Zeralytics logoZERALYTICS

ASCI AI Labelling Guidelines: When Your Ads Need an AI Label

If your business uses AI to make ad images, videos or voice-overs, the ASCI AI labelling guidelines now decide when you must tell customers about it. The Advertising Standards Council of India (ASCI) released its final Guidelines for Responsible Labelling of Synthetically Generated Content in Advertising on 29 September 2026, and they take effect three months after publication, which law firm Khaitan & Co puts at 29 December 2026.

The short answer: you need a label only when AI content could change a buying decision and hiding it would mislead people. Routine edits, decorative backgrounds and AI-written copy do not need one. Fake testimonials and misleading AI "results" are banned even if you add a label. Below: the three categories, how Google and Meta label AI ads, and a checklist for your year-end and Pongal campaigns.

Key takeaways

  • ASCI's AI labelling guidelines were released on 29 September 2026 and apply from three months after publication, around 29 December 2026.
  • Every AI-made ad element falls into one of three groups: prohibited, label required, or no label required.
  • A label is required only when the AI content materially influences buying decisions and leaving it unlabelled would mislead.
  • Fabricated testimonials, misleading AI before-and-after images and unauthorised likenesses are not allowed, label or no label.
  • Google Ads already shows on-ad AI overlays for campaigns targeting India when assets are labelled, so platform settings matter too.

What are the ASCI AI labelling guidelines?

The ASCI AI labelling guidelines are a set of rules from India's advertising self-regulator that explain when ads using AI-generated or AI-altered audio, images or video must carry a disclosure. They focus on the effect on consumers, not the tool. Ads using AI must still follow the full ASCI Code, so a label is an extra layer, not a loophole.

Synthetically generated content (SGC) is any audio, visual or audio-visual content in an ad that a computer creates or materially alters so it looks real and could be believed to show a real person or event. Text is not included, which is why AI-written copy needs no label.

A May 2026 draft sorted ads into high, medium and low risk. According to Fox Mandal, the final version uses three plain categories instead and adds assessment questions.

When do the ASCI AI labelling guidelines come into effect?

The guidelines say they apply on the expiry of three months from publication. ASCI published them on 29 September 2026, so Khaitan & Co reads the start date as 29 December 2026. Some trade coverage, such as exchange4media, cites 1 January 2027. The safe plan is to be ready before Christmas.

Timing matters for Tamil Nadu businesses in particular. Thai Pongal falls on Friday, 15 January 2027, according to Drik Panchang. Pongal creatives designed in December will run after the guidelines apply, so review them now rather than pulling ads mid-festival.

Which AI ads need a label?

An ad needs a label when its AI-generated content materially influences consumer decisions and leaving it undisclosed would mislead people. ASCI suggests three questions: Is the AI content central to the ad's persuasion or likely to sway a purchase? Would people be misled about the product's claims or performance without a label? Would the missing disclosure create a false impression?

The guidelines list these examples where labelling is mandatory:

  • Synthetic influencers and ambassadors who front your brand.
  • Replicated faces or voices: a real person's likeness or voice recreated with their consent for personalised messages.
  • Fabricated events or settings that could affect how people understand product performance.
  • Products that do not exist yet: ASCI's own example is a 3D model of an unbuilt housing complex, which is directly relevant to real estate developers.
  • Core-feature sound effects, such as audio quality in a headset ad.
  • Sponsored AI recommendations: paid product suggestions from an AI tool must say "Sponsored by [Brand]".

The last point matters as ads move into AI answers, such as ads in Google AI Mode in India. ASCI's example is a chatbot recommending a moisturiser for Mumbai weather because the brand paid for it.

What is banned even with an AI label?

Some uses are prohibited whether or not you add a label. These are ads that are illegal, infringe someone's rights, make misleading claims or break the ASCI Code. A label cannot fix a false message. As MediaNews4U reported, ASCI's Secretary General and CEO, Manisha Kapoor, said responsibility for honest communication "remains with advertisers".

The guidelines give four examples of prohibited content:

  1. Fabricated endorsements or testimonials: using AI to show a real person recommending a product they never used or approved.
  2. Exaggerated AI results: for example, an AI image of a toned body in a weight-loss ad when the results are not real or substantiated.
  3. Fake locations: such as a travel agency showing a hotel or facility that does not exist as if it were real.
  4. Unauthorised material: copyrighted work, deepfakes or likenesses used without consent.

For small businesses, two common risks are AI "customer" review videos and AI before-and-after images for clinics, salons or gyms. Replace both with real, documented results.

Which AI uses do not need a label?

No label is needed when AI only makes minor changes or adds elements that do not affect a customer's ability to make an informed choice. This covers routine editing, decorative backgrounds, obvious fantasy, AI-written copy and accessibility features, as long as the substance and core claims of the ad stay the same.

  • Minor enhancements: colour correction, lighting fixes, noise reduction and standard blemish removal.
  • Decorative elements: backdrops, jingles or crowd sounds unrelated to what the product can do.
  • Obvious fantasy: effects nobody would take as real.
  • Text and accessibility: ad copy, subtitles, captions, translations and audio descriptions.

