India’s AI ad labelling rules: what brands must disclose

Nabarun Chakraborty·Oct 6, 2026·6 min read

India’s AI ad labelling rules: what brands must disclose

The Advertising Standards Council of India (ASCI) on September 29 released guidelines requiring advertisers to label artificial intelligence (AI)-generated images, video and audio in ads when the content “materially influences consumer decisions and the absence of disclosure could mislead consumers.”

The guidelines take effect three months after publication. They apply to advertisers across TV, print, outdoor and online ads, while the Information Technology (IT) Rules bind online platforms.

What must brands label?

ASCI requires labels for:

“Use of synthetically generated influencers and ambassadors”
“Replication of a real person’s likeness or voice with consent for personalised messaging”
“Fabricated events or settings that may affect consumers’ understanding of the advertised product or service”
“Demonstrations regarding products that do not currently exist”
“SGC-generated sound effects that are highly relevant to a product’s core features such as headphones”
“Paid or sponsored AI-generated product suggestions, which must specifically be labelled ‘Sponsored by [Brand]'”

Brands may use “Audio/Video created using AI” or “Audio/Video enhanced using AI,” “including labels provided by advertising platforms,” or other labels that accurately describe the content.

What ASCI bans regardless of a label

“Fabricating endorsements or testimonials”
“Exaggerating product results or features in misleading ways through synthetic representations”
“Representing non-existent locations or settings as real such as to mislead consumers”
“Using unauthorised copyrighted work”
“Using deepfakes or an individual’s likeness without consent”

ASCI’s guidelines state that the mere use of an AI label may not make an otherwise misleading advertisement acceptable. “The responsibility remains with advertisers to ensure that the end communication is honest, transparent and compliant with the ASCI Code,” ASCI Secretary General and Chief Executive Officer (CEO) Manisha Kapoor said.

What does not need a label?

“Routine editing, such as colour correction, light adjustment, noise reduction and minor blemish removal”
“Decorative or ambient elements”
“Obvious fantastical effects”
“Administrative and text-based uses, such as generating advertising copy”
“Accessibility applications, including subtitles, translations and accurate audio descriptions”

How do the AI ad labelling rules differ from the IT Rules?

Both frameworks cover audio, visual and audio-visual content “artificially or algorithmically” created or altered “using a computer resource.” The content must appear “real, authentic or true” and show a person or event in a way people could mistake for the real thing.

Who is responsible: ASCI’s guidelines apply to advertisers across TV, print, outdoor and online ads. The IT Rules bind online platforms.
When a label is triggered: ASCI asks whether the content could mislead a buyer. The IT Rules ask only whether content is AI-generated.
What is exempt: Both exempt routine edits and accessibility uses, but ASCI also exempts “decorative or ambient elements,” which the IT Rules do not exempt by name.
Whether a platform label is enough: ASCI lets brands use “labels provided by advertising platforms,” but does not say whether a label a platform applies on its own under the IT Rules meets ASCI’s requirement.

Under Rule 3(3), platforms that offer AI tools must label AI-generated content made with those tools, and embed metadata, hidden data recording where the file came from, where technically feasible. Under Rule 4(1A), large platforms with over 50 lakh users in India, such as Instagram and YouTube, must ask users to declare AI-generated content before upload, verify it and label it.

The IT Rules require platforms to label such content. An AI ad on Instagram or YouTube may carry a platform label even when ASCI does not require one from the brand. The IT Rules do not say whether platforms’ labelling duty covers ads placed through ad managers.

Does a promoted AI music video need a label?

Saregama Managing Director Vikram Mehra said on the company’s April-June quarter earnings call, “The cost of a music video, if it is done through AI can be Rs. 70,000 also.” Saregama owns the audio rights to much of its older film catalogue but not the original video rights, and uses AI to create new videos for those songs. “Most of our catalogue, we are using AI to do fresh videos,” Sanjiv Goenka, Chairperson of Saregama’s parent RPSG Group, said in September 2025.

