Most Indian advertisers are now using generative AI somewhere in their creative pipeline — a generated background, an upscaled product shot, a synthetic voiceover, a virtual presenter. Through 2026 that stopped being a purely creative decision and became a compliance one. Google began rolling out AI labelling for image and video ad creatives in July 2026, driven by regulation in the EU, India and New York. ASCI published draft guidelines for labelling synthetically generated content in advertising in May 2026, aligned to the IT Intermediary Guidelines amendment rules of the same year. Neither is optional reading if you run ads in India. Here is what they actually say, in plain terms.
What Google Now Requires
Google's July 2026 update covers image and video ad creatives that were generated or modified using AI, across Google Ads, Display & Video 360, Campaign Manager 360, Merchant Center and Ads Editor. Advertisers can either add text or visual labels directly inside the creative, or use the AI label setting that rolled out across Google's advertising products through July 2026. Election advertisers have a separate obligation: synthetic or digitally altered content must be disclosed through the campaign settings checkbox under the Political Content policy.
- →Applies to image and video creative, not text ads
- →Covers creative that was generated OR modified using AI — editing counts, not just generation
- →Two compliant routes: label inside the creative, or use the platform AI label setting
- →Election advertising carries an additional, separate disclosure through campaign settings
The Catch Most Advertisers Miss
Google states plainly that using the AI label setting in its advertising products does not by itself guarantee compliance with any specific regulation, and advises advertisers to seek legal guidance and meet their local obligations independently. Read that carefully, because it shifts the risk. Ticking a box in Google Ads is not a compliance programme. If you are advertising in India, your obligation runs to Indian law and the ASCI Code, and Google's setting is a tool that helps you meet it — not a substitute for knowing what it requires.
ASCI's Risk-Based Framework for India
ASCI's May 2026 draft takes a more nuanced approach than a blanket "label everything" rule, sorting synthetic content into three tiers. This is genuinely useful, because it means routine retouching does not need a disclaimer while the things that actually mislead consumers are either labelled or banned outright.
- →High risk — prohibited outright, labelling does not rescue it: fabricated testimonials, unauthorised deepfakes, and AI-generated authority figures falsely implying expertise (the AI "doctor" endorsing a supplement)
- →Medium risk — labelling required: virtual influencers, AI replicas of real individuals, synthetic product demonstrations, realistic AI-created settings, demonstrations of products that do not exist, and AI-generated sponsored recommendations
- →Low risk — no labelling required: basic image enhancement, colour correction, noise reduction, decorative backgrounds, and accessibility tooling
How to Label Without Wrecking the Creative
ASCI's draft contemplates labels along the lines of "Audio/Video created using AI" or "Audio/Video enhanced using AI", subject to its disclaimer requirements. The practical craft question is where to put it. Our approach: treat the label as a design element from the first draft rather than something stamped on at the end. Place it where it is legible without competing with the hook — usually bottom-left or bottom-right, outside the safe-area overlays that Reels and Stories impose. Keep it consistent across the campaign so it reads as a standard, not an apology. And never place it where a crop for a different placement will remove it, because an ad that loses its label at 9:16 is an unlabelled ad.
The Line That Actually Matters: Enhancement vs Fabrication
Strip away the tiers and one principle runs through both frameworks. Using AI to make a real thing look its best is enhancement. Using AI to create the impression of something that did not happen is fabrication. A cleaned-up photo of a product you actually sell is low risk. A generated video of a customer you invented praising a result you did not produce is prohibited, and no label makes it acceptable. That distinction is worth training into your creative team, because it generalises better than any checklist and it will survive the next revision of the rules.
A Practical Compliance Routine
This does not need to be heavy. For clients we run a short, repeatable process rather than a legal review on every asset.
- →Tag every creative at production time with how AI was used — generated, modified, or not at all
- →Sort each one into the ASCI tiers; anything landing in high risk goes back to the brief, not to legal
- →Apply labels at design stage for everything in the medium tier, checked across all aspect ratios
- →Set the platform AI label where applicable, treating it as a supplement to the in-creative label rather than a replacement
- →Keep a simple register of what was generated and by which tool, so a query months later has an answer
- →Re-check the register when ASCI moves the draft to final, since the medium-risk tier is where the boundaries are most likely to shift
Key Takeaway
The ASCI guidelines were still in draft as of mid-2026 and the final text may move, so treat the tiers as direction rather than settled law and take your own legal advice on anything borderline. But the direction is not in doubt: disclosure of synthetic advertising content is becoming standard across the EU, India and parts of the US simultaneously. Advertisers who build labelling into their creative process now will find the final rules a formality. Advertisers who wait will be retrofitting labels across a live campaign library under time pressure. We are already tagging and tiering creative for every client account we run — if you want a hand auditing yours, get in touch.
