The IAB AI Transparency and Disclosure Framework is the Interactive Advertising Bureau's industry guidance on when and how advertisers should tell consumers that AI was involved in an ad. Its central rule is that a label is needed only when AI materially affects authenticity, identity or representation in a way that could mislead consumers, not every time AI touches a piece of creative.

TL;DR: IAB published the first version on 15 January 2026 and Version 2 on 18 August 2026. The framework names specific triggers, such as prompt-generated images and video, certain synthetic voices, digital twins and chatbots that could pass for a person, and lists routine uses that should not automatically need a label. US advertisers can disclose with a standardised sparkle icon or a clear text label, and IAB says binding regulation takes precedence over the framework.

What is the IAB AI Transparency and Disclosure Framework?

The framework is voluntary guidance for brands, agencies, publishers, platforms and technology partners on disclosing AI involvement in consumer-facing advertising and marketing content. IAB, the US trade association whose membership comprises more than 700 media companies, brands, agencies and technology firms, launched it on 15 January 2026 as its first framework for AI disclosure in advertising.

Version 2 keeps the original's "risk-based, materiality-driven approach" and turns it into practical protocols for when and how AI involvement should be disclosed, according to IAB's V2 page. It covers text, imagery, video, audio, synthetic voices, digital twins and AI-powered consumer interactions.

The V2 document applies to paid, sponsored or brand-controlled content, and it does not extend to organic social media, editorial journalism or entertainment unless those assets are used as advertisements. It states that the framework "does not carry enforcement authority".

When does the IAB framework say AI use must be disclosed?

Disclosure is triggered when AI creates a material risk that consumers will be misled about what they are seeing, hearing or interacting with. The V2 document defines materiality as AI use that could mislead a reasonable consumer about what is authentic, factual or human-created, adding that "the threshold is consumer impact, not the volume of AI tools used". In its V2 announcement, IAB lists the high-risk cases as:

images and videos generated from prompts
some synthetic voices
synthetic avatars
digital twins of deceased individuals
digital twins of living individuals in fabricated situations outside standard brand endorsement
AI chatbots or assistants that could be mistaken for human representatives

The January announcement clarified the edges: new statements from deceased people's synthetic voices need a label "even with estate authorization", while authorised, scripted endorsements by living people generally do not.

What does not need an AI label under the IAB framework?

IAB is explicit that many everyday uses should not automatically require a label. Its V2 announcement lists routine post-production, internal workflows, clearly fantastical imagery, authorised synthetic voices of real individuals, generic synthetic voices, background music, standard audio enhancement, digital twins in standard brand endorsement, obvious cartoon or stylised avatars, and text or copy.

The reasoning concerns the value of the label itself. Caroline Giegerich, VP, AI at IAB, said at the V2 launch that "labeling everything teaches consumers to ignore labels and could negatively impact advertisers". She described the working group's task as avoiding two failure points: "Under-disclosure leaves consumers at risk of being misled. Over-disclosure could risk negatively impacting advertisers."

A label is also not a licence. The V2 document says fabricated testimonials, false claims and unauthorised use of a person's likeness or voice are prohibited regardless of AI involvement, "and disclosure does not cure them".

How would the framework apply inside an agentic workflow?

A worked example shows where the line falls once agents trade media on a team's behalf. Alkimi, an enterprise platform for agentic advertising, runs a Gym where buyers and sellers build their own agents and set the rules those agents work to before any real money moves. A buyer there could write the framework's triggers into its agent's rules, so that a seller's offer built on a prompt-generated, photorealistic video is flagged for a label, while an agent drafting copy or resizing assets is left alone.

The test is the one a human creative lead would apply, but it has to be written down as a rule in advance.

How should advertisers label AI-generated ads?

The framework uses two layers of disclosure, according to IAB. The first is consumer-facing: standardised text labels or visual cues such as watermarks, badges or icons, interactive elements such as tap or hover information icons, and placement next to the creative rather than on it.

The second layer is machine-readable metadata using C2PA content credentials. The V2 document says this includes IAB-specific fields that record whether disclosure is required and whether a label was applied, and that advertisers may meet the consumer-facing requirement by encoding that metadata and relying on platforms that reliably render labels.

Version 2 also settles the US choice of label. IAB says advertisers in the US can use a standardised sparkle icon or a clear text label, either of which fully satisfies the disclosure requirement, and the framework includes accessibility requirements for the standardised label.

