This guide summarises ASCI's published Guidelines for Responsible Labelling of Synthetically Generated Content in Advertising and its 29 September 2026 press release. It is not legal advice. Confirm your own obligations against the guideline text, the full ASCI Code, or with counsel.
What Are ASCI's AI Labelling Guidelines in 2026?
The Advertising Standards Council of India's Guidelines for Responsible Labelling of Synthetically Generated Content in Advertising set out when Indian advertisers must disclose AI-generated content and which uses are not allowed at all. The guideline was signed by the Chairman of ASCI's Board of Governors on 17 September 2026, and ASCI's press release dated 29 September 2026 says ASCI "today released" it after "stakeholder consultations on the draft guidelines released in May."
The guideline sorts AI use into four buckets: prohibited content, mandatory labelling, no labelling required, and how to label. It states that "All advertisements containing SGC AI must comply with the entirety of the ASCI Code", and that its "focus is on managing consumer outcomes rather than regulating the technology." In practice that means the question is never "did we use AI?" but "would a consumer be misled about the product without being told?"
What Counts as Synthetically Generated Content Under ASCI in 2026?
ASCI defines synthetically generated content, or SGC, as "any audio, visual, or audio-visual content in an advertisement that is artificially or algorithmically originated, created, generated, modified, or materially altered using a computer resource, to appear real, authentic or true and depicts or portrays any individual or event in a manner that is likely to be understood, perceived or believed to be true or as indistinguishable from a natural person or a real-world event."
Two features of that definition matter. It covers audio, visual and audio-visual content, so text alone sits outside it, which is consistent with clause 3D exempting AI-written copy. And it is about content made "to appear real", which is why obvious fantasy is exempt under clause 3C. ASCI adds that SGC "is only considered misleading or harmful when it creates unfulfillable expectations, exploits vulnerable populations, depicts unsafe situations, or replicates a natural person's likeness without consent, or provides false or misleading information by way of depiction."
How Do You Decide Whether an Ad Needs an AI Label in 2026?
ASCI gives three assessment questions. The requirement "is based on the nature of communication and the risk of consumers being misled or harmed", and brands should ask whether SGC "is central to an ad's persuasive messaging or materially influences purchase decisions", whether its use "misleads consumers about the product's claims, performance, or benefits, if it is not labelled", and whether "the absence of a disclosure" would "create a false or misleading impression of claims, benefits or performance of the product or service."
| Decision in 2026 | ASCI clause | What falls here |
|---|---|---|
| Not allowed, even with a label | Clause 1, Prohibited Content | Fabricated endorsements or testimonials; exaggerated product results through visuals; non-existent locations presented as real; unauthorised copyrighted work, deepfakes, or likeness used without consent. |
| Needs a label | Clause 2, Mandatory Labelling Required | Synthetic influencers and ambassadors; replicated likeness or voice even with consent; fabricated events affecting understanding of product performance; products that do not yet exist; sound effects highly relevant to core features; paid AI product suggestions. |
| No label needed | Clause 3, No Labelling Required | Routine edits and colour correction; decorative backgrounds and ambient music; obvious fantastical effects; AI copy and administrative uses; accessibility such as subtitles and translations. |
| How to label | Clause 4, Disclosure Labels and Execution | Platform labels or brand labels such as "Audio/Video created using AI"; sponsored AI recommendations must say "Sponsored by [Brand]". |
Each bucket has its own detailed section in this guide. The section on AI label exemptions includes a fourteen-row decision table for common production situations. The virtual influencer and AI likeness section covers personas, voice cloning and consent. The sponsored AI recommendations section covers paid answers inside chatbots.
What Is Prohibited Even With an AI Label in 2026?
Clause 1 lists content that is illegal, infringes rights, makes misleading claims or violates the ASCI Code, and states plainly: "These will violate the ASCI code even if an AI label is used." This is the most important sentence in the guideline for anyone who hoped a disclosure would make risky creative acceptable.
