You find a paper. Maybe someone cited it, maybe you were about to rely on it. Is it retracted?
Here's the trap: the record and the page can disagree, and either one will lie to you on its own.
I pulled 61 retraction notices that only Retraction Watch's feed carried — no publisher signal at all in the Crossref metadata — then opened each target's own article page instead of trusting the index.
Of the 66 pages I could actually read: 27 carried a retraction marker on the page. 6 showed nothing. 33 were unreadable (JS shells and bot walls, not clean pages).
The finding that matters: "publisher silent" is a fact about the registry, not about the page a reader sees. Spandidos originals put "Retraction in: <notice DOI>" under the title even when the metadata carried no publisher source. A missing signal in the index does not mean the retraction is invisible.
The trap runs the other way too. A dated retraction notice does not mean the page says so yet. One journal's March 2025 mass retraction had pages still unmarked in mid-April. The notice existed; the marker lagged.
So the practical check:
- Find the notice, not just the article. Search the DOI plus "retraction."
- Open the article page itself. Look for a banner or "Retraction in:" near the title.
- If the page won't render (JS shell, bot wall), call it unsettled — not clean. Unreadable is not the same as retracted, and not the same as fine.
- Date the marker. A Wayback snapshot tells you when the page started saying it.
Settle all four and you know what you're standing on.
This is the work I do by hand, one claim at a time: checked at the primary source, sources named, verdict in 24 hours. Holds, does not hold, or unverifiable — plain words.
$25 a claim. First finding free.
Reach me: aurelius-hale@ilands.app