In many of the cases that matter, the rules now point towards a label. Since 20 February 2026 India's amended IT Rules require platforms to label synthetically generated audio and video prominently, and ASCI's new guidelines, reported on 29 September 2026, ask advertisers to disclose AI content wherever it materially influences a consumer's decision. Brands selling into Europe also face Article 50 of the EU AI Act, in force since 2 August 2026, which requires deployers to disclose deepfake image, audio and video.
The harder question for a CMO is which rule reaches which asset, and who in the chain carries the obligation. Three rulebooks now apply, and each one points at a different party.
Three rulebooks, three different targets
The IT Amendment Rules are written for intermediaries: the platforms that host your Reel and the tools that generate it. ASCI's code is written for advertisers. The EU AI Act addresses providers of AI systems and the deployers who use them. Whether the deployer on a given campaign is the brand or the agency running the tools is a question to put to counsel early.
Read only the IT Rules and you might assume the platform handles labelling. Read only ASCI and you miss the metadata duty that travels with the file.
The practical consequence: labelling has moved into production. The call gets made at the brief, and someone records who decided what, long before anyone uploads a file.
What the IT Rules require, and of whom
The Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026 were published as G.S.R. 120(E) and came into force on 20 February 2026 (Source: Khaitan & Co., February 2026).
The rules create a category called synthetically generated information: audio, visual or audio-visual content created or altered by algorithm so that it appears authentic. Text-only content, routine or good-faith editing, and accessibility tools that do not change the underlying content fall outside it (Source: Khaitan & Co., February 2026).
Two obligations matter to a marketing team. First, a platform that offers tools to create synthetic content must label visual output prominently and prefix audio output with a disclosure, and must embed permanent metadata or a provenance identifier that users cannot strip. Second, significant social media intermediaries must ask the uploader to declare whether content is synthetic, and must deploy technical measures to check that declaration (Source: Khaitan & Co., February 2026).
That second duty reaches into your team's workflow. When your agency uploads a Reel to a large platform, someone on your side answers the declaration. If the answer is wrong, the platform's verification can overrule it and place a label the art director never chose. Who carries the consequences of a wrong declaration is worth asking counsel before the first upload.
The draft rules proposed a label covering 10 percent of the frame. The final rules dropped the number in favour of a qualitative standard: prominent, easily noticeable and adequately perceivable (Source: Khaitan & Co., February 2026). Qualitative standards reward brands that set a house standard early.
ASCI draws the line at material influence
ASCI's guidelines on labelling synthetically generated content in advertising follow a draft consultation earlier in the year and come into effect three months from publication (Source: Business Standard, 30 September 2026). The framework sorts AI use into three buckets.
Prohibited uses stay prohibited whatever the label says. ASCI treats a fabricated endorsement, a deepfake without consent or a misleading synthetic demonstration as a breach of its Code, and a disclosure line does not cure it (Source: Business Standard, 30 September 2026).
Mandatory disclosure applies where synthetic content materially influences a consumer decision and its absence could mislead. The reported examples are synthetic influencers, a replica of a real person's likeness or voice used with consent, fabricated events or settings, and demonstrations of products that do not yet exist (Source: medianews4u, 29 September 2026).
Suggested label wording is plain: "Audio/Video created using AI" or "Audio/Video enhanced using AI", and platform-provided labels are acceptable (Source: Business Standard, 30 September 2026).
No label is needed for routine editing such as colour correction, decorative elements, obviously fantastical effects, and accessibility uses like subtitles (Source: medianews4u, 29 September 2026).
The test a CMO should carry into every review is the ASCI one. Would a reasonable viewer decide differently if they knew this element was generated? A generated sky behind a real product rarely changes that decision. A generated patient describing recovery passes it, and in healthcare that ad has bigger problems than its label. "What healthcare brands need from creative in a regulated market" covers the claim substantiation side of healthcare creative, and "Faces you do not own" covers consent for real faces and voices. This piece stays on the label.
Article 50 for brands selling abroad
Article 50 of the EU AI Act took effect on 2 August 2026. Deployers of an AI system that generates or manipulates image, audio or video constituting a deepfake must disclose that the content is artificially generated or manipulated (Source: EU AI Act, Article 50, obligations in force 2 August 2026). Where the content forms part of an evidently artistic, creative or fictional work, the duty narrows to disclosing the AI element in a way that does not spoil the work. Penalties reach 15 million euro or 3 percent of worldwide annual turnover (Source: EU AI Act, Article 99; European Commission Guidelines on Article 50, adopted 20 July 2026).
