ASCI's new AI ad rules, explained in plain English: what you must label, what you needn't, and what's banned

In June 2026, the Advertising Standards Council of India (ASCI) released draft guidelines for AI-generated content in advertising. So far the only people who've explained them are law firms writing for other lawyers. Here's the business-owner translation: what the rules actually ask of you, what they don't, and the five-minute audit that keeps your ads clean.

Two honesty notes before the useful part. First: these are draft guidelines under public consultation as of this writing — details can change before they're final, and we'll update this post when they do. Second: this is a marketer's explainer, not legal advice; if you're running high-stakes campaigns in regulated categories, spend the money on a lawyer.

The one idea behind the whole framework

ASCI didn't write an anti-AI rulebook. The draft is built on a principle Indian advertising law has always had: an ad must not deceive. AI just created new ways to deceive, so the regulator sorted AI usage into three risk tiers. Everything below hangs off that one idea — if a reasonable viewer would be misled about what's real, you have a problem; if not, you mostly don't.

Tier 1 — No label needed: AI as a production tool

Using AI the way you'd use Photoshop or a colour grade doesn't require disclosure under the draft: touch-ups, background cleanups, upscaling, generated set-dressing and environments, style edits — anything that doesn't create a false impression of reality about people, products or claims.

Practically: the compliance-safe product-imagery workflow — real product pixels, generated context — sits comfortably in this tier. So does most well-made AI brand video, where nobody is being passed off as a real person and the product shown is the product sold.

Tier 2 — Label it: synthetic content that could be mistaken for real

Where AI generates content a viewer could reasonably take as real — synthetic humans presenting as actual people, realistic depictions of events that never happened, AI-generated "results" or demonstrations — the draft requires clear disclosure, along the lines of a "created using AI" label a viewer can actually notice (not grey micro-text in a corner).

The mindset shift this asks of advertisers is small but real: disclosure moves from "optional transparency flex" to "expected default" for realistic synthetic content. If your first instinct is that a label would kill the ad's effectiveness, take that as data — audiences punish concealed AI far harder than disclosed AI, and the surveys behind that predate any regulation.

Tier 3 — Banned outright: no label makes it okay

The draft draws hard lines that no disclosure cures:

• Deepfakes of real people without consent — a celebrity's face or voice endorsing you via AI is prohibited, full stop. (With consent and disclosure, licensed digital likenesses are a different, legal animal.)
• Fabricated testimonials and endorsements — AI-generated "customers" praising your product, invented reviews, synthetic before/afters. If the persuasive force of the content is "a real person experienced this" and no real person did, it's out.
• Deceptive claims dressed as reality — synthetic evidence for product performance that never occurred.

If any of this stings, the cheap version of compliance is one design decision: never use AI to manufacture social proof. Generate worlds, scenes, motion, product contexts — not people vouching for you.

This isn't only ASCI

Two adjacent facts worth knowing. MeitY amended the IT Rules in early 2026 to cover synthetic media, so platform-level labelling obligations are arriving from the government side too — the platforms you advertise on will increasingly ask you to self-declare AI content at upload. And ASCI's existing influencer guidelines already require disclosure when influencers are virtual/AI personas. The direction of travel is one-way: labelled synthetic content, everywhere. Building the habit now costs nothing; retrofitting it after a takedown or a press cycle costs plenty.

The five-minute self-audit

Run your current ads through four questions:

1. Does any AI-generated element present as a real person, event or result? If no → Tier 1, carry on. If yes →
2. Is it labelled clearly enough that a normal viewer notices? If yes → Tier 2, compliant. If no, fix it. Then:
3. Is anyone endorsing or testifying who doesn't exist, or without consent? If yes → Tier 3. Pull it today; that one's not negotiable, draft or no draft.
4. Would you defend the ad, unedited, on a news channel? The regulator's tiers are basically this question in legal clothing.

Most honest advertisers pass in the promised five minutes. The ones who don't are usually one asset swap away — replace the synthetic testimonial with a real customer on camera (the one place we always say use real humans) and the rest of the AI pipeline, with its very real cost advantages, stays fully available.

What we'd do this quarter

• Add an "AI content" line to your creative sign-off checklist — tier every asset before it ships. (If your team needs broader ground rules for AI at work, start with an AI policy.)
• Keep records of what's synthetic in each campaign; consultation drafts become enforcement realities faster than teams update their archives.
• Watch the disclosure norm as a brand choice, not just a legal one — early honest labellers are accumulating trust while competitors get outed. The same transparency logic is reshaping how brands surface in AI search.

Regulation like this is good news for anyone doing AI creative properly. Rules that kill fake testimonials and covert deepfakes don't threaten directed, honest AI production — they clear the slop out of the auction.