IAB updates standard for disclosing AI in ads

The IAB released Version 2 of its “AI Transparency and Disclosure Framework” today, updating its guidance for when advertisers, agencies, publishers, platforms, and technology companies should tell consumers that AI was used in advertising and marketing content.

The framework draws a line between AI that helps create content and AI that changes what consumers could reasonably believe is real. IAB calls for disclosure when AI materially affects authenticity, identity, or representation, while rejecting the idea that every use of AI needs a label.

The update comes as that distinction carries more regulatory weight. Since IAB released the first version in January, disclosure requirements have taken effect in several markets, including California, New York, South Korea, and the European Union. IAB’s framework provides the advertising industry with a common approach to making disclosure decisions across markets with different requirements.

What gets disclosed

Under the framework, realistic synthetic content that could affect how consumers understand what they’re seeing generally warrants disclosure. That includes images and videos generated from prompts, some synthetic voices and avatars, digital twins of deceased people, and digital twins depicting living people in fabricated situations outside standard brand endorsements.

The same principle applies to conversational AI. Chatbots and assistants should be identified as AI when consumers could reasonably mistake them for human representatives.

The framework takes a different approach to AI that is used behind the scenes or in ways unlikely to confuse consumers about authenticity. Routine post-production, internal workflows, text and copy, standard audio enhancement, background music, generic synthetic voices, and obviously cartoon or stylized avatars don’t automatically require disclosure. Some authorized synthetic voices and digital twins used in conventional endorsements also fall outside the automatic-disclosure category.

That makes the nature of the AI use more important than the simple fact that AI was involved. A marketer using generative AI to draft copy faces a different disclosure question from one using it to create a realistic video of a person doing something that never happened.

Consumers want disclosure in some cases

IAB’s approach draws partly on consumer research conducted with Sonata Insights for the original framework. The research found mixed reactions to AI in advertising, with some consumers viewing its use in creative work positively and others considering it inauthentic.

More than half wanted brands to disclose when an ad was fully generated by AI or contained AI-generated imagery or video. The finding supports targeted disclosure around synthetic content without establishing a consumer demand to label every marketing use of AI.

IAB’s framework provides industry guidance for marketers running the same creative across jurisdictions with different rules.

New York’s synthetic performer law took effect in June, while California’s SB 942 and Article 50 of the EU AI Act took effect Aug. 2. South Korea also introduced AI labeling requirements earlier this year.

For U.S. advertising, the IAB framework provides two disclosure options: a standardized sparkle icon or clear text. Those are IAB recommendations and don’t override applicable legal requirements.

The EU takes a different approach. Article 50 requires disclosure for covered AI-generated content and deepfakes, but doesn’t mandate a specific icon. A voluntary Code of Practice provides guidance on the design and placement of disclosures, while a common EU icon remains unsettled.

The growing patchwork turns AI disclosure into another classification problem for marketing operations. Knowing that AI touched an asset isn’t enough. Companies increasingly need to track what AI did, whether it altered the authenticity or identity represented in the content, where that content will appear, and which disclosure requirement applies.