At the end of July, Senator Adam Schiff introduced two new bills aimed at advertising. The first would amend the Federal Election Campaign Act to require disclaimers on paid political content from influencers, and the second, the AI Ads Act, would extend the existing ban on impersonating candidates so that it covers AI-generated content too.
Both bills are narrow by design, aimed squarely at political advertising because of what happened during California's primaries, where one gubernatorial campaign paid news influencers hundreds of thousands of dollars for supportive posts while an LA mayoral run leaned on quick-turn AI video ads to promote itself and go after the competition. That said, disclosure rules in digital advertising have a habit of starting narrow and then widening out, so I wouldn't bet on these staying in political advertising for long (nor should we expect that the advertising tactics seen in California will remain confined to political advertising).
Which makes this a useful moment to look at where the commercial side of advertising actually stands on AI disclosure, because the honest answer is that we're in better shape than most people assume, with one specific gap that's very fixable this year.
Start with influencers, since that's where the political bills land first. BBB National Programs' National Advertising Division surveyed 3,720 US consumers aged 18 to 65 for its Influencer Trust Index and found that 87% of them trust company advertisements while only 74% trust influencer ads, which means the channel we all sold as being more trusted than traditional advertising is now viewed as less trustworthy. Payment looks like the specific trigger, because in a June 2025 Clutch survey of 277 consumers, 53% said they trust a product recommendation less when the influencer is paid, and nearly half hadn't bought an influencer-recommended product in the past year.
That isn't the only reading available, and I'd rather flag the counter-evidence than get accused of picking the numbers I liked. Other 2026 research finds credibility going the other way, with 61% of consumers saying influencer content has become more credible, and in a channel this large and this varied I think both findings are probably true of different slices of it.
This is part of an ongoing series by Dave Byrne, TrustRaise founder and BSI Advisory Board Member, about where we find the practice of brand safety & suitability in 2026. You can find all the articles here.
Add AI into the same picture and things get more opaque. Schilke and Reimann ran thirteen experiments for a paper in Organizational Behavior and Human Decision Processes called The Transparency Dilemma, and found consistently that people who disclose AI usage end up trusted less than people who don't, driven by a drop in perceived legitimacy, with the effect holding whether the disclosure was voluntary or required. Advertising-specific research replicates it: across three studies, AI disclosure labels produced less positive attitudes toward the ad, and the damage ran through reduced trust in the advertised brand rather than stopping at the creative, with further work showing it reaching the organization behind the ad as well.
Then IAB's data points the other way entirely, with 73% of Gen Z and millennials saying clear AI disclosures would increase or not change their likelihood to purchase. Sprout Social landed somewhere in the middle with a three-way split of 37% more interested, 37% more distrustful, and 27% either indifferent or unable to tell the difference.
The contradiction mostly dissolves once you notice that the academic studies are testing one actor disclosing against a background where nobody else does, while IAB is testing how consumers feel about labeling as a general practice.
The penalty attaches to the one who admits it. When a label is unusual it reads as a confession, and the reader fills in a reason for the confession themselves, which is rarely a generous one. When the label is on everything, it reads as metadata, in the same way that nobody discounts a product because it carries a nutrition label.
The backlash Hank Green has been dealing with is that mechanism happening to someone in public, and he's about as trusted as a creator gets (Green is a prominent creator who comments primarily on science-related topics, with over eight million followers on TikTok). Viewers of a recent Complexly video flagged a line that sounded like it might have been a leftover AI prompt response, and when Green clarified that the line was an ad lib but that he had used ChatGPT for research on the script, the reaction got louder rather than quieter. He's since apologized and said he'll be posting less for a while.
What's instructive is that he disclosed voluntarily and almost immediately, and it still landed as a confession, because nothing in the environment around him makes AI-assisted research a normal thing to say out loud yet. One post going around during the backlash put the underlying problem well: some creators talk openly about using AI without any trouble at all, others would be torn apart for it, and which camp you fall into looks close to arbitrary.
That has a consequence worth sitting with, because it means the trust cost of disclosure depends on how many other people are disclosing, which makes this a coordination problem rather than a judgment call any individual brand can get right on its own. It also means the cost drops fastest for whichever part of the industry standardizes earliest, which is a decent argument for moving now rather than waiting to see how it shakes out.
