Preprints, versions, and what you are actually citing
Lesson 5 of 6 · Citations that survive
By the end you can tell a preprint from its published version and cite the one that still says what you quote.
1 · Learn the move · Comparison matrix
A preprint and its published version are different documents with different DOIs, and the differences are not cosmetic: numbers move between versions, claims soften under review, figures get replaced, and occasionally the published paper no longer contains the sentence you are citing. Version is therefore part of the citation, and part of its verification state - VERIFIED means verified against the version cited. The comparison matrix makes the drift visible: one row per claim you rely on, one column per version, a changed-or-unchanged call per cell. You cite the version you actually read, you say which one it was, and if a published version exists, you check whether your claim survived the journey before pointing at either.
I cite [preprint citation + version] for these claims: [list]. The published version is [citation]. Build a comparison matrix: rows = my claims, columns = preprint version / published version, cells = what each version says, from the passages I paste. Flag every cell where the claim changed, and tell me which version my citation must point at for each claim. Passages: [paste both versions' relevant sections].
2 · Your turn. You write the prompt
You cited a preprint eight months ago for a fold-change and a mechanism claim. The paper has since appeared in a journal. Your methods cite the preprint DOI, your discussion quotes the fold-change, and you have not opened the published version. The manuscript goes to the senior author on Friday.
Write the prompt for this scenario as given, or swap in a real one from your own work. The score works the same either way.
Remember: the AI sees only your prompt, not this page. If the situation isn't in your prompt, it doesn't exist.
Optional. These shape the output when you run your prompt below, not your score.