India’s new synthetic-ad guidance asks what the image makes people believe

ASCI’s September final guidance has a future effective interval and focuses attention on what a synthetic element makes consumers believe.

An AI label can explain how an advertisement was made while leaving its central product promise misleading. That is the practical distinction in India’s newly released self-regulatory guidance on synthetic advertising. For an agency, the useful work begins before choosing a label: identify what the audience could reasonably infer from the generated element and whether the advertiser can support that inference.

This source-based analysis was checked on 16 October 2026. The home-fixture exercise below is fictional and unrun. It is not legal advice, an ASCI approval or a report of an actual advertiser’s conduct. No advertising asset was generated or distributed for this article.

Final guidance, with a future effective point

ASCI’s [final-guidelines press release is dated 29 September 2026](https://www.ascionline.in/wp-content/uploads/2026/09/Press-Release_Guidelines-for-Responsible-Labelling-of-Synthetically-Generated-Content-in-Advertising.pdf). It says the guidance takes effect three months after publication. On the date of this analysis, that announced interval has not elapsed. The release should therefore be described as final guidance announced ahead of effectiveness, not as a newly enforced rule already governing every advertisement in India.

The [May consultation release](https://www.ascionline.in/wp-content/uploads/2026/05/AI-Draft-Guidelines-1.pdf) belongs to an earlier stage: it was dated 12 May 2026 and invited comments until 13 June. Its draft risk framing should not be substituted for the final wording. For an agency updating a review checklist, retaining the document stage matters as much as retaining the date. A saved consultation summary can otherwise become a misleading statement of current policy.

The [published ASCI Code and guidelines](https://www.ascionline.in/the-asci-code-guidelines/) separate prohibited content, material uses requiring labels, uses not requiring labels and execution of disclosures. They identify misleading product representations, fabricated testimonials and unauthorised likeness or copyrighted material as problems that labels do not cure. They also distinguish routine or non-material changes from synthetic elements affecting consumer understanding. This is a self-regulatory source, not a universal legal clearance for a campaign.

ASCI’s [March 2025 AdNext AI research announcement](https://www.ascionline.in/wp-content/uploads/2025/03/AI-report-press-release.docx-1.pdf) provides earlier context about industry governance concerns. Its described method combined practitioner and thought-leader interviews, focus groups and other research; it was not presented as a representative census of all Indian advertisers. It is useful background for asking who owns review decisions, not a basis for inventing a national adoption rate or assuming every agency has the same capabilities.

Review the implication, not just the production technique

A useful proposed review starts with the consumer-facing proposition. What does the communication invite someone to believe about the product, the place, the person or the experience? Write that proposition plainly before discussing the generation tool. A realistic image may make an implied claim that never appears in the copy, so the visual and the words should be examined together.

Separate a correction from a disclosure. If the asset depicts an unsupported product feature, the first task is to correct or remove that depiction. A note explaining that AI was involved should not become a way to keep the same unsupported promise. The reviewer should record which product evidence supports the final depiction rather than accepting a production team’s confidence in the prompt.

The final Code’s examples make material context important. Generated settings or sound relevant to product capability can require different treatment from decorative or routine editing. The proposed agency question is therefore specific: would this generated change alter a reasonable understanding of what is being offered? The answer should refer to the actual composition, not a blanket rule that every background is harmless or every automated edit is deceptive.

Give the decision a named owner. The product team can verify what exists; the creative team can describe what changed; the responsible reviewer can assess the complete claim and its disclosure. None of those roles should silently stand in for all the others. Preserve the original image and the reason for a rejected version so that a later handoff does not accidentally reintroduce it.

Fictional exercise: storage hardware in a proposed kitchen

Imagine a fictional Pune home-fixture retailer commissioning an advertisement for a wall-mounted storage rail. The exercise has no real client, sales data or generated image. Its proposed inputs are an authorised photograph of an existing rail, an approved product description and an imagined kitchen setting. The setting is an illustration; it must not be described as a completed customer installation or a verified local property.

In the proposed first review, the designer explains which parts of the composition would be synthetic. The product reviewer checks whether the rail’s fittings and supported accessories remain the same as the real item. If an imagined cabinet conceals a necessary mounting component, the problem is not merely the absence of an AI label. The composition may leave a purchaser with an incomplete understanding of installation.

The exercise includes a deliberately unsuitable variation: the proposed image depicts an accessory that the retailer does not sell. The team would reject that version rather than add a small disclosure and retain it. The worksheet records the reason as a product-representation issue. This is an invented decision scenario, not evidence that any image model produced the error or that an actual retailer used it.

A different proposed variation keeps the photographed product intact and changes only the surrounding room. Reviewers would examine whether the room implies an available service, a particular installation outcome or a feature not supported by the product record. They would then choose the appropriate treatment for the actual communication. The exercise does not pre-approve that variation merely because the object itself was photographed.

For the final hypothetical release, the agency would inspect the advertisement at the intended phone size with the selected disclosure and offer copy together. A caption that cannot be read in the actual placement has not completed the communication task. The agency should retain the exact reviewed version, its supporting product record and the person responsible for release, with unresolved decisions still visibly held.

Turn a policy update into a usable review record

Keep a policy-version field in the working brief. It should distinguish the final guidance, the earlier consultation and the date on which the team checked the source. Record the stated effective interval without assuming that the day an agency updates its checklist is also the day a rule becomes effective. Future publication requires a fresh check for changes, clarifications or other applicable requirements.

Use clear internal outcomes such as correct the product depiction, obtain the missing permission or review the disclosure in context. These are proposed human decisions, not official ASCI classifications or legal conclusions. They help a producer understand what to change and preserve work that can safely be reused. A vague instruction to make the advertisement compliant gives little practical direction.

For international work, keep the market-specific record attached to the actual creative. A decision for an Indian placement should not be copied automatically into another jurisdiction, and another market’s caption should not be assumed sufficient here. This article has not compared every applicable law, platform rule or sector restriction. Its bounded contribution is a source-backed method for exposing the decision that needs qualified review.

The final guidance is useful because it pushes attention towards consumer understanding rather than a superficial AI badge. An agency preparing now can improve its evidence trail before the announced effective interval ends. The goal is an advertisement whose product depiction, origin explanation and permissions can all be examined honestly, including when the correct outcome is to withhold a proposed version.