In every meeting room today, someone asks: should we put an AI label on it? One camp says honesty lasts longest, so label everything. The other camp says a label can just as easily be a dark red flag? Stay silent and hope.
Both camps see the truth on their side. The truth is: they are both right.
One experiment, two effects
Koning and Voorveld looked into it. Their experiment appeared in 2025 in the Journal of Interactive Advertising. They showed 304 people an ad made with generative AI. One half read that it was AI, the other half did not. Then they measured what happened to trust, and why.
The label did two things at once. It made people think about what AI is and can do (the study calls this conceptual AI knowledge). And it made them more alert to the fact that someone wants to sell them something (attitudinal persuasion knowledge).
That alertness had another result. It raised trust too. Because then honesty may not last longest, but it does last a little longer. The same label that warned people: someone is trying to sell me something here, also produced more trust.
Which reaction wins out in the mind of the average consumer, the study does not say. It doesn't need to. The point is that both exist.
Why that undermines the rule "just be transparent"
"Transparency builds trust" could just as well appear in the Cosmopolitan horoscope. It seems so obvious that nobody questions it. This study shows something different: a label sets a thought in motion, and that thought determines what happens next.
Take "made with AI" under a bank's ad. One reader thinks: a machine wrote this, nobody cared. Another thinks: at least they say so upfront, they don't hide their sales pitch. Same label. The brand around it decides which thought wins out.
Awkward news for anyone who likes a single truth. Unless the truth is that well-made campaigns always win. A well-made campaign with a clear disclosure gives people a reason to read that disclosure as honesty.
What you do with it
Maybe, to begin with, we should stop treating an AI disclaimer as a tick box that we have to cross off for the legal team. Compliance for the sake of compliance
The real question is a brand question. It can't be answered with research or a survey. Ask people whether they want to know and they guess what you'd like to hear. So start with yourself. What is your message? How do you want to treat the people you speak to?
And then there is a technical matter: what exactly do you label? Generative AI that made the ad and AI-assisted creation are two different things.
What the CMO should remember
"Should we disclose AI?" is the wrong question. The right one is: what do people think of us when they notice we used AI? What do they think if we deliver this message with or without an AI label? That is a brand question, and it is exactly the same as the one you answered when asked "may we retouch that portrait of a real employee with Photoshop".
Source
Koning, B. & Voorveld, H.A.M. (2025). Disclaimer! This content is AI-generated: How AI-disclosures influence trust in advertisements and organizations. Journal of Interactive Advertising, 25(3), 240-253. dare.uva.nl
