India’s advertising self-regulator ASCI issued final guidance on 29 September 2026 for labels on synthetic content in ads. AI advertising now faces a consumer-impact test: material synthetic influencers, staged product demonstrations and realistic fabricated scenes may need disclosure, while misleading endorsements and non-consensual likenesses remain unacceptable even with a label. The guidance takes effect three months after publication.
Key takeaways
- ASCI asks advertisers to judge whether synthetic content changes a consumer’s understanding or decision, rather than label every AI-assisted edit.
- A disclosure does not cure a deceptive claim, fabricated endorsement, or use of someone’s likeness without consent.
- Synthetic influencers, personalised voice or likeness replicas made with consent, and realistic fabricated product scenes are among the uses that require a label.
- Routine colour correction, background decoration, copy generation and accessibility work are among uses for which ASCI says a label is unnecessary.
- The industry guidance is due to take effect three months after its 29 September 2026 publication.
What changed in AI advertising in India?
The Advertising Standards Council of India, or ASCI, has moved from a May consultation draft to final guidance on synthetically generated content in advertising. Its central question is whether a synthetic image, voice, scene or demonstration materially affects what a consumer believes about a product or service. The council says advertisers should disclose such content where the absence of a label could mislead a buyer. It also names uses that remain prohibited regardless of labelling.
The distinction matters because generative tools touch many stages of advertising. An editor might correct lighting in a product photo, generate draft copy or create captions. Another campaign might create a fictional person who appears to endorse a product or show a building that does not exist as though it were ready to occupy. Treating those jobs as equivalent would blur the difference between routine production and a representation that could change a purchase decision.
ASCI’s release, dated 29 September, says the guidance will take effect three months from publication. Independent coverage by exchange4media, Business Standard Hindi and MediaNews4u also reported the final announcement on that date. These are separate publisher reports on the same ASCI action, not separate rules.
The scope should be described accurately. ASCI is the advertising industry’s self-regulatory council, not Parliament or a new statutory regulator. Its code and guidelines govern how the council assesses ads and works alongside applicable law and platform policies. The announcement does not itself specify a new government fine, a universal watermark requirement or an immediate ban on every AI-made ad. Brands should still evaluate other legal obligations separately where they apply.
Which synthetic ads are prohibited even with a label?
ASCI says an AI disclosure cannot legitimise an ad that is illegal, infringes rights, makes misleading claims or breaches its advertising code. It lists fabricated endorsements or testimonials; synthetic depictions that exaggerate product results; non-existent settings shown as real in a misleading way; unauthorised copyrighted material; and deepfakes or a person’s likeness used without consent. These are not simply items on a labelling checklist. If the underlying representation is deceptive or unauthorised, the creative must change.
Consider a skincare campaign. A brand could put “AI-generated” beneath a before-and-after visual, but that notice would not fix a fabricated result that makes the product appear more effective than evidence supports. The same principle applies to a travel ad that presents an invented hotel facility as a real amenity. The label tells a viewer how an image was made; it does not make an untrue product claim true.
Consent is a separate boundary. A campaign that digitally places a recognisable person’s face or voice into an endorsement without permission risks misleading viewers and infringing the person’s rights. ASCI places use of an individual’s likeness without consent among prohibited uses. Even a consensual voice or likeness replica may require disclosure if it is used for personalised messaging, because viewers could otherwise mistake a synthetic message for a natural recording.
This is the most useful operational distinction for an agency: ask first whether the ad itself can run, then ask whether the permitted ad needs a label. Reversing that order encourages a team to treat a disclosure as a safety net for a creative claim it has not substantiated. ASCI’s answer is clear: honesty and consent remain the foundation.
When does an AI advertising label become mandatory?
ASCI says disclosure is required when synthetic content materially influences consumer decisions and an undisclosed synthetic element could mislead. The examples in its release include virtual influencers or ambassadors; replication of a real person’s likeness or voice with consent for personalised messages; fabricated events or settings that shape what buyers understand; demonstrations of products that do not yet exist; and synthetic sounds tied closely to a product’s core feature.
The sound example shows why the test is about effect, not just format. In a headphone ad, AI-generated audio that purports to demonstrate a product’s signature performance may give a listener a false impression of what they will hear. By contrast, an ambient sound bed that does not claim to demonstrate a feature may have a different impact. The advertiser needs to evaluate the actual claim the creative communicates, including what is implied rather than only what is written.
ASCI also calls out paid or sponsored AI-generated product suggestions. Its release says those suggestions must specifically carry a “Sponsored by [Brand]” disclosure. That instruction targets a familiar advertising problem in a new interface: a recommendation can look like neutral advice even when a brand paid to place it. The source of the recommendation matters to the person deciding whether to trust it.
These examples do not create a safe list in which anything unnamed is automatically acceptable. A synthetic ad can still mislead in ways that a short list does not anticipate. For each campaign, a reviewer should ask what an ordinary viewer is likely to conclude about the product, endorser, place or event, and whether that conclusion would differ if the viewer knew the content was synthetic.
What AI uses do not need an ASCI disclosure?