- Fabricated endorsements: "using AI to create, alter or attribute an endorsement or testimonial in a manner that falsely represents that a real person has used, experienced, approved, recommended or endorsed a product or service when they have not."
- Exaggerated results: ASCI's example is "a weightloss product showing a before-after image of weight reduction from 100 kgs to 70 kgs through an AI generated image of a lean, toned body, when the visual does not represent actual or substantiated product results." Routine clause 3A edits are carved out.
- Non-existent places as real: "a travel agency depicting a hotel in a non-existent location or having non-existent facilities."
- Rights and consent: "Using unauthorised copyrighted work, deepfakes or a person's likeness without consent."
Which AI Uses Need a Mandatory Label in 2026?
Clause 2 says labelling "is mandatory in advertisements which contain SGC in a manner that materially influences consumer decisions, and the lack of a disclosure label would mislead consumers." ASCI lists six examples and states "Labelling is mandatory in all such cases to help consumers understand the nature of the representation."
| Clause | Mandatory-label case in 2026 | ASCI's example |
|---|---|---|
| 2A | Synthetic influencers and ambassadors | "Using synthetically generated influencers and ambassadors." |
| 2B | Replicated likeness or voice, even with consent | Personalised messages using "a digitally replicated version of a real person's face or voice, with that person's consent". |
| 2C | Fabricated events or settings | Where they "may impact consumer understanding of the product performance or service advertised". |
| 2D | Products that do not yet exist | "a 3D model of an unbuilt housing complex". |
| 2E | Sound effects relevant to core features | "audio quality in an ad for a headset". |
| 2F | Paid AI product suggestions | A chatbot recommending a moisturiser for Mumbai weather because the brand paid; label "Sponsored by [Brand]". |
Which AI Uses Need No Label in 2026?
Clause 3 says "No labelling is required when advertisements feature minor modifications or use of SGC in ways that have no material impact on a consumer's ability to make an informed choice." The five groups are minor enhancements, background and ambient elements, fantastical elements, administrative and text uses, and accessibility. This is deliberate: ASCI says the guidelines aim to ensure "transparency while avoiding consumer label fatigue."
For most creative teams this is the reassuring part. Colour correction, blemish removal, AI-written copy, decorative backgrounds, jingles and subtitles all sit here. The boundary is substance: clause 3A edits qualify only where they "do not alter the substance or core claims of the ad", and clause 3B elements only where they are "unrelated to the product's actual capabilities or promise."
How Do You Decide Whether an AI Edit Needs a Label?
Run the three questions ASCI sets out under "Assessment for labelling". The guideline says the requirement "is based on the nature of communication and the risk of consumers being misled or harmed", and asks brands to consider whether SGC "is central to an ad's persuasive messaging or materially influences purchase decisions", whether its use "misleads consumers about the product's claims, performance, or benefits, if it is not labelled", and whether "the absence of a disclosure" would "create a false or misleading impression of claims, benefits or performance".
The decision table below applies those questions to common production situations. Where a row says "depends", the guideline's own wording shows which way it tips.
| AI use in your ad | Label? | ASCI's reason |
|---|---|---|
| Colour correction, light adjustment, noise reduction on product photos | No | Clause 3A minor enhancement, provided it does "not alter the substance or core claims". |
| Standard blemish removal on a model | No | Clause 3A lists "standard blemish removal". |
| Retouching that shows results the product cannot deliver | Not allowed | Clause 1B prohibits exaggerating product results through visual representations, label or not. |
| An AI-generated decorative city skyline behind the product | No | Clause 3B, "purely decorative", unrelated to the product's capabilities. |
| AI-generated ambient music or a jingle | No | Clause 3B lists "ambient music, jingles". |
| AI sound effects demonstrating a headset's audio quality | Yes | Clause 2E: sound effects "highly relevant to the product's core features". |
| A fairy fixing a product with magic dust | No | Clause 3C, obvious and unrealistic. |
| A realistic AI scene of the product performing in an event that never happened | Yes | Clause 2C, fabricated events that "may impact consumer understanding of the product performance". |
| A 3D render of an unbuilt housing project | Yes | Clause 2D, "a product that does not currently exist". |
| A hotel ad showing a location or facility that does not exist as real | Not allowed | Clause 1C prohibits depicting non-existent locations as real. |
| AI-written ad copy | No | Clause 3D, "Generating or enhancing advertising copy". |
| AI subtitles, captions or translations | No | Clause 3E, accessibility. |
| Voice modulation that does not impersonate anyone | No | Clause 3A example: "voice modulations which do not purport to impersonate or mislead". |
| Voice cloning of a real person for personalised messages | Yes | Clause 2B, "even with their consent". |
Why Is "No Label Required" Not the Same as "Anything Goes"?
