CoChat
Sign Up
PreprintCited by 8Topic ModelingEthics and Social Impacts of AIArtificial Intelligence Applications

ChatGPT Hallucinates Nonexistent Citations: Evidence from Economics

Open access
Buchanan, Joy, Hill, Stephen, Shapoval, Olga·2023·SSRN Electronic Journal
DOI 10.2139/ssrn.4632842
Open this paper in CoChat — free account

Abstract

In this study, we generate prompts derived from every topic within the Journal of Economic Literature to assess the abilities of both GPT-3.5 and GPT-4 versions of the ChatGPT large language model (LLM) to write about economic concepts. ChatGPT demonstrates considerable competency in offering general summaries but also cites non-existent references. More than 30% of the citations provided by the GPT-3.5 version do not exist and this rate is only slightly reduced for the GPT-4 version. Additionally, our findings suggest that the reliability of the model decreases as the prompts become more specific. We provide quantitative evidence for errors in ChatGPT output to demonstrate the importance of LLM verification.

Integrity signals

Registry

Crossref · 10.2139/ssrn.4632842

Cites

6 of 6 references verified0 citation errors · 0 not checkable

Statistics

Recomputing p-values needs the paper’s full text.

Claims the paper makes

We couldn’t retrieve this paper’s full text, so its claims and self-check weren’t run. The registry and citation checks still apply.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • ChatGPT and similar large language models tend to fabricate plausible-sounding but non-existent references (hallucinated citations).

    “One of the most pressing concerns is the phenomenon of AI hallucinations, i.e.ChatGPT's tendency to fabricate plausible-sounding but non-existent references.”

    From Training Students to Identify and Correct Fabricated References in ChatGPT-Generated Literature Reviews (2025) · cites this paper as Buchanan et al. (2023)

  • Buchanan et al. (2023) found that over 30% of citations produced by GPT-3.5 in economics contexts did not exist, and even GPT-4 failed to eliminate this problem entirely.

    “Similarly, Buchanan et al. (2023) found that over 30% of citations produced by GPT-3.5 in economics contexts did not exist, and even GPT-4 failed to eliminate this problem entirely.”

    From Training Students to Identify and Correct Fabricated References in ChatGPT-Generated Literature Reviews (2025) · cites this paper as Buchanan et al. (2023)

Cites

Flagged first, then by citations

Cited by

Sorted by citations

Authors
Buchanan, J., Hill, S., & Shapoval, O.
Affiliations
Samford University (US)
Corresponding author
Buchanan, Joy
ORCID
Venue
SSRN Electronic Journal
Publisher
Elsevier BV
Published
2023
ISSN
1556-5068 (ISSN-L)
DOI
10.2139/ssrn.4632842
OpenAlex
W4389161477
Cited by
8
References
6
Citations by year
2025: 5 · 2024: 3
Type
posted content
Language
en
Open access
Locations
SSRN Electronic Journal (acceptedVersion)
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
  • Social Sciences › Safety Research › Ethics and Social Impacts of AI
  • Physical Sciences › Computer Science › Artificial Intelligence › Artificial Intelligence Applications
Keywords
ChatGPT, hallucinations, citation errors, large language models, GPT-3.5, GPT-4, prompt specificity, model reliability, economics writing, Journal of Economic Literature

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review