Look. I have been watching this conversation about AI disclosure for two years now, and I keep waiting for the part where someone says the actual quiet thing out loud.
The quiet thing is this: most creators already know whether their work used AI. They know exactly which paragraphs were drafted by a model, which images were generated and then painted over, which music stems were AI-produced before a human touched the mix. They know. They are just not telling you. And the reason they are not telling you is not philosophical confusion about what counts as AI use. It is much simpler than that. They are scared of what honesty costs them.
This is not a framework problem. It is a market problem. And no labeling system, no matter how elegantly designed, can solve a market problem by adding more labels.
I tested this in a small, informal way last spring. I asked twelve independent creators across photography, writing, and music production the same question: if a verified, neutral AI disclosure badge existed and you could apply it to your work, would you? Three said yes without hesitation. Nine said some version of "it depends." When I pushed on what it depended on, the answers clustered around the same thing: what will my clients think, what will my audience do, will I lose the next commission. One commercial photographer put it plainly. She said, "my clients are already nervous. I am not going to hand them a reason to cut my rate."
That is not a bad person. That is someone doing the math honestly.
The Framework Keeps Solving the Wrong Problem
Here is the thing about every AI transparency proposal I have read in the last eighteen months. They are all designed as if the problem is technical. As if creators are failing to disclose because they lack a standardized vocabulary, or because no one built the right metadata field, or because disclosure mechanisms are not embedded in the right platforms yet.
Adobe tried with Content Credentials. The C2PA coalition tried. The EU AI Act has carve-outs and requirements that took a small army of lawyers to write. And usage is... minimal. Voluntary adoption rates for Content Credentials in creative work are genuinely hard to find, which tells you something. When the industry people most invested in a system cannot produce clean numbers, the system is not working.
The problem is not technical. Creators know how to add metadata. They know how to write a caption. They know how to add a footnote. The information is not hard to surface. The issue is that surfacing it feels like attaching a discount sticker to your own work in a market that has not decided yet how to price AI involvement. Some audiences devalue it immediately. Some clients use it as a negotiating lever. Some platforms have started quietly deprioritizing content flagged as AI-generated without saying so explicitly.
So the rational move, for a solo creator with rent to pay and a client list to protect, is silence. Not lying, necessarily. Just... not volunteering information that the market has not learned to handle yet.
I find that genuinely depressing. Not because I think all AI use is equal or that disclosure is unimportant. But because the people pushing hardest for labeling frameworks are often not the ones absorbing the cost of wearing the label.
What Transparency Actually Requires
There are creators doing this honestly. A few illustrators I follow on Substack have been upfront about using AI for reference generation or color palette exploration while doing all the final rendering themselves. One writer I read regularly notes at the bottom of each post when she used a model for structural feedback or research synthesis. A sound designer I know from the modular community has started including a short process note on every commercial release.
What they have in common is not virtue, exactly. It is audience trust that was built before AI entered the conversation. They had enough relationship capital to survive the disclosure. And in most cases, their audiences responded with more respect, not less. The honesty created signal. It told their people: this person is not hiding things.
But here is what none of the framework proposals grapple with honestly: not every creator has that buffer. A new commercial photographer with eighteen months of client work does not have the relationship capital to absorb an AI disclosure conversation right now. A freelance writer who just started getting consistent assignments is not in a position to test whether their editor cares. The transparency frameworks are being designed for established creators and then pushed onto everyone.
The other thing the frameworks miss is that AI use is not binary. It is a spectrum with genuinely hard edges. Is a writer who used a model to check their argument structure disclosing AI use? What about a designer who ran three color options through an AI tool and then threw all of them out but kept thinking about the logic? What about music produced entirely by a human who spent six months training a custom model on their own archive? These are not gotcha questions. They are what creators are actually navigating, and no label resolves them cleanly.
I do not think the answer is no transparency. I think the answer is that transparency becomes real when audiences start rewarding it consistently, and right now they are not. Some do. Most have not formed a stable preference yet. The market is still reading the room.
Until that changes, the label frameworks are going to stay mostly theoretical. Creators will keep making the rational calculation. And the people most hurt by that silence are the ones trying to build genuine trust with their audience in a space where trust is the only asset that actually compounds.
Look. If you are a creator who is using AI and staying quiet about it, I am not calling you a coward. I am saying you are responding to the incentives in front of you, and those incentives are real. But I also think some of you are underestimating your audience. The ones who have been with you long enough to notice your voice... they might handle the honesty better than you think. And the ones who would punish you for it were probably not going to stay anyway.
The label problem is not going to be solved by a better label. It is going to be solved, if it gets solved at all, by enough creators betting on their audience and winning. One disclosure at a time. Not because a framework told them to. Because they decided the honesty was worth it.
That is a much slower fix than a metadata standard. But it is the one that actually works.