Because the exemptions are bounded by the rest of the guideline and the whole ASCI Code. Clause 3A's routine edits only qualify where they "do not alter the substance or core claims of the ad". Clause 3B's decorative elements only qualify where they are "unrelated to the product's actual capabilities or promise". And clause 1B notes that exaggerated product results through visuals are prohibited, while clarifying it "does not include routine edits and enhancements as laid out under Clause 3A."
ASCI's own example of the line is useful: "a weightloss product showing a before-after image of weight reduction from 100 kgs to 70 kgs through an AI generated image of a lean, toned body, when the visual does not represent actual or substantiated product results." The same tools that perform a permitted colour correction can produce that prohibited image. The test is what the consumer is led to believe, not which software was used.
If the AI changed how the ad looks, it is usually exempt. If the AI changed what the ad claims, shows the product doing, or shows a person saying, it needs a label or is not allowed at all.
Which Label Should You Use When One Is Required?
Clause 4 gives brands latitude. They "may use appropriate labels, either those provided by ad platforms or their own labels such as but not limited to 'Audio/Video created using AI' or 'Audio/Video enhanced using AI.'" They "may use alternative labels that accurately inform the consumer." The one fixed wording is for sponsored AI recommendations, which must say "Sponsored by [Brand]", covered in the section on sponsored AI recommendations. For AI personas and replicated voices, see the virtual influencer and AI likeness section.
What Are the Common Mistakes About AI Label Exemptions?
- Labelling everything. ASCI explicitly wants to avoid "consumer label fatigue". Over-labelling dilutes the labels that matter.
- Treating retouching as always exempt. Retouching that shows unachievable results falls under clause 1B and is prohibited.
- Assuming sound is always ambient. Sound that demonstrates a core feature, such as audio quality, needs a label under clause 2E.
- Calling a realistic scene "fantastical". Clause 3C covers effects audiences recognise as not depicting reality.
- Forgetting AI copy still has to be true. Clause 3D exempts it from labelling, not from the ASCI Code.
- Timing to the signing date. The three months run from publication; ASCI released the guideline on 29 September 2026.
How Should Brands Label AI Content in 2026?
Clause 4 gives brands room. They "may use appropriate labels, either those provided by ad platforms or their own labels such as but not limited to 'Audio/Video created using AI' or 'Audio/Video enhanced using AI.'" They "may use alternative labels that accurately inform the consumer." The one fixed form is for sponsored AI recommendations: "Where AI recommends a product that is sponsored, the disclosure should clearly state 'Sponsored by [Brand].'"
The guideline also says "Disclaimers should follow the ASCI Code on disclaimer guidelines where applicable", and repeats the limit that defines the whole document: "The use of SGC in some cases may be considered misleading or objectionable regardless of labels, if the end effect is likely to mislead or harm the consumer."
In the press release, Manisha Kapoor, Secretary General and CEO of ASCI, said the guidelines clarify "where the mere use of an AI label may not make an otherwise misleading advertisement acceptable", and that "The responsibility remains with advertisers to ensure that the end communication is honest, transparent and compliant with the ASCI Code."
