The IAB AI Transparency Framework v2: A Buy-Side Disclosure Standard for Video Ad Ops

The IAB AI Transparency Framework is a standard, not a statute

The IAB AI Transparency Framework v2, released on August 18, 2026, gives video ad operations teams a buy-side disclosure standard rather than a new law. It tells advertisers when AI-generated or synthetic content in an ad needs a consumer-facing label, and how that label should travel with the asset through the supply chain.

The original framework launched in January 2026 alongside proprietary research from Sonata Insights. Version 2 refines it with more granular, format-specific guidance now that several AI disclosure laws have moved from proposal to enforcement. The IAB is explicit that the document carries no enforcement authority and is meant to complement, not replace, state and international regulation.

That timing is the point. In the weeks before the release, New York's synthetic performer law took effect in June 2026, California's SB 942 went live on August 2, 2026, South Korea revised its AI Basic Act, and Article 50 of the EU AI Act became applicable on August 2, 2026. The IAB framework gives advertisers one industry baseline instead of a jurisdiction-by-jurisdiction scramble. The IAB's own consumer research sharpens the stakes: more than half of consumers want brands to disclose when an ad is fully AI-generated, and 73 percent of Gen Z and Millennial respondents said clear disclosure would increase or not change their likelihood to buy. Yet 82 percent of advertisers assumed consumers felt positive about AI in advertising, while only 45 percent of those younger respondents actually did, a 37-point perception gap the framework is built to close.

When AI use triggers a label

The framework rests on a single test: does the AI involvement create a material risk that a reasonable consumer is misled about what they are seeing, hearing, or interacting with. Materiality is about consumer impact, not the volume of AI tools in the production chain. If the use could mislead about what is authentic, factual, or human-made, it warrants a label.

The document lists specific triggers. Disclosure is required for images and video generated from prompts, synthetic voices of deceased persons even with estate authorization, synthetic voices of living persons placed in fabricated scenarios, photorealistic AI influencers, digital twins of the deceased, and digital twins of living people shown in events that never happened. Conversational agents get an 'AI-powered' label rather than an 'AI-generated' one. For living talent, the line between a sanctioned synthetic performer and an undisclosed one is exactly why teams need a clearance process {{link}} before a generated face fronts a brand.

The IAB leans on an NYU Stern finding that adding an AI disclosure to an advertisement cut click-through rate by 31.5 percent. That number is the tax the trade body uses to argue against blanket labeling: label everything and you train consumers to ignore labels while quietly paying a measurable performance cost. Targeted disclosure is the compromise. The framework also documents that 76 percent of US adults said distinguishing AI content from human content is extremely or very important to them. The disclosure is therefore a trust instrument, not merely a legal checkbox, and the 31.5 percent click cost is the price of honesty the industry has chosen to measure rather than ignore.

For living talent, the line between a sanctioned synthetic performer and an undisclosed one is exactly why teams need a clearance process synthetic performer clearance before a generated face fronts a brand.

What stays unlabeled

Just as important is the exemption list, because it tells video teams where they can stop worrying. Routine post-production stays unlabeled: color correction, dust and blemish removal, contrast optimization, and upscaling. Removing safety rigging from a stunt sequence counts as standard visual effects practice, not a disclosure event.

The carve-outs are broad. Stylized or fantastical imagery, authorized voice clones used for scripted commercial endorsement, generic synthetic voiceovers with no identifiable person, background music, standard audio enhancement, text and copy generation, and obvious cartoon or animated mascots all fall below the threshold. IAB research with Sonata Insights found that more than half of consumers wanted brands to disclose when an ad was fully AI-generated or used AI imagery or video, which is why the exemption list is drawn so narrowly around human-authored or merely enhanced content. Copy generation being exempt matters, since text is where most production AI spend currently sits.

The practical read for a video team: generating the hero footage from a prompt is a label event, but polishing it in post is not. The disclosure obligation attaches to the act of creation that could mislead, not to every downstream touch of the same asset.

Where the label goes and how it looks

For advertisers in the United States, the framework offers two compliant options: a standardized sparkle icon, the Unicode character U+2728, or a clear text label reading 'AI-generated'. Either satisfies the requirement. The label must render at a contrast ratio of at least 4.5:1 to meet WCAG AA, so it is legible, not a barely visible watermark in the corner. The framework also includes accessibility requirements for the standardized label, so the disclosure is legible to assistive technology, not just visible to sighted viewers.

Placement is specified down to the frame. Video labels belong in the first frame of the creative. Audio labels go before the segment they describe begins. The visible layer is only half the system: underneath sits a metadata layer built on C2PA content credentials, with two custom assertions, com.iab.threshold and com.iab.disclosure, that carry tool identification and timestamps through the supply chain. That same provenance layer is where cross-market disclosure gets complicated {{link}}, because one asset now has to satisfy five different regional rules.

The rollout is deliberately long. Companies get 60 days to designate an AI Disclosure Lead. Months zero to six cover team training and a pre-launch checklist. Months six to twelve are for controlled pilots that measure how disclosure affects performance. Months twelve to twenty-four are for scaling and automation. The framework is asking organizations to build a function, not just add a sticker.

That same provenance layer is where cross-market disclosure gets complicated cross-market AI video disclosure, because one asset now has to satisfy five different regional rules.

How this stacks against the EU AI Act and Google

The IAB document is voluntary and buy-side; the law is not. Under Article 50 of the EU AI Act, applicable since August 2, 2026, providers must mark synthetic audio, image, video, or text outputs in a machine-readable format that is detectable as artificially generated. Deployers of deepfake systems must disclose that content was artificially manipulated, with a lighter obligation for clearly artistic or satirical work. The law prescribes no specific icon, leaving design to a voluntary Code of Practice finalized in June 2026.

Platforms are moving on their own timeline. Google began adding a 'How this ad was made' panel to ads across Search, YouTube, and Discover in July 2026; creative built with Google's own generative tools is labeled automatically, while third-party AI must be flagged through a manual control, and on-ad labels appear in the EU, India, and New York where local law requires them. The existing compliance baseline for this shift is the earlier disclosure checklist {{link}} that many teams already run.

The operating reality is three disclosure layers on one asset: the IAB self-regulatory standard, the binding legal rules in the EU, California, and New York, and the platform-level label from Google or Meta. A governance playbook {{link}} is what keeps those layers from contradicting each other when a campaign spans markets and tools.

The existing compliance baseline for this shift is the earlier disclosure checklist AI video disclosure compliance checklist that many teams already run.

A governance playbook AI video governance playbook is what keeps those layers from contradicting each other when a campaign spans markets and tools.

A disclosure checklist for video ad ops

For a video team, the framework translates into a short operating checklist. Name an AI Disclosure Lead within the 60-day window. Build a pre-launch checklist that decides, per asset, whether the AI involvement crosses the materiality line. Embed C2PA provenance at export so the metadata travels with the file. Pick the label format per market. Run the first pilots as controlled experiments that measure the 31.5 percent question at your own scale.

Documentation is the quiet multiplier. A creative audit trail {{link}} lets every approver see what changed and who approved it, which is also what a regulator or platform will ask for first. For assets using synthetic people, keep the clearance record alongside the provenance file so the disclosure and the consent chain are never separated.

The label is the tax, but over-labeling is its own penalty: it trains consumers to ignore the very signals brands need them to trust. The IAB's v2 is best read as an operating model for targeted disclosure, not a compliance chore. Teams that build the function now will ship AI video that is both faster and honestly labeled.

A creative audit trail AI video creative audit trail lets every approver see what changed and who approved it, which is also what a regulator or platform will ask for first.