The IAB published Version 2 of its AI Transparency and Disclosure Framework on 18 August, seven months after the original January release — and the interval explains the update. In those seven months, AI disclosure went from proposal to enforcement in three jurisdictions: New York’s synthetic performer law took effect in June, and both California’s SB 942 and Article 50 of the EU AI Act became enforceable on 2 August. South Korea’s AI Basic Act joined them earlier in the year. What was a voluntary best-practice conversation in January is now a compliance question with statutes attached.
The framework’s core move is a line most operators will find workable: disclose when AI materially affects authenticity, identity or representation — not whenever AI touched the workflow. Fully AI-generated images and video, synthetic avatars, digital twins of deceased people, digital twins of living people placed in fabricated situations, and chatbots a consumer could mistake for a human representative all sit on the disclose side. Routine post-production, internal workflows, clearly fantastical imagery, generic synthetic voices, background music and AI-assisted text sit on the other. For US campaigns, the framework offers a standardised sparkle icon or a clear text label, with accessibility requirements attached.
Just as notable is what the IAB argued against: over-labelling. The framework warns explicitly about label fatigue — as the IAB’s Caroline Gigerich put it, labelling everything teaches consumers to ignore labels. The consumer research underneath supports drawing a line rather than painting the whole surface. More than half of consumers surveyed wanted disclosure on ads that are fully AI-generated or built on AI imagery. Clear disclosure would maintain or increase purchase likelihood for 73% of Gen Z and Millennial consumers, the research cited by the IAB found. Trust, in other words, is not damaged by admitting AI use — it is damaged by being caught concealing it.
01The framework is voluntary; the laws underneath it are not
It is worth being precise about what this document is. The IAB says plainly that the framework carries no enforcement authority — it exists to help the industry self-regulate. But it is scaffolding built over laws that do enforce, and that is its practical value. A publisher that adopts the framework’s disclosure logic is not just aligning with a trade body’s preference; it is building one consistent operational answer to New York, California, Brussels and Seoul simultaneously, instead of running four parallel compliance interpretations. IAB CEO David Cohen framed the stakes simply: trust is foundational to the growth of AI. The commercial version of that sentence is that disclosure discipline is becoming a condition of running AI-assisted advertising at all.
The adoption numbers say this is no longer a niche concern. AI use in creative work is reported by 83% of advertising executives, up 23 percentage points from 2024. 72% of marketers believe an industry standard for AI disclosure should exist. The gap between those two figures — near-universal use, and a standard that until this week was a first-draft document — is exactly the exposure the framework is trying to close.
02Publishers sit on both sides of the disclosure line
Most coverage of the framework reads it as advertiser guidance. Publishers should read it twice, because they hold two distinct positions. First, as content producers: AI-assisted articles, synthetic voiceovers on video, AI-generated imagery in editorial and branded content all run into the same authenticity test, and branded content studios are squarely in scope. Second, as the medium of record: when a non-compliant AI ad runs on your pages, the regulator’s screenshot has your masthead on it. The framework names publishers and platforms among the parties with disclosure responsibilities, and the ecosystem has a long history of compliance obligations rolling downhill from buyers to the sell side via contract clauses. The time to decide what your ad quality policy says about undisclosed synthetic media is before a demand partner’s lawyer decides it for you.
03Why this matters for publishers
| The regulatory floor has already moved | New York, California and the EU are enforcing now — the framework is a map of laws in effect, not a forecast. Any publisher running AI anywhere in content or ad ops has a compliance surface today. |
|---|---|
| Ad quality review just gained a category | Creative that reaches your pages through direct deals or programmatic can now be non-compliant in ways your existing malware-and-misleading checks were never built to catch. Undisclosed synthetic media is a new rejection reason. |
| Branded content studios are squarely in scope | Publisher-produced advertiser content that uses generated imagery, synthetic voice or avatars carries the disclosure obligation, and the publisher's name is on the output. |
| Disclosure is a trust asset, not a tax | The consumer research points the same way as the Perplexity–Time dispute we covered last week: the transparency high ground is worth holding, and the party with the clearest labelling wins the argument when a format is challenged. |
04What publishers should do
05The bottom line
Version 2 is the rare industry framework that arrives after the laws rather than before them, and it is more useful for it: the question is no longer whether AI disclosure becomes mandatory but whether your operation converges on one coherent answer or four improvised ones. The line it draws — label what changes what a consumer could believe is real, and leave the routine tooling alone — is defensible, research-backed and operationally cheap to adopt while it is still voluntary. Publishers who sort their AI use against it this quarter get a trust asset and a compliance posture at the same time. Publishers who wait will do the same sorting later, under an enforcement letter’s timeline instead of their own.