When Do the ASCI AI Labelling Guidelines Take Effect in 2026?
The guideline text says: "These Guidelines shall come into effect on the expiration of 3 months from the date of their publication." The document is signed by the Chairman of ASCI's Board of Governors and dated 17 September 2026. ASCI's press release, datelined Mumbai, 29 September 2026, says ASCI "today released" the guidelines and that they "will come into effect three months from the date of publication."
ASCI does not itself state the calendar date on which the guidelines take effect. Two dates appear in its documents, and they do different jobs. The 17 September 2026 date is when the guideline was signed. The 29 September 2026 date is when ASCI says it released it. If 29 September is treated as the date of publication, three months runs to 29 December 2026. Plan to that date at the latest, and confirm the effective date with ASCI before relying on it in a contract or compliance calendar.
How Should Indian Brands and Agencies Prepare in 2026?
- Inventory live and planned creative that uses AI for audio, visuals or video, including personalised and programmatic variants.
- Sort each item into ASCI's four buckets using the three assessment questions. Most will be clause 3 exempt.
- Remove anything prohibited now, rather than labelling it: fabricated testimonials, exaggerated before-and-after imagery, non-existent locations shown as real, unconsented likeness.
- Choose a house label for mandatory cases and apply it consistently across platforms.
- Write the sponsored-recommendation wording into AI ad contracts, since that label is fixed.
- Add the labelling decision to creative briefs, so it is made before production, not at review.
- Brief creators and agencies, especially on the consent-versus-disclosure distinction for replicated voices.
What Does the Guideline Not Settle in 2026?
- The exact effective calendar date. ASCI gives the release date and the three-month rule, not the resulting day.
- Label placement, size and duration, beyond following the ASCI Code's disclaimer guidelines.
- Who applies a label when a platform and a brand are both involved.
- Separate penalties. All SGC advertising "will continue to be assessed against the ASCI Code in its entirety", per the press release.
What Does ASCI Say About Sponsored AI Chatbot Recommendations?
ASCI's 2026 guideline on synthetically generated content lists paid AI product suggestions as a case where labelling is mandatory. Clause 2F covers "Using AI for paid or sponsored product suggestions, which must specifically be labelled as 'sponsored by'." Clause 4 then fixes the wording: "Where AI recommends a product that is sponsored, the disclosure should clearly state 'Sponsored by [Brand].'" It is the first Indian advertising code to address paid answers inside AI assistants directly.
The guideline illustrates the rule with a scenario every marketer will recognise. A consumer asks an AI chatbot, "What is a good moisturiser for Mumbai weather?" The AI recommends a specific brand or product because the brand has paid for or sponsored that recommendation. In ASCI's words, "the recommendation should be clearly identified as 'Sponsored by [Brand]', so that the consumer understands that the recommendation is commercially influenced."
| Element | ASCI's wording in 2026 | What it means for a brand |
|---|---|---|
| Category | Clause 2, "Mandatory Labelling Required" | Not optional, not a best practice. It sits with the mandatory cases. |
| The trigger | "Using AI for paid or sponsored product suggestions" | Payment or sponsorship behind the recommendation is what triggers it. |
| The label | "Sponsored by [Brand]" | A specific form of words, unlike the other mandatory cases. |
| The purpose | "so that the consumer understands that the recommendation is commercially influenced" | The test is consumer understanding, not technical compliance. |
Why Does This Rule Matter More Than It First Appears?
Because a recommendation inside a conversation looks like advice, not like an advertisement. A banner or a sponsored search result has a familiar visual grammar that tells people it is paid. An answer from an assistant arrives in the same voice as every other answer, which is exactly why ASCI singled it out and why it specified a label rather than leaving the wording to the advertiser.
For Indian brands the timing is direct. Paid placements inside AI assistants are now a real buying option, as our ChatGPT Ads API guide and the overview of advertising inside AI in India set out. Any brand planning to buy those placements for Indian consumers now has an ASCI code requirement to account for alongside whatever the platform itself does.