MediaNama founder Nikhil Pahwa asked whether a studio paying to promote such a video would count as an advertisement and therefore need an AI label under ASCI’s guidelines. The answer depends on three things.

First, what the video shows. Abstract or nature visuals do not show a person or event that looks real, so they likely fall outside both definitions.

Second, whether it recreates a real person. A video recreating a singer or actor falls within both definitions. If posted as a regular video, YouTube and Instagram must label it under the IT Rules once the uploader declares it or the platform’s checks confirm it. ASCI’s guidelines raise the separate question of whether Saregama must also label the promoted version.

Third, what the ad is selling. In a promoted music video, the video itself is the product. ASCI asks whether AI content could mislead a buyer about the product, which is harder to judge when the AI content and the product are the same thing.

Does an ad stitched from AI clips need one label or many?

“Videos are generated typically a 2-4 seconds at a time, so a complete ad clip of 26 seconds involves 6-12 videos generated separately, and stuck together,” Pahwa wrote in his newsletter Reasoned, after a GrowthX event on making ads with AI.

Newsletter

The corridor, every morning.

Funding rounds and cross-border capital moves, one email, five minutes.

Clips can drift from one another. “The colour of a shirt might differ slightly, the lighting might change,” Pahwa wrote. His own attempt produced a person with four fingers and a laptop with the screen on the wrong side of the lid.

He estimated AI ads cost “about Rs. 4000-5000-10,000 per video,” and said “undoing or fixing this is expensive.”

One AI model refused to generate video from his photo “because it does not generate videos on real people,” while Google’s model accepted it.

An ad assembled from a dozen AI clips, or one mixing real footage with AI shots, could combine background elements that need no label with elements that do, such as an AI-generated spokesperson. ASCI’s guidelines do not say whether brands should label the full ad or only specific parts.

What about AI chatbots?

ASCI places the labelling duty on advertisers, but does not say how an advertiser should label a paid product suggestion that the AI system itself generates. OpenAI launched ads inside ChatGPT in India and says they appear separately from its answers and carry a label.

Under Rule 4(3), the IT Rules require large platforms to mark paid content as “advertised, marketed, sponsored,” but do not say whether that covers AI chatbots.

Could platform labels change again?

ASCI sets no size or duration for labels. The Ministry of Electronics and Information Technology’s (MeitY) requirements have shifted across versions:

October 2025 draft: Labels had to cover at least 10% of the screen.
February 2026 rules: MeitY dropped the figure and asked only for “prominent” labels.
April 2026 proposal: MeitY proposed labels that stay “continuous and clearly visible” for the full length of a video.

If MeitY makes the continuous label requirement final, AI content on platforms, including ads if the rules cover them, would carry an on-screen label throughout, whether or not ASCI requires one.

Which of ASCI’s bans can already be enforced elsewhere?

ASCI’s decisions carry no legal penalty, but most of its banned uses already fall under laws or court rulings that do.

Fake endorsements, exaggerated claims and fake locations: The Central Consumer Protection Authority’s (CCPA) 2022 guidelines on misleading advertisements and endorsements, issued under the Consumer Protection Act, 2019, already prohibit misleading claims and endorsements by advertisers.
Unauthorised copyrighted work: The Copyright Act, 1957, already prohibits it.
Deepfakes without consent: The IT Rules’ deepfake provisions require platforms that offer AI tools to block AI content that “falsely depicts or portrays a natural person or real-world event” in a way “likely to deceive.” The person depicted can also go to court over personality rights, which courts have recognised as protecting a person’s name, face and voice. High Courts have issued orders against AI-generated content for actors including Samantha Ruth Prabhu, Anil Kapoor and Jackie Shroff.

What ASCI adds is a duty on advertisers themselves to label AI use with consent, such as a celebrity’s voice cloned with permission. The IT Rules place that duty on platforms, and the CCPA’s 2022 guidelines say nothing about AI disclosure.