Hadn't come across this one — just looked into it. The auto-PDF finder and retraction alerts seem genuinely useful for day-to-day research. Appreciate the rec.

Day 218.

Everyone calls the same journal a different name. Abbreviations, typos, old titles, merged publishers.

Building a system to prove a reference is NOT fake is convoluted.

But if something falls outside this framework, it's almost certainly fake beyond anything I could imagine.

Here's the weird part - ask GPT to verify a GPT-fabricated title and it says "correct." But Gemini catches it. Flip it around, same thing. Each LLM can't detect its own fabrications.

Which honestly means you could probably narrow down which model did it. Am I reading too much into this?

Something I keep finding in student references - titles with 2-3 words swapped to synonyms. The DOI is real, journal is real, but the title has subtle substitutions.

If you cited through Zotero or "Cite This," the title comes in verbatim. So why would only specific words change?

A librarian found a "preposterous number" of fake references in a Springer Nature journal paper.

Librarians are already chasing down papers that don't exist because students used AI. Now the journals are doing it too?

Who's actually checking?

retractionwatch.com

the editor who warned everyone about AI hallucinations got suspended for publishing AI-hallucinated quotes in his own newsletter. you can't make this up 🫠

$7,000 to publish a 4,000-word open access paper.

Meanwhile, the citations in those papers? Nobody's checking if they're even real.

Interesting priorities.