CV Google Scholar MSU
Demos · What this means for research

Thirty citations

We asked for journal articles, with DOIs, then checked every one. Guide: Verification →

What we asked

Give me 30 peer-reviewed management journal articles on [topic].

Two topics. A thin literature, “algorithmic legitimacy in organizations,” and an established one, “prospect theory and decision under risk.” Thirty citations each, every DOI checked against Crossref.

What came back

Thin literature

“algorithmic legitimacy in organizations”

DOI exists
17 of 30
It is the paper named
8 of 30

Established literature

“prospect theory and decision under risk”

DOI exists
20 of 30
It is the paper named
18 of 30
gpt-4o via the API · temperature 1.0 · September 2026 · 30 citations per topic · checked against Crossref

What to notice

  1. The DOI existing is not the check. Most of us would stop there. In the thin literature 17 DOIs resolved, but only 8 went to the paper the model named. The other 9 went to a different paper.
  2. A real DOI on the wrong paper is the hardest error to catch. It passes the quick check, so it looks verified when it isn’t.
  3. Check hardest where the literature is thin. That’s where the model has the least to predict from, and it’s where you are working when the question is worth asking.

Limits

Two topics, one model. Checked against Crossref by DOI and title.