IAB updates AI disclosure framework as legal mandates take effect across three continents

The Interactive Advertising Bureau has published a second version of its AI disclosure framework, seven months after the first edition, as advertising laws requiring explicit labelling of synthetic content take effect across Europe, Asia and the United States. The update does not reverse the group's earlier recommendations but widens them to accommodate a regulatory landscape that has moved faster than many in the industry anticipated.

A patchwork of legal requirements

When the IAB released its initial guidelines in January, the European Union's AI Act had been agreed in principle but its transparency obligations for general-purpose AI and specific provisions for synthetic content were not yet enforceable. Since then, the Act's phased implementation has begun. Member states are designating national competent authorities, and the rules on labelling AI-generated or manipulated image, audio and video content, Article 50 of the Regulation, are becoming operational. In the United States, a mix of state-level statutes and Federal Trade Commission guidance has created a de facto national standard. Several Asian jurisdictions, including South Korea and China, have also introduced mandatory watermarking or disclosure rules for synthetic media.

The practical consequence for multinational advertisers is a compliance matrix that varies by market. The IAB framework now maps its voluntary labels to these legal requirements. US advertisers can adopt a standardised sparkle icon or a plain-text label, both recognised across state lines. In the EU, however, a common icon intended to harmonise disclosure across the single market has not been finalised. The European Commission is still consulting on the design and technical specifications, leaving companies to improvise or wait.

Adoption has outpaced policy

The framework cites data from IAB and Sonata showing that 83% of advertising executives have used AI in the creative process, a 23 percentage-point increase from 2024. That figure alone explains the urgency. Generative tools for image, video, voice and copy have moved from experimental pilots to daily production workflows in agencies and in-house teams. A separate AdvertiserPerceptions study referenced in the document found that 72% of marketers believe an industry-wide disclosure standard is necessary, an acknowledgement that self-regulation alone is insufficient when legislatures are writing their own rules.

The speed of adoption also reflects commercial pressure. Brands are evaluating AI not only for cost savings but for speed and versioning at scale. A single campaign can now spawn hundreds of localised variants in hours rather than weeks. The framework's authors argue that transparency must not become a bottleneck, but they also warn that inconsistent labelling erodes the very trust the industry says it wants to protect.

Consumer attitudes are not uniform

The guidelines highlight a generational split. According to the cited research, 73% of millennial and Generation Z respondents said an advertisement generated with AI would have no effect on, or would even increase, their likelihood of purchasing. Older cohorts were not broken out in the summary, but the implication is clear: younger audiences, having grown up with filters, deepfakes and synthetic influencers, are less likely to treat AI origin as a negative signal.

Other studies, however, point to a measurable performance penalty. The framework acknowledges that AI-generated ads can depress click-through rates compared with human-produced creative. The reasons are debated: some attribute it to a subtle uncanny-valley effect, others to platform algorithms that may downrank synthetic content, and others still to creative mediocrity when tools are used without sufficient art direction. The IAB's position is that marketers must balance transparency with business outcomes, a formulation that reads as both a pragmatic concession and a strategic hedge.

What requires a label, and what does not

The framework draws a line at consumer-facing synthetic media. Disclosure is expected for synthetic images or video, digital twins of living or deceased persons, synthetic voices in specific situations, and conversational agents deployed in advertising contexts. These categories align closely with the EU AI Act's Annex III and Article 50 definitions of high-risk and transparency-obligation use cases.

Equally important are the exemptions. Standard post-production tasks, colour correction, noise reduction, routine retouching, do not trigger a label. Neither does "clearly stylised or fantastical imagery" where no reasonable viewer would mistake the output for a documentary record. The distinction is intentional. The IAB argues that labelling every algorithmic touchpoint, from an auto-masking tool in Photoshop to a background blur in a video call, would desensitise audiences and dilute the signal for genuinely deceptive content.

The icon problem

A single, recognisable symbol would simplify compliance and consumer education. The US advertising ecosystem has coalesced around a sparkle icon, a small, four-pointed star, promoted by the Coalition for Content Provenance and Authenticity (C2PA) and adopted by major platforms. In Europe, the process is more fragmented. The European Commission's AI Office is responsible for developing the common icon mandated by the AI Act, but the timeline has slipped. Industry sources suggest a finalised design may not appear before the first quarter of 2027, well after the first wave of enforcement actions.

Until then, EU advertisers face a choice: adopt the US sparkle icon voluntarily, create their own labelling system, or wait for the official mark. The IAB framework stops short of recommending a specific interim solution, instead advising members to ensure whatever label they use is "clear, conspicuous and accessible", language that mirrors the Act's own wording.

Business risk cuts both ways

Giegerich's statement captures the tension. Over-disclosure risks "label fatigue", a phenomenon documented in nutrition labelling and cookie consent banners where users learn to ignore ubiquitous warnings. Under-disclosure invites regulatory sanctions and reputational damage. The EU AI Act provides for fines of up to 3% of global annual turnover or 15 million euro, whichever is higher, for violations of transparency obligations. For a multinational advertiser, the cost of non-compliance dwarfs the production budget of the campaigns in question.

There is also a competitive dimension. Platforms are beginning to enforce their own disclosure policies. Meta, Google and TikTok now require advertisers to self-declare AI-generated content and may attach automatic labels. An advertiser who discloses proactively avoids the platform's generic badge, which may carry a stigma the brand would rather avoid. The IAB framework is, in part, a play for industry control over the visual language of disclosure before platforms impose their own.

The framework is available on the IAB website. The European Commission's AI Act transparency obligations are detailed on the Commission's press portal and the EUR-Lex database.