Sometime before your next campaign ships, someone in the room is going to ask whether you should tell people AI helped make it. I hear this almost every day. That’s the AI disclosure question, and most brands are filing it under legal. Wrong drawer. The legal part is mostly decided. Regulators and platforms spent the past year deciding it for you. What’s left is the brand part: when the label shows up next to your work, what does it say about you?
We’ve watched two versions of this play out over the past year. One brand disclosed its AI and got buried for it. Another swore off AI entirely and turned that promise into its best-performing content in years. Same transparency. Opposite outcomes. The label didn’t decide either one. The work behind it did.
AI disclosure just became the default
For a couple of years, brands had a third option between “proudly AI” and “proudly human”: say nothing. Use the tools, pocket the savings, let the audience assume whatever it assumed. That option is expiring on a schedule you can look up.
Since August 2, 2026, Article 50 of the EU AI Act has required AI-generated content to be marked in a machine-readable way, deepfakes to be disclosed to people on first exposure, and chatbots to identify themselves as software. If you sell into Europe, that’s not a trend piece. That’s your compliance calendar. New York got there first on one front: in December 2025, the governor signed a law requiring advertisers to conspicuously disclose when a synthetic AI performer stands in for a human being. Google reworked its ad policies in July 2026 so advertisers can carry AI labels right inside the creative, because regulators in multiple markets now expect them. YouTube has required disclosure of realistic synthetic content since 2024. Meta stamps “AI Info” on posts (see photo above). The IAB published a transparency framework built on C2PA provenance metadata, which travels inside the file itself, quietly, whether or not anyone writes a caption.
Add it up and the direction is obvious. Disclosure is becoming infrastructure, like the nutrition panel on a cereal box. Nobody voted on whether they wanted one. It just became part of the packaging, and then everyone learned to read it.
Which means the real question has changed shape. You don’t get to decide whether people find out. You get to decide what finding out means.
Two labels, two outcomes
In late 2025, Valentino released a fully AI-generated campaign for a handbag: surreal, kaleidoscopic visuals, limbs morphing into logos, bodies pouring out of leather. The house was upfront about how it was made. It didn’t matter. Commenters called the work “cheap” and “lazy,” and the critique that stuck wasn’t about the technology at all. It was about the mismatch. This is a house that charges couture prices because human hands spent hundreds of hours on the product. An ad assembled by a model in minutes read as the opposite of the thing being sold. Coca-Cola walked into the same wall with its AI-generated holiday spots. Disclosure didn’t soften either backlash, because disclosure wasn’t the problem.
Now the other direction. In October 2025, Aerie pledged to keep its marketing “100% real,” promising no AI-generated bodies and no AI-generated people, ever. The pledge became the brand’s most-liked post in a year. By spring 2026 the company had Pamela Anderson fronting the relaunch and was reporting double-digit awareness gains alongside a 23% sales lift in the quarter after the pledge. Take the brand-reported numbers with the usual salt. The shape of the result still holds: a promise about process became the most effective creative the brand had run in years.
Here’s what those two stories share, and it’s not the obvious thing. Honesty didn’t sink Valentino, and honesty didn’t lift Aerie. Fit did. Valentino’s label exposed a contradiction between what the house sells and how the ad was made. Aerie’s pledge sat on top of a decade of unretouched photography, a promise the brand had been keeping since 2014. One label said “we cut a corner you’re paying for.” The other said “we’re still who you thought we were.”
A label is a receipt
Strip away the policy language and an AI disclosure does one simple thing: it tells people where your money and effort went. That’s why identical labels produce opposite reactions on different brands. A receipt means nothing on its own. It means everything next to the price you charged.
Luxury sells hours and hands, so an AI label on a luxury film reads as an admission. Aerie sells the absence of manipulation, so an AI-generated model wouldn’t just be off-brand, it would dissolve the brand entirely. A software company shipping a product demo made with AI tooling? Nobody blinks, because nobody was buying craft from them in the first place. The label lands on top of an existing promise, and the reaction is the audience auditing that promise in real time.
The research on consumer sentiment keeps pointing the same direction: when people believe a brand leaned hard on AI, trust drops, and the effect is strongest for brands that built their story on authenticity or craftsmanship. People can’t reliably spot AI work anymore. What they’ve learned to punish is the feeling of it, that thin, averaged, nobody-home quality the internet now calls slop. We wrote earlier this year that the anti-AI design wave isn’t a backlash, it’s a standards reset, and the disclosure era is that reset with paperwork. People don’t read the label to learn about your workflow. They read it to find out whether you still care.
