Walters & Wilder 2023 show that fabricated references are a systematic failure mode of ChatGPT-assisted writing: across studies the share of fabricated citations typically falls in the 47–69% range, and one radiology evaluation found 64% of 343 citations could not be found in PubMed or on the open web.
E11 states ChatGPT tends to cite non-existent works, that 6 studies systematically investigated it, that fabricated citations are 'typically in the 47–69% range', and that 64% of 343 radiology citations 'could not be found in PubMed or on the open web'.
Written by GLM-5.3 Flash via Ollama Cloud · 30 Sept, 22:56
Source chain
AEvery quote below was checked, without a model, to appear verbatim in its source.
- 01
“the proportion of fabricated citations is typically in the 47–69% range, with a higher rate in geography than in medicine”
Full-text passage · no page number · evidence E11
Passage read from www.ebi.ac.uk, which may be a preprint rather than the published version.
- 02
“Of the 343 citations, 64% were fabricated (i.e., could not be found in PubMed or on the open web)”
Full-text passage · no page number · evidence E11
Passage read from www.ebi.ac.uk, which may be a preprint rather than the published version.