Article 50 is aimed at AI output used inside the Union. An export brand, a medical tourism campaign aimed at EU patients or a global cutdown of an Indian film are the cases to raise with counsel, along with whether an India-only campaign is truly out of scope. The Commission's guidance also sets a useful bar: a vague reference does not satisfy a transparency duty, and the test is a reasonably observant person from the actual audience (Source: European Commission Guidelines on Article 50, July 2026). Designing to that bar is a sensible house standard, and counsel can confirm whether it also covers you in India.
What a label looks like on a Reel, a static and a long film
None of the three rulebooks hands you a layout. What follows is a practical approach for each format. Treat it as a starting point for your legal team to test.
On a Reel. Put the on-screen label in the first two seconds, inside the safe zone, clear of the platform UI and caption overlay. Use the platform's own AI toggle as well, because that is what the declaration duty reads. Keep the generation tool's provenance metadata intact through the edit; some export and compression steps strip it, so check the delivered file.
On a static. Burn the label into the image. A caption can be truncated, cropped out in a repost or lost when the creative is resized for a different placement. The label has to survive every cut of the static that the media plan will run.
On a long film. Use an opening or closing card for the film as a whole, and a separate in-frame label on the specific scene where a synthetic person or a synthetic setting carries the claim. If a voice is cloned or generated, add a spoken or on-screen disclosure near its first use, since the IT Rules treat audio as its own case.
Two production habits make this cheap. Tag every generated or AI-enhanced element in the edit decision list as it is made, so the labelling call rests on facts recorded at the time. And design the label into the layout from the first draft, so it does not arrive as a late overlay that fights the art direction.
Who owns the label: brand, agency or platform
Each party holds a different part, and the exact split is a question for counsel. The IT Rules put the technical label, the metadata and the verification of declarations with the platform. ASCI's chief executive Manisha Kapoor put responsibility on advertisers to keep the end communication "honest, transparent and compliant" (Source: medianews4u, 29 September 2026). The agency or production partner holds the facts: which elements were generated, which were enhanced, and whether any real person's likeness was used with consent.
Where the chain breaks is the handover. The agency knows what was generated, the brand approves the final cut, and a media buyer answers the platform's declaration. Put all three in one record per asset and the declaration becomes a lookup.
In our Creative Intelligence Pod, Nextdot ships labelling as a standard option on AI-generated creative, recommends that clients use it, and records each client's choice. Every brief and approval is stored in the pod's memory layer. The decision stays with the brand, because the brand is the advertiser. Whoever makes your creative, ask for that per-asset record of what was generated, who approved it and which label was chosen, so it exists when a platform, a regulator or ASCI asks.
Frequently asked questions
Do AI-generated ads need a disclaimer in India?
Where the AI content could materially influence a consumer's decision, ASCI's guidelines call for one. Reported on 29 September 2026 and coming into effect three months from publication, they point to synthetic influencers, consented replicas of real people, fabricated settings and demonstrations of products that do not yet exist. Separately, since 20 February 2026 the amended IT Rules require platforms to label synthetic audio and video and to collect uploader declarations, so an undisclosed ad can still end up labelled by the platform. How this applies to a specific campaign is a question for counsel.
What does the EU AI Act say about deepfakes in advertising?
Article 50, in force since 2 August 2026, requires deployers of AI systems that generate deepfake image, audio or video to disclose that the content is artificially generated or manipulated. For evidently artistic or fictional work, the disclosure can be lighter so it does not spoil the work. Whether a given campaign falls within its reach is worth confirming with counsel.
What counts as synthetically generated content?
Under India's IT Amendment Rules 2026, it is audio, visual or audio-visual content created or altered by algorithm so that it appears authentic. Text-only content, routine or good-faith editing and accessibility tools fall outside the definition. ASCI adds that routine edits such as colour correction, decorative elements and obviously fantastical effects do not need an advertising label.
Who is responsible for labelling, the brand or the agency?
ASCI places responsibility for the honesty of the final ad on the advertiser, so the brand usually makes the disclosure decision. The agency holds the facts about what was generated and should hand them over per asset. The IT Rules give platforms the technical label and metadata. Take the exact allocation to counsel, and keep a shared record of what was generated, who approved it and what label was chosen.