Transparency is not the same as disclosure.
Disclosure is showing everything. Transparency is showing what helps people understand.
The right idea, applied too broadly
Content Credentials — the C2PA standard backed by Adobe and others — is built on the right idea. Provenance matters. For news photography, legal evidence, electoral content, knowing where an image came from and how it was processed is genuinely important.
But the same badge that makes sense on a Reuters photograph looks different on a motivational quote card. "App or device used: ChatGPT. Issued by: OpenAI. Issued on April 15, 2025." That's not transparency. That's a receipt.
The tool name is not the trust signal
Knowing which tool made an image is not the same as knowing whether to trust it. A Photoshop badge doesn't make an image more credible. A ChatGPT badge doesn't make it less. Context does.
The label is also an interface. Where it appears, how prominent it is, what words it uses — these shape interpretation. A visible badge on a creative visual doesn't say "here is useful information." It says "something here needs your attention." That's a different message.
Transparency works when information matches the decision
Transparency works when the information matches the decision the viewer needs to make. For provenance-sensitive content, show provenance. For a robot illustration or an abstract visual, naming the tool is like putting the harvesting equipment on a coffee label. It may be true, but it is rarely the information the customer needs.
The question is not whether AI use should ever be disclosed. It should, in the right context.
The question is: disclosed to whom, at what granularity, and where in the interface.
Transparency is not a setting you turn on everywhere. It's a design decision.