Artificial intelligence is increasingly used in advertising to edit content, create virtual influencers and synthetic voices, personalise messaging, and produce realistic product demonstrations. As AI-generated content becomes more sophisticated, consumers may find it difficult to distinguish between authentic depictions and content that has been generated or materially altered using AI.

When does AI-generated content need to be labelled?

ASCI bases the labelling assessment on the nature of the communication and the risk that consumers may be misled or harmed. Brands should consider the following questions together:

1. Whether the SGC is central to the advertisement’s persuasive messaging, or if it materially influences the consumer’s purchase decisions?
2. If unlabelled, will the SGC and its manner of use mislead consumers about the product’s claims, performance or benefits?
3. Would the absence of a disclosure create a false or misleading claim regarding the benefit or performance of the product/service?

Prohibited content and mandatory disclosure

ASCI distinguishes between prohibited content, content requiring disclosure and uses for which no label is required. Advertisements that are illegal, infringe rights, make misleading claims or otherwise violate the ASCI Code remain non-compliant even if they carry an AI label. Separately, labelling is mandatory where SGC materially influences consumer decisions and its non-disclosure would mislead consumers. This includes advertisements using synthetic influencers or ambassadors; digitally replicating a real person’s face or voice, even with consent, for personalised messaging; depicting fabricated events, settings or situations relevant to product performance; demonstrating a product that does not yet exist; or using synthetic sound effects that are highly relevant to a product’s core features. Where an AI service recommends a product because the recommendation has been sponsored by a brand, the commercial relationship must be disclosed through a label such as “Sponsored by [Brand]”. Other permissible labels include “Audio/Video created using AI” and “Audio/Video enhanced using AI”, although brands may use alternative wording that accurately informs consumers.

When no label is required

ASCI has also clarified that labelling is not required where SGC has no material impact on a consumer’s ability to make an informed choice. This category includes routine editing and minor enhancements, such as colour correction, lighting adjustments, noise reduction, standard blemish removal and copy refinement, provided they do not alter the substance or core claims of the advertisement. It also covers decorative backgrounds and ambient elements unrelated to the product’s actual capabilities or promise; administrative and text uses undertaken in good faith without creating false records; accessibility features such as subtitles, closed captions, translations and accurate audio descriptions; and obviously unrealistic or fantastical content that an ordinary consumer would not understand as depicting reality.

The Guidelines were published on 17 September 2026 and will take effect upon the expiry of three months from that date.

What does this mean for advertisers and digital platforms?

For brands, advertising agencies, influencers and other stakeholders involved in advertising, the Guidelines add a distinct SGC assessment to the advertising review process. The practical task is to identify how SGC has been used, determine whether it affects a claim or purchasing decision, and decide whether the content is prohibited, requires disclosure or falls within the no-labelling category. Stakeholders should therefore consider:

1. identifying the extent and manner in which SGC has been used in an advertisement prior to publication. This may involve tracking edits made to video or audio content and requiring those involved in its production to appropriately flag the use of SGC;
2. determining whether the SGC materially contributes to a claim or is likely to influence a purchasing decision. Stakeholders may refer to the assessment questions and examples in the Guidelines when evaluating whether a particular use of SGC requires labelling. Where the position remains unclear, further legal or compliance review may be appropriate;
3. using an appropriate disclosure where required under the Guidelines and ensuring that its presentation complies with the ASCI Code’s disclaimer requirements, where applicable;
4. ensuring that AI-generated product demonstrations and representations are supported by facts;
5. obtaining appropriate consent before digitally replicating an individual’s likeness or voice; and
6. clearly identifying sponsored recommendations made through AI systems.

These requirements are particularly relevant to digital advertising, where content can be generated, adapted and personalised rapidly and at scale. Compliance processes may therefore need to account not only for the original advertisement, but also for materially different AI-generated versions, personalised variants and sponsored recommendations delivered through AI systems. Advertisers and platforms should also ensure that disclosures remain clear when content is reformatted, redistributed or displayed across different interfaces.

What may require further clarity?

Although the Guidelines provide an assessment framework and illustrative examples, their application may still raise questions about when SGC materially influences a consumer’s decision. The concept is not separately defined, and its application will depend on the nature of the claim, the product or service being advertised, and the likely effect of the synthetic representation on consumer understanding. For example, if SGC used in an advertisement for a fairness cream makes the depicted skin tone appear marginally brighter than the substantiated result, the issue may be whether that alteration changes the consumer’s perception of product efficacy. Similarly, if SGC depicts a vehicle’s braking distance as marginally shorter than its actual braking distance, the representation may require scrutiny even where braking performance varies with external conditions. These examples illustrate that materiality may turn not merely on the extent of the alteration, but on whether it affects a claim or attribute that is significant to the purchasing decision. More complex uses of SGC are therefore likely to require fact-specific legal and compliance review.

A consumer-centric framework

The Guidelines are ultimately consumer-centric. They seek to provide additional information where technology may blur the distinction between an actual event, person or product experience and a synthetic representation. By requiring disclosure where SGC may materially influence consumer understanding, while exempting uses that do not affect informed choice, the framework seeks to promote transparency without creating unnecessary label fatigue.

For advertisers, the practical message is that increasing use of AI must be matched by a proportionate review of what has been generated or materially altered, how consumers are likely to understand the representation, whether the underlying claim is substantiated, and whether consent has been obtained where necessary. That review should lead to a clear determination: whether the content is prohibited, requires an effective disclosure, or falls within the no-labelling category. Labelling is a transparency measure, not a defence to an advertisement that is otherwise misleading, harmful or non-compliant with the ASCI Code. This framework is likely to become an increasingly important part of advertising compliance as synthetic and personalised content becomes more common across digital media.