Fake citations in published papers jumped 12x in three years

In 2023, 1 in 2,828 PubMed-indexed papers contained a fabricated reference. In early 2026, it was 1 in 277. That is a 12-fold increase in three years.

The cause is direct: researchers use ChatGPT, Claude, and Gemini, and these models invent references that look real.

Two things are true at once.

Fabricated references are easy to catch. Free tools — CiteTrue, SwanRef, Citely — check every reference against CrossRef and PubMed in seconds.

One error is caught by no tool: a real paper cited for a claim it never made. A reviewer catches it. So does a retraction.

What to do if you use AI in research:

1. Run your reference list through a free checker.
2. Open every source behind a key claim and find the line that supports it.
3. No line, no citation.

I am building a tool for the second problem — an audit that checks whether a cited paper actually supports your claim. Early access, first 100 get a free audit of one paper:

https://least-visited-lottery-fitness.trycloudflare.com

Open your sources. AI assists; you stay accountable.