On 18 August 2026, MarTech reported that the IAB's updated framework anchors AI disclosure in a single question. Does the AI change what the viewer believes to be real?
That is a sharper line than the industry has been drawing. Most disclosure debates so far have assumed a binary. Either an ad used AI or it did not, and audiences deserve to know either way. The IAB's move rejects that binary. Disclosure attaches to authenticity and identity, not to tooling.
Why A Blanket Label Was Never Going To Work
Modern campaigns run on AI at almost every step. Copy variants are generated and tested at scale. Static assets are resized into dozens of placements by model-driven tooling. Voiceovers are localised, background plates are cleaned up, product shots are relit. If every touch of a model triggered an "AI generated" chip in the corner of the creative, the label would appear on nearly everything and mean nearly nothing. Warnings that fire on all traffic stop being warnings.
The IAB framework, as MarTech describes it, sidesteps that failure mode by asking a different question. It looks at the persuasion, not the pipeline. When a synthetic person, place, or claim is doing the convincing, the viewer is owed disclosure. When the model is doing production plumbing, they are not.
That distinction is defensible. It is also operationally useful, which is rarer.
The Two-Tier Workflow Advertisers Should Build
Creative operations teams should stop trying to design one governance path for "AI in advertising" and build two.
The light path covers everything that leaves the source material's meaning intact. Resizing a hero image for social placements. Generating twenty headline variants from an approved message. Localising a voiceover into five markets from a signed-off script. Removing a boom mic from a plate. Colour-matching a product shot to a brand guideline. In each case a human authored the claim and a model executed the production. The IAB threshold, as reported, does not require disclosure here, because the viewer's understanding of what is real has not shifted. Light-path assets still need brand safety review, rights clearance, and legal sign-off. They do not need a synthetic-media label.
The heavy path covers everything that manufactures identity, presence, or evidence. A face that does not belong to a real person endorsing the product. A real spokesperson's likeness placed in a scene they were never in. A location composited to look like documentary footage. A demonstration of the product doing something the product did not do in camera. Anything that answers yes to "would a reasonable viewer think this happened" gets a disclosure treatment, a heavier legal review, and a documented provenance chain. C2PA credentials, watermarking, and metadata should travel with the file, not sit in a separate compliance folder.
Two paths, one intake. The decision about which path a piece of work belongs on should be made at brief stage, not at delivery. That is the operational shift.
What Buyers And Platforms Should Do Next Week
Media buyers should ask agencies and production partners a single procurement question. Show me your intake form and show me where the authenticity flag sits. If the answer is a shrug or a promise to figure it out per campaign, that is the risk. Governance that lives in individual producers' heads does not scale, and it does not survive an audit when a synthetic asset slips into the light path by accident.
Platforms should read the IAB line as cover to stop demanding blanket AI labels that punish legitimate production workflows, and start demanding provenance on the assets that actually mislead. Meta, Google and TikTok have all shipped some version of an AI disclosure toggle. Most of them are calibrated to the tooling question, not the authenticity question. That gap will close, and the buyers who moved first will not have to retrofit their libraries.
Brand-side marketing leaders should stop treating AI disclosure as a communications risk to be managed by the legal team after the fact. The IAB has told the industry where the line is. Everything upstream of the line is a creative operations problem. Everything at or across the line is a disclosure and provenance problem. Those are different disciplines with different tools, and pretending they are one workflow is why creative teams keep getting caught out.
The IAB has done the useful work of separating production from persuasion. The advertisers who wire that distinction into their intake, their asset management, and their contracts this quarter will be governed by a rule that already fits how they make ads. The ones who wait will be governed by whatever the platforms decide next.