On August 18, 2026, the IAB published version 2 of its AI Transparency and Disclosure Framework, the first update since the original landed in January. MarTech and Marketing Brew covered it the same day, mostly as a compliance story, and it is one. But read the framework as a marketer rather than a lawyer and it says something more specific: for the first time, an industry body has defined when an AI label is material to the consumer, and handed advertisers a standard way to show it. That label is about to appear on a great deal of ad creative. The question the framework does not answer, and the one that will decide whether the label builds trust or spends it, is what happens after the click. If the ad says AI and the page does not, the two are telling different stories, and the visitor is the one who notices.
What changed in version 2
The core of v2 is a materiality test. Disclose AI use only when it materially affects authenticity, identity or representation in ways that could mislead consumers. A synthetic spokesperson, a generated testimonial, a product image that shows something the product does not do: those are material. Routine production AI, the retouching, the resizing, the copy assistance that touches nearly every ad now, needs no label. That is a deliberate narrowing. The January original left many teams unsure whether every asset touched by a model had to be flagged; v2 says no, flag the ones where the AI changes what the consumer would believe.
The second change is the format. In the United States, advertisers can use a standardized sparkle icon or a plain-text label. A common icon matters because a disclosure nobody recognizes is not a disclosure, and the industry has spent two years inventing its own badges.
The timing is not an accident. Between the two versions, enforcement moved from proposal to live law. The EU AI Act's Article 50 became binding on August 2, 2026. California's SB 942 took effect the same day. New York's synthetic performer law, signed in December 2025, has been in force since June. We covered what Article 50 asks of advertisers and what New York's law means for synthetic talent when each landed, so we will not repeat the legal detail here. The IAB framework is the industry's attempt to give those obligations a single, recognizable expression in the ad itself.
The perception gap nobody closes with a label
The framework cites a number worth sitting with. In the IAB and Sonata Insights study "The AI Ad Gap Widens," 82 percent of advertising executives believe Gen Z and Millennial consumers feel positively about AI-generated ads. Only 45 percent of those consumers actually do. That is a 37-point gap, and it has widened from 32 points in 2024.
Two things follow. First, the people deciding how much AI creative to ship are systematically more optimistic about how it lands than the people it lands on. Second, a disclosure label does not close that gap on its own. It makes the AI visible, which is the legal requirement, but visibility is not trust. Trust is built or lost in the seconds after the click, when the visitor checks whether the thing the ad showed them is the thing the page delivers. A sparkle icon on the ad is a promise about honesty. The page is where the promise is kept or broken. Disclosure is a trust problem before it is a legal one, and trust is evaluated across the whole journey, not on the creative alone.
Disclosure is an ad-to-page signal
Here is the practical consequence most compliance checklists will miss. Under a materiality test, the thing being disclosed is the representation: this person is synthetic, this scene is generated, this testimonial was not spoken by a customer. Representation does not stop at the ad. If the ad shows a synthetic spokesperson with a sparkle icon and the landing page shows a real founder with no note, the two are inconsistent about who is speaking for the brand. If the ad's hero image is labeled as generated and the page reuses the same image unlabeled, the page has quietly dropped a disclosure the ad made. If the ad is honest about AI and the page is silent, a visitor who cares about the label, and more than half of the young consumers in the IAB study do not feel positively about AI-generated ads, sees a brand that discloses when it has to and stops when it can.
This is a new species of alignment drift. We usually describe drift as message or tone diverging between ad and page over time. Disclosure drift is a divergence on a compliance-relevant signal, and it happens the same way: the creative gets a label because the creative team was told to add one, and nobody tells the web team. The page never changes. The two disagree from day one.
The scale problem makes this worse. Three weeks before the IAB published v2, TikTok previewed Symphony Agent, which turns brand and campaign details into a finished TikTok ad through a guided, prompt-based workflow, and Meta's Advantage+ keeps modifying creative after upload. We have written about where TikTok's Symphony tooling is heading and how to audit Meta's AI modifications to your creative. Every generated variant is a new promise and a potential new disclosure, and each one points at a page that was written once, by a person, some time ago. The number of ad-to-page pairs that need congruent disclosure grows with creative volume, and creative volume is what the platforms are optimizing for. That is the governance gap in its newest form.
What we'd check before the icon ships
None of this requires a new department. It requires adding one check to a loop you should already be running.
1. Inventory the materially AI-affected assets. Apply the v2 test honestly: which ads use AI in ways that change authenticity, identity or representation? Those are the ones that carry a label, and those are the pairs that need the check.
2. Confirm the page tells the same story. For each labeled ad, open the destination. Does the page carry a consistent disclosure where the same synthetic element appears? Does the person, scene or claim the ad represented appear the same way on the page? Inconsistency in either direction is the finding.
3. Score the pair, with disclosure as an added check. AdAlign scores each ad against its landing page across four dimensions, Visual Match, Message Match, Above the Fold Continuity and Tone Alignment, with one score out of 10 and the weakest dimension surfaced. Treat disclosure consistency as a fifth check on top of that congruence score: a pair can score well on message and still fail on provenance.
4. Monitor. Labels will be added to creative as the framework is adopted, pages will be redesigned, and platforms will generate variants you did not brief. Drift is the default. Continuous re-scoring is how you catch a label that appears on the ad and never makes it to the page.
Where to start
If your creative volume is outpacing your ability to check what each ad promises and how it discloses, run a free audit on your highest-spend pairs, three analyses, no card required. Teams that want every live ad and page re-scored as creative changes should look at the Pro plan; agencies governing AI creative across a client roster will want the multi-account controls in Agency, described in the agency overview.
This is the third of three shifts this month that push responsibility onto the page. Meta is removing the placement exclusions that let you control an ad's context, so message match has to be placement-proof. An appellate court has ruled that what your landing page links to can count as your own advertising claim. And now the industry's disclosure standard puts a label on the ad that the page has to honor. Three different sources, one conclusion: the click is where your promises get checked.
Frequently asked questions
What is the IAB AI Transparency and Disclosure Framework version 2? It is the August 18, 2026 update to the IAB's industry framework for disclosing AI use in advertising, the first since the January 2026 original. It introduces a materiality test for when disclosure is required and, in the United States, a standardized sparkle icon or plain-text label for making it.
When do I have to disclose AI in an ad under the materiality test? When AI materially affects authenticity, identity or representation in ways that could mislead consumers, such as a synthetic spokesperson, a generated testimonial or an image showing something the product does not do. Routine production uses of AI, such as retouching, resizing or copy assistance, do not require a label under the framework.
What is the sparkle icon? It is the standardized visual disclosure the IAB framework offers US advertisers as an alternative to a plain-text label. A common icon is meant to make AI disclosure recognizable across brands rather than each advertiser inventing its own badge.
Does AI disclosure on the ad need to appear on the landing page too? The framework governs the ad, but the representation it discloses does not stop there. If the landing page reuses the same synthetic element without disclosure, or presents the brand's spokesperson or claims differently from the labeled ad, the pair is inconsistent on a compliance-relevant signal. Treating disclosure as part of ad-to-page congruence keeps the story the same on both sides of the click.
Which laws does the updated framework respond to? The EU AI Act's Article 50 transparency obligations, binding since August 2, 2026; California's SB 942, effective the same day; and New York's synthetic performer disclosure law, signed in December 2025 and in force since June 2026.