India's advertising watchdog has told brands to label synthetic influencers, and has added that a label will not rescue an ad that misleads. The Advertising Standards Council of India (ASCI) published its Guidelines for Responsible Labelling of Synthetically Generated Content in Advertising on September 29, 2026. They apply three months after publication. An influencer generated by software now has to be presented as such.
What the guidelines require
ASCI describes itself as an "independent, voluntary self-regulatory organisation", and the new text follows a draft released in May. The law firm Khaitan & Co dates that draft to 8 May 2026; comments were invited until 13 June. The trade site MN4U reports that the framework looks at what synthetic content does to consumers, not at the technology used to make it.
Trade coverage describes four areas. The first is a list of uses that stay prohibited whether or not an AI label is attached: fabricated endorsements or testimonials, misleading synthetic depictions of product results, non-existent locations presented as real, unauthorised use of copyrighted work, and deepfakes or a person's likeness used without consent.
The second area is mandatory labelling. Disclosure is required when synthetic content materially influences a consumer's decision and leaving it out could mislead. That covers synthetically generated influencers and ambassadors, the replication of a real person's likeness or voice with consent for personalised messages, fabricated events or settings, demonstrations of products that do not exist yet, and sound effects tied to a product's core features. Paid or sponsored AI-generated product suggestions must carry the words "Sponsored by [Brand]".
The third area lists what needs no label: colour correction, lighting adjustments, noise reduction, blemish touch-ups, decorative or ambient elements, obvious fantasy such as dragons, ad copy written with AI, and accessibility tools such as subtitles, translations and audio descriptions. The fourth sets the wording. Brands may write "Audio/Video created using AI" or "Audio/Video enhanced using AI", use a label supplied by an ad platform, or find their own words if they describe the content accurately.
One detail from the draft did not make it into the summaries of the final text. The May version gave, as an example of prohibited content, an AI-generated fake doctor promoting a supplement to imply medical endorsement. Neither Storyboard18 nor MN4U repeats that example in its account of the final guidelines, and ASCI's own document could not be retrieved for this article.
A label is not a licence
The sentence that sets this text apart came from ASCI's secretary general and chief executive. "The responsibility remains with advertisers to ensure that the end communication is honest, transparent and compliant with the ASCI Code," Manisha Kapoor said, adding that the guidelines clarify "where the mere use of an AI label may not make an otherwise misleading advertisement acceptable."
In practice, a brand cannot invent a testimonial, stamp "AI" on it and call the matter closed.
Every ad using synthetic content is still judged against the whole ASCI Code.
Britain's Committees of Advertising Practice (CAP) reached a similar position, though by another route. According to the law firm RPC, which summarised the guidance on 23 September 2025, the UK has no AI-specific rules requiring disclosure, existing rules apply however an ad is made, and disclosing AI use alone is very unlikely to cure a fundamentally misleading message. CAP's own example is an AI-generated image of a cosmetic's effect that does not reflect real results. The same summary notes that telling people an influencer is AI-generated could help an advertiser avoid a misleading impression. India has turned that hint into a duty for one category of content.
The European Union takes a third road. Article 50 of the AI Act obliges deployers of an AI system that generates a deep fake to disclose that the content was artificially generated or manipulated. The text on artificialintelligenceact.eu gives 2 August 2026 as its start date. The Act defines a deep fake as content that resembles existing persons, objects, places, entities or events and would falsely appear authentic. An invented influencer resembles no one in particular, so on a plain reading the question of whether Article 50 reaches it is left open.
Why a made-up face persuades
The reason regulators care is a mechanism, and one 2025 experiment lays it out. Zoe Olbermann and Magdalena Nagl, of the University of Würzburg, published it in Frontiers in Communication on 5 June 2025. They showed 169 participants a fictitious Instagram post: a brand ambassador for an invented fitness-tracker company, Motion 360, apologising after a newspaper article about data-protection concerns. The ambassador was a human influencer, an animated virtual one or a humanised virtual one, depending on the group.
Their model runs as a chain. The more human a virtual influencer looks and seems, the more similar to themselves viewers judge it. Similarity feeds wishful identification, the wish to be like the person on screen. Identification, the authors presume, leads to less critical processing of the message, and a message accepted more readily lifts the brand's image. The authors found that mental human-likeness, meaning the sense that the character has a mind, was the strongest predictor of perceived similarity.
