The Public Drew the AI Disclosure Line. The Industry Framework Drew a Different One

Gallup asked 3,270 American adults where AI belongs in advertising and got back a line, not a mood. 75% accept AI for brainstorming and early drafts when the brand discloses it. 53% accept AI producing the final text, images, or video. And 62% say using AI to recreate a person’s likeness or voice is unacceptable even when the brand disclosed it. The survey ran May 4 to 11, 2026, with a margin of error of ±2.4 points, and Gallup published the results on August 19.

In the same August, the IAB shipped version 2 of its AI Transparency & Disclosure Framework. Its trigger is different in kind, not just in degree: disclose when AI involvement is “materially relevant” to consumer decision-making, and avoid what the framework calls “unnecessary labeling and disclosure fatigue” everywhere else.

The public’s line runs through likeness. The framework’s line runs through materiality. Those two lines cross a marketing organization at different angles, and the wedge between them is now a governance problem with a name on it.

The Public’s Line Runs Through Likeness

The three Gallup numbers form a gradient, and the gradient is the finding.

At the bottom of the ladder, AI as a process tool: brainstorming, early drafts. With disclosure, 75% of respondents accept it. In the middle, AI as the producer of the finished ad: acceptance drops to 53%, a majority, but a thin one. At the top, AI recreating a human face or voice: 62% call it unacceptable, and the phrase attached to that number is what makes it expensive. Unacceptable “even when disclosed.”

That phrase changes what disclosure is worth. On the lower rungs, disclosure functions as a purchase: tell the audience, and most of them accept the practice. At the likeness rung, the purchase fails. Respondents were asked to assume the brand had told them, and a majority rejected the practice anyway. Disclosure does not buy you likeness.

Two more numbers set the ambient temperature. Overall sentiment on AI in advertising sits at 49% negative against 19% positive. And the age curve inverts the intuition that younger audiences, being AI-native, would be AI-tolerant: the 18 to 29 group is the harshest of all, at 66% negative.

The Framework’s Line Runs Through Materiality

IAB Framework v2 asks one question: is the AI involvement materially relevant to the consumer’s decision? If yes, disclose. If no, the framework’s own stated design goal kicks in, avoiding “unnecessary labeling and disclosure fatigue.”

As a piece of standards engineering, this is a defensible choice. Materiality gives a compliance team a single test it can operationalize across thousands of assets. A blanket-label regime would drown the signal: if everything carries an AI label, the label stops informing anyone, which is precisely the fatigue the framework says it is designed to avoid.

But materiality is a judgment about the consumer’s decision, made inside the company. The Gallup data is a measurement of the consumer’s tolerance, made outside it. There is no mechanism forcing those two to agree, and in August 2026 we got the first population-level evidence that they do not.

Where the Wedge Opens

Run the two tests against the same hypothetical assets and the divergence shows up at both ends of the ladder.

Take the bottom rung first. Suppose a brand uses AI to brainstorm and draft copy that humans then rewrite entirely. Under a materiality reading, that involvement is plausibly immaterial to the consumer’s decision, so IAB v2 would counsel no label, in the name of avoiding fatigue. But Gallup’s 75% acceptance for exactly this use was measured with disclosure present. The poll tells us the public accepts disclosed process use. It does not tell us the public accepts silent process use, because that is not what was asked. A framework that removes the label is operating outside the measured zone.

Now the top rung. Suppose a brand licenses a synthetic voice for a voice-over, with consent and a disclosure line in the credits. A materiality analysis could go either way, and a disclosure requirement is satisfiable either way. The Gallup result says the analysis is beside the point for 62% of the audience: the likeness itself is the violation, and the label does not repair it. Materiality treats disclosure as the remedy. The public, at this rung, is saying there is no remedy.

So the wedge is two-sided. At the bottom, the framework may label less than the measured tolerance assumed. At the top, the framework offers a cure the audience has already rejected. A marketing governance team that adopts IAB v2 verbatim imports the materiality line and silently discards the likeness line, without ever making that decision on purpose.

Two Lines Are Still Better Than Zero

None of this is an argument against the framework. Until this August, a brand writing an AI disclosure policy had no population-level measurement of where the public draws the line to calibrate against. We wrote about that vacuum from two directions, the cost clock running on undisclosed AI in marketing and the absence of brand-level disclosure policy for AI content. And the AI-washing incident of May 2026 showed what it costs when marketing claims about AI and operational reality diverge in public.

What changed in August 2026 is that the inputs exist. There is a standard, and there is a measurement. The governance work is no longer inventing a line from nothing. The work is reconciling two lines that do not coincide, and documenting which one wins in which case.

Do This Now

Put a one-page addendum on top of whatever disclosure policy you have, this quarter, with four clauses.

Adopt IAB v2 as the floor, not the policy. The materiality test is your minimum disclosure trigger and your defensible baseline in an industry argument. Nothing in it is your ceiling.

Write a likeness rule that ignores materiality. Synthetic faces and voices get an affirmative senior signoff, every time, regardless of the materiality analysis, because 62% of the measured public rejects the practice with disclosure already in place. Treat likeness as a category decision, never an asset-level one.

Keep the label on process use where the framework would drop it. The 75% acceptance for AI-assisted drafting was measured with disclosure present. If you remove the label citing fatigue, record that you are stepping outside the measured zone, and own the residual risk explicitly rather than by omission.

Price the age curve into audience decisions. If a campaign targets the 18 to 29 band, the ambient negativity is 66%, not 49%. The same asset carries different risk in front of different audiences, and the media plan should say so.

The wedge between the public’s line and the framework’s line will not close on its own. Standards bodies optimize for operability, and polls measure sentiment. The only place the two get reconciled is inside your own policy, in writing, before a campaign forces the question.

---

This analysis synthesizes Americans Aren’t Sold on Businesses Using AI in Advertising (Gallup, Julie Ray, August 2026) and the AI Transparency & Disclosure Framework V2 (IAB, August 2026).

Victorino Group helps marketing and governance teams turn AI disclosure standards and public-sentiment data into enforceable policy. Let’s talk.