How does the framework relate to the EU AI Act and US state laws?

The framework is self-regulation and does not replace the law. IAB says it "is designed to represent the advertising industry perspective, not replace state and international regulation, which takes precedence over industry self-regulation" (IAB).

IAB's V2 announcement sets out what changed after January 2026. New York's synthetic performer law, signed in December 2025, took effect in June 2026, and California's SB 942, which introduced metadata and labelling requirements, took effect on 2 August 2026, the same date Article 50 of the EU AI Act went into effect.

IAB also notes that Article 50 does not prescribe a specific icon. For the wider European picture, see this explainer on the EU AI Act and agentic advertising.

Why did IAB choose a materiality test instead of labelling everything?

IAB's own research points to a perception gap. In The AI Ad Gap Widens, from IAB and Sonata Insights, 82% of ad executives believed Gen Z and Millennial consumers felt positive about AI-generated ads, against 45% of those consumers who actually did, a gap that widened from 32 points in 2024 to 37 points in 2026.

The same study found that disclosure tends to help with that audience. 73% of Gen Z and Millennial respondents said knowing an ad was made with AI would increase or make no difference to their likelihood to purchase, and more than half wanted disclosure when an ad is 100% AI-generated or uses AI images or video.

Practice has not caught up. IAB found that 89% of advertisers who have used generative AI to create ads disclose at least sometimes, but less than half always do. The survey covered 505 US Gen Z and Millennial consumers and 104 US ad executives.

Who is accountable for AI disclosure decisions?

The advertiser is. The V2 document says accountability rests with advertisers, brand and/or agency, "who bear ultimate responsibility for disclosure assessments and ensuring consumers receive appropriate disclosure".

That holds when an AI agent selects, assembles or approves the creative. A buyer still needs to know afterwards which assets used AI, whether any met a trigger and which label was applied, which is why a buying agent's decision log should record those choices alongside the brand safety checks it already runs.

How can advertisers put the framework into practice?

The V2 document sets out a phased roadmap: assign an AI Disclosure Lead within 60 days, establish governance foundations in the first six months, pilot disclosures and begin building technical infrastructure through month twelve, and scale automated systems with ongoing monitoring within 24 months. A checklist drawn from the framework:

Name an AI Disclosure Lead and list every point where AI touches creative.
Apply the materiality question asset by asset: could a reasonable consumer be misled about authenticity, identity or representation?
Check each asset against the named triggers and non-triggers before launch.
Choose a label method per format, the sparkle icon or a clear text label in the US, and meet the accessibility requirements.
Capture C2PA metadata, including whether disclosure was required and whether a label was applied.
Check local law first, since binding regulation takes precedence over the framework.
Write the disclosure rules into briefs and agent mandates, and keep a record of each decision.

Where does a shared deal record fit?

Disclosure decisions are easier to defend when both sides of a deal can see the same record of what was agreed. In the Alkimi Marketplace, humans keep approval rights at the deal level, and the DealSheet is written at agreement, timestamped and held by both parties, so either side can retrieve the agreed terms independently. Whatever tools a team uses, the IAB framework works best when the disclosure decision is made before launch and written down where buyer and seller can both find it.

Frequently asked questions

Is the IAB AI Transparency and Disclosure Framework legally binding?

No, it is voluntary industry guidance without enforcement authority. IAB says state and international regulation takes precedence over it.

Does every AI-generated ad need a label?

No, IAB calls for targeted disclosure rather than universal labelling. A label is needed only when AI use materially affects authenticity, identity or representation in a way that could mislead consumers.

What is the IAB sparkle icon?

It is the standardised visual indicator Version 2 offers as one way to disclose AI involvement. In the US, a clear text label is the alternative, and IAB says either fully satisfies the requirement.

Do AI chatbots in ads need to be disclosed?

Yes, when they could be mistaken for a human representative. Both versions name chatbots and conversational agents that simulate human interaction as cases needing disclosure.

Does AI-written ad copy need a label?

Not automatically, as IAB lists text and copy among uses that should not by default require a label. Advertisers remain responsible for the accuracy of every claim, whoever or whatever drafted it.

Is the framework the same as IAB Tech Lab's AAMP?

No, the disclosure framework is IAB guidance on consumer-facing creative. AAMP is IAB Tech Lab's Agentic Advertising Management Protocols, which concern how agents in advertising work together.

IAB's AI Transparency and Disclosure Framework Explained for Advertisers