Viewers also separated what they saw from what they believed. Appearance and mind were judged independently: recipients can recognise that a virtual influencer looks like a person while granting it nothing human in its mental abilities. The human influencer still scored highest on similarity, and the humanised virtual influencer beat the animated one. The old puzzle of why nearly human machines unsettle us sits underneath this result, because the same dial that makes a face persuasive can also make it uncomfortable. Companion apps, another kind of synthetic character, have been studied at a different moment of the relationship: how they answer a user's goodbye.
The authors drew a design lesson from it: "Virtual influencers should be developed through storytelling into characters that have feelings, care and stand up for values." Read next to ASCI's rule, that sentence marks the tension. A marketer is advised to give the character an inner life, and the regulator requires a label saying the character is generated. The study has limits its authors name, and they matter here. The influencer was fictitious and female, and the scenario was a single corporate apology, so the experiment tested crisis communication rather than everyday product promotion, and no real virtual influencer with a following of its own was involved.
What the label does to trust
The ASCI rule assumes disclosure helps the consumer. Marketing research is less clear about what disclosure does to the brand. In an experiment published on 22 November 2025 in the Journal of Marketing & Social Research, Fatema Juzer Ujjainwala ran an experiment on 320 social media users, comparing AI and human influencers, with and without explicit disclosure. AI influencers reduced perceived authenticity and brand trust compared with human ones, and explicit disclosure made those effects worse. People with high digital literacy reacted less negatively.
Treat that result with care. It comes from one between-subjects experiment in a small open-access journal, and only its abstract and introduction were read for this article. An earlier paper in a peer-reviewed journal, Lim and Lee's 2023 study in Computers in Human Behavior, ran two experiments on how disclosing a virtual influencer's origin, and using positive or negative emotional narratives, affect its credibility through perceived parasocial interaction, the one-sided bond a viewer feels with a media figure. Its abstract says earlier studies found virtual influencers engage consumers as effectively as humans, while their artificial identity raises credibility concerns.
If the Ujjainwala finding holds, honesty has a price: brands will pay in trust to tell the truth. That is the bargain behind the label. Kapoor's stated test is whether a consumer's understanding of an advertisement may be adversely affected without an AI label.
Kyra, Naina and the market for virtual faces
India's virtual influencers are not a thought experiment. Forbes India reported on 23 October 2024 that Kyra, created in 2022 by the Bengaluru studio Futr Studios, had 260,000 followers on Instagram and 103,000 on YouTube, and that some of her brand videos had reached 74 million views. A leading digital influencer could quote between ₹5 lakh and ₹10 lakh for a single video, the same range as human creators, the magazine said.
"We looked at the success of virtual influencers like Lil Miquela from Brazil (2.5 million Instagram followers) and Imma from Japan (392,000 Instagram followers) and noticed a gap in India, and the potential to have an early-mover advantage," co-founder Himanshu Goel told the magazine.
Kyra is openly digital and often stars in futuristic settings. Another character, Naina, from Mumbai's Avtr Meta Labs, lives and creates content in the real world, working with real people and celebrities. As trade coverage summarises it, ASCI's wording covers both, since it speaks of synthetically generated influencers and ambassadors without a test of how real they look.
The magazine also cited the Influencer Marketing Hub, a Copenhagen-based research outfit: 60.4 percent of respondents to its 2023 benchmarking report had used virtual influencers, and 62.2 percent a year later. These are marketers surveyed by an industry publisher, and the figures are two years old.
What the rules leave open
ASCI is a voluntary body, and the coverage read for this article does not describe how breaches of the new text will be handled. The effective date is stated only as three months after publication.
The guidelines also sit beside a separate government rule. India's Ministry of Electronics and Information Technology notified amendments to the IT Rules on 10 February 2026, in force from 20 February, which require labelling of what they call synthetically generated information. Law firm Freshfields summarises the definition as audio or visual content created or altered so that it appears real and portrays a person or event in a way that is, or is likely to be perceived as, indistinguishable from a natural person or real-world event. On the wording of that definition, a visibly animated character like Kyra might fall outside it. ASCI's rule would still reach her.
The IT Rules have critics. WITNESS, the human rights organisation, welcomed that the final rules apply only to audio and visual content and dropped a requirement that a visible label cover 10% of the content. Its executive director Sam Gregory still warned: "provenance that doesn't travel across platforms isn't provenance, and detection tools that can't reliably tell synthetic from authentic content shouldn't be the basis for legal enforcement." ASCI's text puts the responsibility on advertisers, as Kapoor said, and does not rest on detection tools.
The detail left for last is a design choice. Manifest Media reported in May that the draft was aligned with the IT Rules to ensure transparency while avoiding consumer label fatigue. ASCI's list of what needs no label is the other half of that aim.