- The label wording is fixed. Elsewhere in the guideline brands may choose their own label. Here ASCI names the words.
- The illustration is consumer advice. A moisturiser question is the kind of everyday query where a paid answer would be mistaken for a neutral one.
- It applies to the recommendation itself. The clause is about the AI's product suggestion, not only about a separate ad unit next to it.
How Does the Sponsored Label Differ From Other AI Labels?
Most of ASCI's 2026 guideline lets brands pick appropriate wording. Clause 4 says brands "may use appropriate labels, either those provided by ad platforms or their own labels such as but not limited to 'Audio/Video created using AI' or 'Audio/Video enhanced using AI.'" It adds that "Brands may use alternative labels that accurately inform the consumer." The sponsored-recommendation case is treated differently: the disclosure "should clearly state 'Sponsored by [Brand].'"
| Situation | What the label discloses | Wording flexibility in 2026 |
|---|---|---|
| AI-created or AI-enhanced audio or video in an ad (a mandatory case) | That the content was created or enhanced using AI | Flexible. Platform labels, ASCI's examples, or alternatives "that accurately inform the consumer". |
| A paid or sponsored product suggestion by AI | That the recommendation is commercially influenced | Specified: "Sponsored by [Brand]". |
The two labels disclose different things. An "Audio/Video created using AI" label tells a consumer how content was made. A "Sponsored by [Brand]" label tells them why they are seeing a recommendation. A sponsored AI recommendation that also uses AI-generated video could plausibly need both, because each answers a separate question the consumer would otherwise get wrong. The guideline does not address combined labels explicitly, so treat that as a judgement to confirm, not a rule.
What Does the Guideline Not Say About Sponsored AI Answers?
The clause is short, and several practical questions sit outside it. Planning around them honestly is better than assuming an answer ASCI has not given.
- Where the label must appear in the interface, its size, or its placement relative to the answer. The guideline says the disclosure should "clearly state" the wording, and that disclaimers "should follow the ASCI Code on disclaimer guidelines where applicable."
- Who applies the label, the brand or the platform. Clause 4 allows "labels provided by ad platforms", but the guideline does not allocate responsibility between them.
- Any threshold of payment or relationship below which the label is not needed. The clause says "paid or sponsored".
- Enforcement or penalties. The guideline states that "Advertisements containing SGC are subject to compliance with the entire ASCI Code." It does not set out separate penalties.
How Should Brands Buying AI Ads Prepare?
Treat the sponsored label as a contractual and creative requirement now, rather than as a platform feature you will check later. The practical work is making sure that whatever surface carries your paid AI recommendations to Indian consumers can show the ASCI wording, and that someone on your side has confirmed it does.
- List every AI surface you pay for that could recommend a product to an Indian consumer, including tests and pilots.
- Check how each one labels paid placements today, and whether that label reads as "Sponsored by [Brand]" or something weaker.
- Put the label requirement in your insertion orders or platform briefs so the obligation is written down, not assumed.
- Brief your agency and creative team that this is a mandatory case with fixed wording, not one of the flexible labels.
- Keep the rest of the ASCI Code in view. A labelled sponsored answer still has to be truthful. The label does not cure a misleading claim.
ASCI is explicit elsewhere in the same guideline that some content "will violate the ASCI code even if an AI label is used." A sponsored recommendation for a product that cannot deliver what the answer implies is still a misleading advertisement. The label tells consumers who paid; it does not make an untrue recommendation acceptable.
What Are the Common Mistakes With Sponsored AI Recommendations?
- Using a softer label. "Promoted" or "partner" may not meet a clause that names "Sponsored by [Brand]".
- Assuming the platform handles it. The guideline allows platform labels but does not say the platform carries the obligation.
- Treating it as best practice. It is listed under Mandatory Labelling Required.
- Confusing the two label types. "Created using AI" and "Sponsored by [Brand]" disclose different facts.