There’s a flip side that gets less attention, and it’s good news if your work is actually human. In 2026, brands are getting accused of using AI when they didn’t. Illustrators post process videos to prove their hands were involved. Photographers publish contact sheets. Once an audience is primed to see slop everywhere, real craft eats false accusations, and a shrug isn’t a defense. A working disclosure regime fixes that too. The same provenance metadata that outs a generated image can certify an unretouched one. For a brand that genuinely does the work, the receipt isn’t a threat. It’s the proof you’ve been wishing you could show.
The pledge is the expensive option
Which brings us to the tempting move of the moment: the no-AI pledge. Aerie’s numbers make it look like free money. It isn’t. A pledge is the single most expensive form of disclosure, because it converts a workflow preference into a public promise with enforcement risk. We argued this summer that brand honesty is a risk you take, not a message you send. A no-AI pledge is that idea with a legal team. The moment you make it, every asset you ship becomes evidence. One AI-touched image from a freelancer, one vendor who quietly generated the background plates, and the story stops being about tools and starts being about the lie.
Dove could make its 2024 promise to never use AI to represent real bodies because that promise is structural. Real bodies have been the entire brand since 2004. Aerie could follow because unretouched photography was already a decade deep in its identity. In both cases the pledge didn’t add a value. It fenced one that was already there.
If your brand’s story isn’t about realness, human craft, or bodies, a no-AI pledge is a costume. Audiences have seen enough purpose-washing to recognize the move, and the first inconsistency will cost you more than the pledge ever earned. Aerie’s pledge isn’t brave. It’s consistent. That’s the part worth copying.
Your vendors are your disclosure now
Here’s the part that bites brands who think they’ve already made their decision: you don’t control most of the hands that touch your marketing. The photographer’s retoucher, the video house compositing your product shots, the freelancer building social cutdowns, the influencer reading your brief, the stock library feeding your decks. Any of them can introduce AI into your creative without a single person at your company choosing it. Under the new rules, and under your audience’s new scrutiny, their workflow becomes your label.
The EU’s framework makes the point explicit by putting obligations on deployers, not just toolmakers. If a synthetic performer ends up in your ad, “our production company handled that” isn’t a defense a regulator or a comment section will accept. The brand wears it. Every time.
So the disclosure decision isn’t one decision. It’s a policy, and it has to reach your contracts. The brands handling this well right now are doing unglamorous things: asking vendors to declare AI use per deliverable, writing provenance requirements into SOWs, deciding in advance which asset classes can be generated and which can’t ever be. None of it is exciting. All of it is cheaper than explaining to your customers why the “100% real” campaign had a synthetic crowd scene in the background. If you’ve made a promise, your procurement is now part of your brand team. If you haven’t, your vendors are making your promise for you, quietly, one deliverable at a time.
Disclose the hand, not the hammer
So where’s the sane middle? Start with an observation nobody puts in a compliance memo: no brand has ever labeled its Photoshop use, its render farm, or its spellcheck. Audiences never asked for that, because those tools never touched the thing being bought. The outrage arrives only when AI replaces the thing the customer believed they were paying for. The real body in the swim campaign. The couture hand in the fashion film. The lived point of view in the writing.
That’s the test, and you can run it on any asset before the law makes you. Three questions, in order. The promise: what is this audience actually buying from us, effort, realness, taste, speed, price? The proof: if a label appeared on this asset tomorrow, would it contradict that promise or leave it untouched? The price: if we pledged the opposite and swore off AI here, could we keep that promise on our worst-staffed week two years from now? Most disclosure decisions collapse into ten minutes once those three are on the table. The brands that got burned skipped straight to the output and never checked it against the promise.
We’ll go first, since it’s fair to ask. We use AI at Jacob Tyler. It reads, it drafts, it pressure-tests, it does the eleventh revision of a resize nobody should do by hand. It doesn’t decide what a brand stands for, and it doesn’t get final say on work that carries a client’s name, because judgment and taste are the product our clients are paying for. AI can hold the hammer. It doesn’t get to be the hand.
When every ad carries a receipt
Play this forward a few years and the labels themselves stop mattering. Provenance metadata rides inside every file. Platforms surface it automatically. Shopping agents check it before a human ever sees the ad. When every piece of creative carries a receipt, disclosure stops being a differentiator in either direction. The scarlet letter fades. So does the halo. What’s left is the question the label was always standing in for: was this worth making, and does it sound like anyone in particular?
That’s the durable version of the AI disclosure question, and it has nothing to do with Brussels or Albany. Labels reward brands that already know what they’re for, and they expose the ones that hoped nobody would look at the receipt. Strong brands can read their own receipt out loud. Weak brands need the fine print.
So run the test now, before the law runs it for you. Put the label on your next campaign in your head and read the work again. If nothing changes, ship it. If the work suddenly reads as a broken promise, the label isn’t your problem, and neither is the AI. If you’d rather not run that test alone, that’s where we come in.