AI Disclosure is Not Enough

Even if AI disclosure were complete and reliable, it would answer the wrong question. What matters is whether the authors remained intellectually engaged, verified the output, and accepted responsibility for the work, not whether a model touched the prose. A disclosure can imply that using the tool is sufficient once declared, or that a paper without such a disclosure deserves greater confidence. But, that does not necessarily follow. Authors must think through and stand behind every claim, and readers should approach every paper critically, regardless of whether AI assistance is disclosed. - Harlan M. Krumholz, writing in The Journal of the American College of Cardiology