Influence of Topic Familiarity and Prompt Specificity on Citation Fabrication in Mental Health Research Using Large Language Models: Experimental Study
Open accessPublished version of a preprint
The DOI matches its Crossref record; 19 of the 19 references it cites were found in a registry or an index; 21 of the 23 sources its claims cite were checked for support.
Abstract
Background: Mental health researchers are increasingly using large language models (LLMs) to improve efficiency, yet these tools can generate fabricated but plausible-sounding content (hallucinations). A notable form of hallucination involves fabricated bibliographic citations that cannot be traced to real publications. Although previous studies have explored citation fabrication across disciplines, it remains unclear whether citation accuracy in LLM output systematically varies across topics within the same field that differ in public visibility, scientific maturity, and specialization. Objective: This study aims to examine the frequency and nature of citation fabrication and bibliographic errors in GPT-4o (Omni) outputs when generating literature reviews on mental health topics that varied in public familiarity and scientific maturity. We also tested whether prompt specificity (general vs specialized) influenced fabrication or accuracy rates. Methods: In June 2025, GPT-4o was prompted to generate 6 literature reviews (~2000 words; ≥20 citations) on 3 disorders representing different levels of public awareness and research coverage: major depressive disorder (high), binge eating disorder (moderate), and body dysmorphic disorder (low). Each disorder was reviewed at 2 levels of specificity: a general overview (symptoms, impacts, and treatments) and a specialized review (evidence for digital interventions). All citations were extracted (N=176) and systematically verified using Google Scholar, Scopus, PubMed, WorldCat, and publisher databases. Citations were classified as fabricated (no identifiable source), real with errors, or fully accurate. Fabrication and accuracy rates were compared by disorder and review type by using chi-square tests. Results: Across the 6 reviews, GPT-4o generated 176 citations; 35 (19.9%) were fabricated. Among the 141 real citations, 64 (45.4%) contained errors, most frequently incorrect or invalid digital object identifiers. Fabrication rates differed significantly by disorder (χ22=13.7; P=.001), with higher rates for binge eating disorder (17/60, 28%) and body dysmorphic disorder (14/48, 29%) than for major depressive disorder (4/68, 6%). While fabrication did not differ overall by review type, stratified analyses showed higher fabrication for specialized versus general reviews of binge eating disorder (11/24, 46% vs 6/36, 17%; P=.01). Accuracy rates also varied by disorder (χ22=11.6; P=.003), being lowest for body dysmorphic disorder (20/34, 59%) and highest for major depressive disorder (41/64, 64%). Accuracy rates differed by review type within some disorders, including higher accuracy for general reviews of major depressive disorder (26/34, 77% vs 15/30, 50%; P=.03). Conclusions: Citation fabrication and bibliographic errors remain common in GPT-4o outputs, with nearly two-thirds of citations being fabricated or inaccurate. Reliability systematically varied by disorder familiarity and prompt specificity, with greater risks in less visible or specialized mental health topics. These findings highlight the need for careful prompt design, rigorous human verification of all model-generated references, and stronger journal and institutional safeguards to protect research integrity as LLMs are integrated into academic practice.
Overview
CoChat Summary
TL;DR
In six GPT-4o–generated mental health literature reviews, 35 of 176 citations (19.9%) were fabricated, and among the 141 real citations, 64 (45.4%) contained bibliographic errors. Fabrication and accuracy varied by disorder familiarity and, within some disorders, by prompt specificity.
Key findings
- Across 6 reviews, GPT-4o produced 176 citations; 35 (19.9%) were fabricated. Of the 141 nonfabricated citations, 77 (54.6%) were fully accurate and 64 (45.4%) contained errors, most commonly incorrect or invalid DOIs (51/141, 36.2%).
- Fabrication rates differed significantly by disorder (χ²₂=13.7; P=.001): major depressive disorder 4/68 (6%), binge eating disorder 17/60 (28%), and body dysmorphic disorder 14/48 (29%). Pairwise comparisons showed higher fabrication for binge eating disorder and body dysmorphic disorder than for major depressive disorder.
