AI Ad Labelling in India: What Sellers Must Know

Festive campaigns in India will lean harder on AI-made images and short videos this season, which makes AI ad labelling in India a practical question rather than a legal footnote. On 8 October 2026, MeitY’s Cyber Laws Division issued an advisory asking social media platforms to label manipulated and synthetic content, especially when it is presented as a current event. If you run a shop, sell on marketplaces or make ads for clients, three questions matter: what has to be labelled, who carries the duty, and what your tool should do for you.

AI ad labelling in India: the rules already in force

The Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026 came into force on 20 February 2026 (Freshfields). They define synthetically generated information (SGI) as audio, visual or audio-visual material that a computer resource creates or alters so that it appears real and is hard to tell apart from a real person or event. Deepfakes, AI-altered photos and videos, and voice clones are covered. Routine or good-faith editing is excluded. AI-generated text is not SGI under these rules, although existing duties on unlawful text still apply.

Non-prohibited SGI must be clearly and prominently labelled. Visual content needs a visual label and audio needs an audio disclosure. The final rules dropped an earlier draft’s 10% watermark requirement. Significant social media intermediaries must ask users to declare whether an upload is SGI and must verify those declarations with technical measures, rather than relying on the declaration alone. Where feasible, platforms must embed permanent metadata that can trace content back to the system that made it, and labels and metadata must not be removed or altered.

Takedown windows also shortened. For lawful court or government orders about unlawful content, the window fell from 36 hours to roughly three hours. For high-risk categories such as morphed images and impersonation, it fell from 24 hours to two hours (Freshfields; Legal500).

The October advisory: broader in scope, aimed at platforms

The 8 October 2026 advisory is addressed to social media intermediaries. It asks them to identify and label, or add context to, manipulated, synthetic or misleading content, particularly when that content is presented as a current event. It also asks platforms to prioritise complaints from authorised government agencies about public-order content, and to encourage users to check the source, authenticity and date of information before sharing it. It takes immediate effect (MediaNama).

Three points matter for advertisers. First, the coverage we read does not mention advertisers or ad creatives at all. Second, the advisory lists no specific penalties, unlike some earlier MeitY advisories. Third, MediaNama notes that its labelling clause is not limited to audio-visual content the way the February amendment is, so text-only material could arguably fall within its scope. Terms such as “authorised government agencies” are not defined, and legal commentary is still developing. Read the advisory as a clear signal of enforcement direction rather than a finished rule for your ad copy.

What is still a draft

MeitY placed draft amendments in the public domain on 30 March 2026 (MediaNama). The draft would require labels on visual content to be “continuous and clearly visible… throughout the duration of the content”. Audio would need a clear disclosure at the start. Permanent metadata and identifiers would be required, and users would be barred from removing them. Consultation was extended to 7 May 2026 after further changes. We have not found evidence that the draft has been notified, so plan for it without treating it as settled law. If it passes in its current form, a small tag in one corner at the start of a 30-second reel would no longer be enough.

How InstaStudio handles labels, metadata and takedowns

A tool cannot make your ad compliant on its own, and the rules place duties on platforms and on the people who publish. Still, the product choices you make matter. Based on the InstaStudio feature set as it stands today, these are the parts that bear on labelling and takedowns:

• A visible AI label on outputs. Paid plans get a discreet tag, and the free plan adds a watermark.
• IPTC AI metadata on outputs, plus signed C2PA Content Credentials. The trust-list certificate that makes C2PA credentials verifiable is still pending, so do not treat them as proof of origin yet.
• A /report grievance and takedown form with 2-hour, 3-hour and 24-hour deadlines, feeding an admin queue.
• DPDP consent records, data export and account deletion.

Pricing is as follows, and all of these prices include GST. The free plan is watermarked. Basic costs ₹799 per month and removes the watermark, with commercial use included. A Day Pass gives Pro features for 24 hours for ₹199. A 500-credit pack costs ₹79. You can see plans on the InstaStudio product page.

What sellers and creators should do this festive season

Label AI-made creatives yourself even when a tool already does it, because a platform’s detection may miss an edit. Keep the original prompt, the source photos and the approval record for each ad, so you can answer a complaint quickly if one arrives. Before a festive push, pick a single person to check every creative against the label and claim checklist, and make sure someone can pull a flagged ad down the same day. Do not use AI to create a scene that looks like a current news event, a real person’s endorsement or a real customer’s testimonial. Get written permission before using a real person’s likeness, and check your marketplace or ad platform’s own AI disclosure settings before you launch. For regulated categories such as health, finance or food, have a lawyer review the claims as well as the labels.

Frequently Asked Questions

Does the IT Rules labelling duty apply to AI-written ad copy?

Not under the amended IT Rules. AI-generated text is not SGI, though other duties on unlawful text still apply. The October advisory’s wording is broader, and MediaNama notes that text could fall within it, so keep copy accurate and avoid invented claims.

Is a watermark enough to meet the labelling rules?

Not necessarily. The rules require clear and prominent labels on non-prohibited SGI, and the draft would require continuous visual labels. A watermark is one part of a labelling process, not a compliance check. Review each output against the current rules before you publish.

Does the draft labelling amendment already apply?

Not as far as we have found. It was put to public consultation on 30 March 2026, and the deadline was extended to 7 May 2026. Check MeitY’s website for any notified text before you rely on it.

How quickly must platforms remove flagged AI content?

Under the amended rules, platforms should act within roughly three hours of a lawful court or government order about unlawful content. For high-risk categories such as morphed images and impersonation, the window is two hours. Other grievance timelines are longer.

Sources

• MediaNama: MeitY widens labelling for social media beyond synthetic content (October 2026)
• MediaNama: MeitY tightens AI label rules, mandates continuous disclosure (April 2026)
• Freshfields: India targets deepfakes and AI-generated content, key changes under MeitY’s 2026 amendments to the IT Rules
• Legal500: IT Rules 2026 deepfake regulation, three-hour takedowns and AI labelling obligations

Want festive creatives with visible labels and embedded metadata from the first draft? Try InstaStudio, or browse more articles on the InstaDataHelp blog.