Labelling of Synthetically Generated Content in Advertising, press release, 29 September 2026. Read 5 October 2026.
- trade body ASCI, ASCI Releases Draft Guidelines for Responsible Labelling of AI-Generated Content in Advertising, press release, 12 May 2026. Read 5 October 2026.
- trade body ASCI, ASCI updates code, media companies must label paid posts on social handles, press release, 13 August 2025. Read 5 October 2026.
- press Business Standard, ASCI issues guidelines for labelling synthetically generated content in ads, 30 September 2026; The Hindu BusinessLine, ASCI releases guidelines for responsible labelling of AI-generated ad content, 29 September 2026. Read 5 October 2026.