ASCI says routine editing and low-impact production work do not require a label under these guidelines when they do not materially affect informed choice. It lists colour correction, light adjustment, noise reduction, minor blemish removal, decorative or ambient elements, and obvious fantasy effects. It also lists administrative and text-based uses such as generating advertising copy, plus accessibility work including subtitles, translations and accurate audio descriptions.
That exemption is easy to oversimplify. A generic “we used AI for copy” disclosure is not required by this ASCI announcement, but the copy itself must still meet the normal standards for truth and fairness. Likewise, “minor blemish removal” should not be used as a label for changing the apparent outcome of a beauty product. The deciding factor is whether the edit changes a consumer’s understanding in a material way, not which software produced the pixels.
The most practical way to apply this is to document the creative process. Record whether the product shown exists, whether a person consented, which visual or audio elements were synthetic, and what evidence supports the performance claim. That record helps a legal or brand reviewer make a consistent decision before an ad goes live. It also helps an agency revise a creative if a platform or ASCI later questions it.
| Ad element | ASCI treatment | Key question |
|---|---|---|
| Fabricated endorsement or non-consensual likeness | Prohibited | Is the claim truthful and authorised? |
| Virtual influencer or realistic fabricated product scene | Disclosure required when material | Could it change a consumer decision? |
| Routine lighting correction or accurate subtitles | No ASCI label required | Does it leave product understanding unchanged? |
How should brands and creators prepare?
First, separate permissions from disclosure. A signed approval to use a creator’s face or voice addresses consent, but it does not eliminate the need to label a realistic AI replica used for personalised messaging. Conversely, a prominent AI label does not grant permission to use a person’s likeness. Agencies should therefore keep a documented consent trail and a separate disclosure decision for each creative asset.
Second, review the product claim at the level a consumer experiences it. If a synthetic visual makes a property appear finished, a cosmetic result appear stronger, or a device sound clearer, the ad may communicate a claim even without explicit words. Ask the product owner to show the evidence for that impression. If the depiction cannot be supported, the appropriate change is to remove or correct the depiction, not simply add a small note.
Third, make disclosures understandable in the place where a viewer encounters the creative. ASCI’s release gives examples such as “Audio/Video created using AI” or “Audio/Video enhanced using AI” and allows other labels that accurately convey the nature of the content. Platform-provided labels can be used, but the advertiser remains responsible for the final communication. A disclosure that is technically present yet impossible to notice would defeat the purpose of transparency.
Fourth, prepare for distribution across formats. A campaign may be cropped for a mobile feed, adapted to an influencer post, translated, or turned into a short video. A label placed only in a master file can disappear in one of those versions. Teams should check the live render of each major placement and keep the consent and claim records tied to the asset versions that actually run.
This is timely for the Indian advertising technology market. Our coverage of Google’s vertical-video buying changes shows how one creative can travel through several inventory formats, while Meta’s small-business AI workflows show why more teams can generate campaign assets quickly. ASCI’s guidance shifts attention from generation speed to what the audience is led to believe.
What is the deadline and what remains open?
ASCI’s release says the guidance takes effect three months after 29 September 2026. That provides a transition period for brands, agencies and platforms to review templates, approvals and disclosure practices. The press release does not quantify how many ads will need labels, and it does not supply new enforcement statistics for this specific guidance. Any prediction about compliance rates or financial penalties would require a separate source.
There is also a difference between the May draft and the September final release. The earlier document helped frame the consultation, but a live campaign should be evaluated against the final guidance and the wider ASCI Code. Where an example or wording differs, the current final source has priority. For a complex campaign, a brand may still need specific professional advice on copyright, publicity rights, consumer law and platform rules beyond this self-regulatory announcement.
The practical result is a cleaner review sequence: verify that the ad is truthful and authorised; test whether a synthetic element materially changes what a consumer understands; disclose it if the guidance requires; and check the finished placement. The sequence protects consumers without treating every use of AI as inherently deceptive. It also makes the burden clear: the advertiser is accountable for the claim people see.
AI advertising guidelines: frequently asked questions
Do all AI-generated ads in India need a label?
No. ASCI says a label is required where synthetic content materially influences consumer understanding or decisions and non-disclosure could mislead. Routine, low-impact edits and some administrative uses do not require a label under its guidance.
Can a label make a deepfake endorsement acceptable?
No. ASCI lists non-consensual likenesses and fabricated endorsements among prohibited uses. An AI disclosure does not cure a misleading or rights-infringing ad.
When will ASCI’s final guidance apply?
ASCI published the final announcement on 29 September 2026 and says the guidance comes into effect three months from that date. Advertisers should use the transition period to review their assets and approval process.
Does consent remove the need to disclose a synthetic voice?
Not always. ASCI names replication of a real person’s likeness or voice with consent for personalised messaging among cases where a label may be required because the synthetic nature could matter to the viewer.
Reporting basis: ASCI’s 29 September 2026 press release is the primary source. Its announcement and timing were corroborated with exchange4media, Business Standard Hindi and MediaNews4u. The workflow examples above are Lapaas Voice explanations of the ASCI framework, not new rules or legal advice.
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