The strongest empirical support sits in that same Schilke and Reimann paper, which found the penalty for disclosing is smaller than the penalty for being exposed by somebody else. Disclosing first lets you keep hold of the legitimacy story, and being found out doesn't.
There's a more practical version of the same point, which is that platforms are already running their own detection regardless of what advertisers declare. Google and Meta both apply AI labels off their own signals, including provenance data passed along from other platforms, so the live question for an advertiser is whether you label on your own terms or wait for a classifier, a competitor, or a reporter to do it for you.
IAB's AI Transparency and Disclosure Framework, released in January, is the first comprehensive industry standard for AI disclosure in advertising, and I think it deserves a lot more credit than it's had.
The materiality test is the genuinely good part of it, requiring disclosure where AI materially affects authenticity, identity, or representation in ways that could mislead consumers, and letting routine production tasks, background tools, and clearly stylized creative through without a label. That distinction is the hardest question in this whole area and they answered it well, because blanket labeling would have produced disclosure fatigue, which is exactly the condition under which labels stop carrying any information at all. The two-layer design is right too, pairing consumer-facing labels for comprehension with machine-readable C2PA metadata for verification and supply chain transparency, and the American Bar Association's antitrust newsletter has treated it as serious standards work rather than a press exercise, which strikes me as the fair read.
What the framework hasn't been given yet is the infrastructure that turns a good answer into actual market behavior, and there are two pieces missing.
The first is that there's no way to prove you're using it. The framework is voluntary and adoption depends on individual agencies and platforms, with no register of adopters, no audit path, and no way for a brand to demonstrate conformance to a buyer, a platform, or a regulator, so adoption is currently invisible and therefore can't be rewarded. Anyone who's worked in brand safety has watched this exact movie before, because shared definitions sat around for years before they changed anybody's behavior, and what eventually changed behavior was third-party verification measuring against them and buyers writing them into contracts. A standard without a measurement layer stays a document. Self-attestation in the RFP, vendor measurement, or a certification path would each turn adoption into a visible signal, and honestly any one of the three would get us most of the way there.
The second gap is that creator content sits outside the pipes. The framework and the C2PA metadata layer both operate inside the ad supply chain, and organic creator content doesn't pass through an ad server, so neither the consumer label nor the provenance metadata travels with it reliably. Meanwhile the FTC already treats virtual and AI endorsers under the same rules as human ones, which means the legal obligation exists in the one channel the technical standard doesn't reach, and that channel happens to be the fastest-growing category of spend we have. It's also, as Green just found out, where audiences are currently making the rules up as they go. Extending the materiality test into creator briefs, contracts, and platform-native disclosure tooling is the highest-value thing available to us right now.
There's a timing point too, which is that Schiff's bills are being drafted as we speak. We've already produced a defensible test for separating AI that materially misleads from AI that's just routine production, and offering that to legislators seems more useful than waiting and reacting to whatever gets written without it. Self-regulation and regulation aren't really competing here, because a workable materiality test is the thing that lets a statute be enforceable instead of overbroad or empty.
IAB's own research found that consumer perception ranked as the second-biggest advertiser concern in 2024 and then fell significantly in the 2026 rankings, with advertisers now prioritizing impact on human creativity, implementation cost, and brand authenticity instead, while legal and ethical concerns declined substantially over the same period.
So the framework arrived at roughly the moment advertisers stopped worrying much about how consumers feel about AI in advertising. That gap comes out of the standard-setter's own data, which is what makes it worth raising rather than something to argue about.
Voluntary standards stay voluntary for as long as they move faster than the statutory layer, and right now the statutory layer is moving faster. New York's synthetic performer disclosure requirement took effect on June 9 with penalties of $1,000 for a first violation and $5,000 after that, EU AI Act Article 50 transparency obligations apply in full from August 2 and require machine-readable marking and clear labeling of synthetic content, and enforcement is already running, with the FTC's Operation AI Comply producing more than a dozen actions.
None of that waited for voluntary adoption, and six months on from a genuinely good framework the statutory layer has advanced at about the pace it would have anyway. Self-regulation hasn't yet bought the goodwill it's supposed to buy, and it'll only start doing that once adoption is something you can see and measure.
And a recommendation for IAB: publish a register of adopters. All it needs is a page listing who has committed and to what, which is a low bar for the cheapest useful thing anyone could ship here.
We got the standard right. What we need next is proof of who's using it, coverage of the channel where the exposure is growing fastest, and a seat at the table while the political rules are still being written.