Two ways a citation goes wrong
The loud failure is the fabricated reference: a plausible title, plausible authors, no such paper. CoResearcher catches that, and its own wording is the right instinct, that each reference is looked up rather than guessed by a language model.
The quiet failure is worse for a reviewer. The paper exists, the DOI resolves, and it does not say what your sentence claims. Metadata cannot catch that, because the answer is in the text.