- Assuming the effective date is the signing date. The guideline was signed on 17 September and released on 29 September 2026; the three months run from publication.
- Relying on the label to fix a weak claim. The whole ASCI Code still applies.
What Does ASCI Say About Virtual Influencers and AI Likeness?
ASCI's 2026 guideline on synthetically generated content puts two influencer-adjacent uses straight into the mandatory labelling category. Clause 2A covers "Using synthetically generated influencers and ambassadors." Clause 2B covers "Replicating a real person's likeness or voice even with their consent, for personalised messaging." Both sit under the heading "Mandatory Labelling Required", and ASCI states that "Labelling is mandatory in all such cases to help consumers understand the nature of the representation."
A third rule sits higher up, in the prohibited category. Clause 1D lists "Using unauthorised copyrighted work, deepfakes or a person's likeness without consent" as content that violates the ASCI Code "even if an AI label is used." Taken together, the three clauses give a clear ladder: an AI likeness used without consent is prohibited, one used with consent still needs a label, and a fully synthetic influencer needs a label too.
| Use in 2026 | ASCI category | ASCI's wording |
|---|---|---|
| A synthetic influencer or brand ambassador | Mandatory label | "Using synthetically generated influencers and ambassadors." |
| A real person's face or voice replicated, with consent, for personalised messages | Mandatory label | "Replicating a real person's likeness or voice even with their consent, for personalised messaging." |
| A real person's likeness used without consent, or a deepfake | Prohibited, even with a label | "Using unauthorised copyrighted work, deepfakes or a person's likeness without consent." |
| AI used to fabricate a testimonial or endorsement | Prohibited, even with a label | "Fabricating endorsements or testimonials". |
Why Does Consent Not Remove the Need to Label?
Because consent and disclosure protect different people. Consent protects the person whose face or voice is being replicated. Disclosure protects the consumer receiving the message, who would otherwise believe they were hearing directly from that person. ASCI's clause 2B is deliberately worded "even with their consent", which closes off the assumption that a signed release is the only obligation.
ASCI's own example makes the point concrete: "A brand sends personalised messages to consumers featuring a digitally replicated version of a real person's face or voice, with that person's consent, without disclosing that the content has been digitally generated or replicated." That is presented as the scenario the label exists to prevent. A celebrity voice greeting each customer by name is exactly what clause 2B describes, and under the 2026 guideline it needs a label.
Consent answers "may we use this person's likeness?" The label answers "does the consumer know this is generated?" ASCI requires both for replicated likeness in personalised messaging. Getting one does not satisfy the other.
How Does the Fabricated Testimonial Rule Apply to Influencer Work?
It is the line a label cannot cross. Clause 1A prohibits "Fabricating endorsements or testimonials i.e. using AI to create, alter or attribute an endorsement or testimonial in a manner that falsely represents that a real person has used, experienced, approved, recommended or endorsed a product or service when they have not." ASCI states this kind of content will violate the Code "even if an AI label is used."
For influencer marketing, that rules out several tempting shortcuts: generating a creator saying something they never said, altering a genuine review so it reads more favourably, or attributing an endorsement to a real person who did not give it. Each is prohibited regardless of disclosure. A synthetic influencer is a different case, because it is not presented as a real person who used the product, and ASCI treats it as a mandatory-label case rather than a prohibited one.
What Should Indian Brands Using Virtual Influencers Do?
Treat every synthetic persona and every replicated likeness as labelled by default, and build the label into the creative, not as an afterthought in a caption. The guideline lets brands choose the wording for these cases: clause 4 permits platform labels or the brand's own, "such as but not limited to 'Audio/Video created using AI' or 'Audio/Video enhanced using AI'", and alternatives "that accurately inform the consumer".
- Inventory your AI personas. Any synthetic influencer, mascot presented as a person, or AI ambassador in Indian advertising.
- Inventory replicated likeness. Any campaign using a real person's generated face or voice, including personalised video or voice messages.
