What Disclosure Should Look Like When AI Touches Published Work
The question is not whether a tool was used but whether a reader is misled. That reframing settles most cases.
Publishers, agencies and businesses producing content have all had to arrive at a position on disclosing the use of generative tools. Debates about it stall because they start from the wrong question — whether a tool was involved — rather than from the one that matters, which is whether a reader would feel deceived on learning how something was made.
A workable test
Ask what the reader is being asked to believe. A reader of a bylined report believes a person investigated and stands behind the claims. A reader of a product description believes it accurately describes a product.
Disclosure is required where the tool's involvement would change what a reasonable reader thinks they are getting. That is why spell-checking has never required disclosure and a synthetic quotation attributed to a real person would be indefensible regardless of any disclosure.
The bright lines
Some things do not become acceptable through disclosure. Presenting generated text as the words of a real person, generating quotations that were never said, or producing statistics without a source are failures of accuracy, not of labeling.
Images are the sharpest case. A synthetic image presented as a photograph of a real event is a fabrication whatever the caption says.
Where a note belongs
Sensible practice is a short, specific note where a tool did substantive work: assisted with a first draft, used to summarize a document set, used to generate an illustration. Specific notes inform. A blanket sitewide statement that AI may have been used anywhere informs nobody.
Verification does not transfer
Whatever the workflow, someone remains responsible for every factual claim published. A tool cannot hold that responsibility, and no disclosure moves it.
The publications handling this well have written the policy down, applied it consistently, and made it public. The ones handling it badly are still deciding case by case, which is how inconsistency becomes a story about the publisher.