Common AI ad uses and where they fall

Here is how typical small-business uses map to the three categories. Rows on unbuilt projects, voice clones, captions and testimonials follow ASCI's own examples; the rest are our reading, so check close calls with a legal adviser.

AI use in your adLikely categoryWhy
AI walkthrough of an apartment project not yet builtLabel requiredASCI's own example of a product that does not exist yet
AI-generated model wearing your saree or showing your productLabel likely requiredClose to a synthetic ambassador and shows how the product looks in use
AI studio background behind a real product photoNo labelDecorative, as long as it does not imply a real store or location
AI-written captions, headlines and Tamil or Hindi subtitlesNo labelText and accessibility uses are exempt
AI clone of your doctor's or founder's voice for personalised messagesLabel requiredReplicated likeness or voice, even with consent
AI "happy customer" testimonial videoProhibitedFabricated testimonial; a label does not make it acceptable

How do Google and Meta label AI-generated ads?

Both platforms have their own AI labels, and ASCI accepts platform-provided labels. On Google Ads, you can mark assets as AI generated or edited, and campaigns targeting India show a visible overlay on the ad. On Meta, ad images made or significantly edited with AI can get an "AI info" label.

As of November 2026, Google Ads Help says labelled assets get a disclosure in the "How this ad was made" section of My Ad Center, with on-ad overlays for campaigns targeting India, the EU and a few other regions. You can set the label when reviewing assets or later in Asset studio. Google may also label assets itself.

If you add your own label to an image, keep it away from corners and switch off image enhancements that may crop it.

As of November 2026, Meta's help page says the "AI info" label applies to ad images created or significantly edited with Meta's generative AI features, or created or edited with third-party AI tools. It appears in the "About this ad" menu and sometimes also next to the Sponsored label. Resizing and colour correction are not labelled, and Meta says not every ad made with its own AI features will carry the label, so do not rely on the platform alone.

Are the ASCI guidelines legally binding?

ASCI is a voluntary self-regulatory body, not a statutory regulator. But ignoring its guidelines is risky: according to Fox Mandal, if a complaint is upheld the ad must be withdrawn or modified within ten business days, ASCI informs the agency and media partners, and it may refer the matter to a regulator. Other laws apply too.

The IT Rules amendment that took effect on 20 February 2026 requires intermediaries such as social platforms to label synthetic content, as Fox Mandal explains. Khaitan & Co notes the difference: platforms label all synthetic content, while ASCI asks advertisers to label only where it matters to consumers. Under the Consumer Protection Act, 2019, the Central Consumer Protection Authority (CCPA) can order misleading ads to be stopped or changed and impose penalties, according to exchange4media. This is general information, not legal advice.

Your 7-step AI ad compliance checklist

Use this before any AI-assisted creative goes live:

  1. List every AI element in live and planned ads: images, video, voice, music and models.
  2. Run ASCI's three questions on each element and note whether it is prohibited, label required or no label.
  3. Remove prohibited content first, such as AI testimonials.
  4. Add clear labels such as "Video created using AI" or "Image enhanced using AI" where required, and follow ASCI's disclaimer guidelines where they apply.
  5. Set platform labels in Google Ads Asset studio and check Meta's "AI info" on your ads.
  6. Update creator and model contracts to cover consent for AI likeness or voice use, how long it can be used and what happens if consent is withdrawn.
  7. Record your decisions for each ad, as Khaitan & Co suggests, so you can show why a label was or was not added.

If you work with influencers, point six matters most. As exchange4media points out, agreements written for recorded content may not cover new AI-generated statements made with a creator's face or voice. If you run influencer marketing campaigns, add AI likeness consent as its own clause in creator agreements.

Frequently asked questions

Do I need to label an ad if I only used ChatGPT to write the copy?

No. ASCI lists generating or enhancing advertising copy as an administrative and text use that needs no label. The guidelines cover audio, images and video that look real. But the claims in your copy must still be true and substantiated under the ASCI Code, so check every number, offer and promise before publishing.

What wording should an AI disclosure label use?

ASCI gives "Audio/Video created using AI" and "Audio/Video enhanced using AI" as examples, and accepts labels provided by ad platforms. You can write your own wording if it accurately tells people what was made with AI. For paid recommendations shown by an AI tool, the disclosure should clearly read "Sponsored by [Brand]" so people know it is a paid placement.

Does an AI-generated background in a product photo need a label?

Usually not. ASCI treats purely decorative backgrounds and abstract settings as needing no label, provided they are unrelated to what the product can do. But showing a showroom, hotel or location that does not exist as if it were real is prohibited.

Do the guidelines apply to small businesses and organic posts?

Khaitan & Co notes the guidelines apply to all advertisers using synthetic content, and the text has no exemption for small businesses. Paid ads clearly fall within them. Separately, the IT Rules make social platforms label synthetic content, so AI-made organic posts may be labelled too. When in doubt, treat promotional posts like ads.

Sources

Tip: AI creative is still a fast, low-cost way to test more ad ideas; the new rules just ask you to be honest about it. Our performance marketing team can review your live Google and Meta ads against this checklist and help you set up AI labels before your Pongal campaigns launch.