- Check consent paperwork for every replicated likeness. Without consent, clause 1D makes it prohibited.
- Choose a label that accurately tells consumers the content is generated or replicated, and apply it consistently.
- Audit testimonials and reviews for any AI alteration or attribution. Clause 1A makes those prohibited even with a label.
- Brief creators and agencies on the difference between a labelled synthetic persona and a fabricated endorsement.
For the wider context of how influencer partnerships are structured in India, see our influencer marketing guide and our look at creator-led D2C brands in India. Content provenance standards such as those in our SynthID and C2PA guide are a separate, technical layer; ASCI's guideline is about what the consumer is told.
What Does the Guideline Not Settle for Influencer Marketing?
- Label placement and duration within a video or a post. ASCI says disclaimers "should follow the ASCI Code on disclaimer guidelines where applicable."
- Whether a recurring synthetic persona needs the label on every post or can rely on profile-level disclosure. The guideline does not say.
- Lightly AI-edited footage of a real creator. Routine edits such as colour correction and "voice modulations which do not purport to impersonate or mislead" are listed under no label required in clause 3A, but the line between enhancement and replication is a judgement call.
- Separate penalties. The guideline states that all SGC advertising is subject to the whole ASCI Code; it does not set out a distinct penalty regime.
What Are the Common Mistakes With AI Influencers?
- Assuming consent is enough. Clause 2B requires a label "even with their consent".
- Presenting a synthetic persona as a real person. It is a mandatory-label case, and drifting into a false endorsement makes it prohibited.
- AI-polishing real reviews. Altering a testimonial so it misrepresents what a real person said is prohibited under clause 1A.
- Treating a deepfake as acceptable if labelled. Clause 1D puts deepfakes and unconsented likeness in the prohibited list, label or not.
- Hiding the label in a caption nobody reads. Follow the ASCI Code's disclaimer guidelines.
- Planning to the signing date. The three months run from publication; ASCI released the guideline on 29 September 2026.
What Are the Common Mistakes in 2026?
- Believing a label fixes misleading content. Clause 1 content violates the Code "even if an AI label is used".
- Labelling every AI touch. ASCI wants to avoid label fatigue; most production use is exempt.
- Treating consent as disclosure. Replicated likeness needs a label even with consent.
- Using soft wording for sponsored AI answers. The label is fixed: "Sponsored by [Brand]".
- Counting three months from the signing date. The clock runs from publication, and ASCI released the guideline on 29 September 2026.
- Assuming text is covered. The SGC definition covers audio, visual and audio-visual content; AI copy is exempt but must still be truthful.
Key Takeaways for 2026
- ASCI's SGC labelling guideline was signed on 17 September 2026 and released on 29 September 2026, after consultation on a May draft.
- It takes effect three months from publication: 29 December 2026 if the release date is the publication date. ASCI does not state the calendar date, so confirm it.
- Four buckets: prohibited even with a label, mandatory label, no label, and how to label.
- Prohibited: fabricated testimonials, exaggerated results, non-existent places as real, deepfakes and unconsented likeness.
- Mandatory label: synthetic influencers, consented likeness replicas, fabricated events, unbuilt products, core-feature sound, paid AI suggestions.
- No label: routine edits, decorative elements, obvious fantasy, AI copy, accessibility.
- Sponsored AI recommendations must say "Sponsored by [Brand]"; other labels can be any accurate wording.
- The whole ASCI Code still applies to every ad that uses AI.
Distk helps Indian brands and agencies build AI-assisted advertising with the disclosure decision made at the brief stage, across creative, influencer and paid AI placements. If you are planning campaigns that will run after the guideline takes effect, that review is where we start.
Sources
- ASCI, Guidelines for Responsible Labelling of Synthetically Generated Content in Advertising, signed 17 September 2026. All quoted rules, definitions and examples come from this document.
- ASCI press release, Mumbai, 29 September 2026, for the release date and the effective-date wording.
- The Advertising Standards Council of India, ascionline.in.