- Overall fabrication did not differ by review type (general 17/102, 16.7% vs specialized 18/74, 24%; P=.21), but stratified analyses found higher fabrication for specialized versus general binge eating disorder reviews (11/24, 46% vs 6/36, 17%; P=.01).
- Accuracy differed by disorder (χ²₂=11.6; P=.003), with results reporting major depressive disorder 41/64 (64%), binge eating disorder 26/43 (60%), and body dysmorphic disorder 10/34 (29%). Within-disorder review-type differences included higher accuracy for general versus specialized major depressive disorder reviews (26/34, 77% vs 15/30, 50%; P=.03) and higher accuracy for specialized versus general body dysmorphic disorder reviews (7/13, 54% vs 3/21, 14%; P=.01).
Methods
- In June 2025, GPT-4o was prompted in new sessions to generate 6 literature reviews (~2000 words; ≥20 citations) on 3 disorders varying in public awareness and research coverage: major depressive disorder (high), binge eating disorder (moderate), and body dysmorphic disorder (low).
- Each disorder was reviewed at 2 specificity levels: a general overview (symptoms, impacts, treatments) and a specialized review (evidence for digital interventions). All 176 citations were extracted and verified using Google Scholar, Scopus, PubMed, WorldCat, and publisher databases.
- Citations were classified as fabricated, real with errors, or fully accurate. Fabrication and accuracy rates were compared by disorder and review type using chi-square tests, with stratified chi-square analyses for within-group differences.
Limitations
- The paper states that findings may not generalize to other psychiatric disorders, specialized subfields, or academic disciplines; results are specific to GPT-4o; no prompt-engineering strategies were tested; and only one output per prompt was analyzed, so intraprompt variability was not assessed.
- The paper also notes that low expected cell counts prevented formal modeling of interaction effects, relying instead on stratified chi-square tests.
- Evident from the text: there is an internal inconsistency in the reported body dysmorphic disorder accuracy rate—the abstract states 20/34 (59%), while the Results section and Table 1 report 10/34 (29%).
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
15 Valid · 1 Not valid · 2 Not checked · 63 Own results and interpretations
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Binge eating disorder occupies an intermediate position in terms of research output and public awareness, whereas body dysmorphic disorder remains comparatively understudied and less widely known [10].”
§ Influence of Topic Familiarity and Prompt Specificity on Citation Fabrication in Mental Health Research Using Large Language Models: Experimental Study > Introduction
What the sources say
- Partly supportedcites [10] sourceExperimental — accuracy not yet measured
The Introduction says BDD has received little empirical attention compared with related conditions, and the Conclusions call for greater awareness of BDD. The paper does not establish where binge eating disorder falls in research output or public awareness, so the claimed intermediate position is unsupported. (LLM, conf=0.86)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“In contrast, when prompted on the specific topic, all 5 citations contained misattributed authorship but linked to real papers with matching titles [9].”
§ Influence of Topic Familiarity and Prompt Specificity on Citation Fabrication in Mental Health Research Using Large Language Models: Experimental Study > Introduction
What the sources say
- Supportedcites [9] source
Section 3.4.3 (p. 6) reports that Bard returned five citations for the more specific topic, all with misattributed authorship. It also states that all five hyperlinks led to real articles with the same title. (LLM, conf=0.95)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“The findings of this study align with previous research across various disciplines showing that citation fabrication and bibliographic inaccuracies are common in LLM-generated outputs [7-9,14].”
§ Influence of Topic Familiarity and Prompt Specificity on Citation Fabrication in Mental Health Research Using Large Language Models: Experimental Study > Discussion > Comparisons With Prior Work
What the sources say
- Supportedcites [7] source
The Methods (page not provided) describe testing ChatGPT citations across 42 multidisciplinary topics. The Results and Discussion (pages not provided) report substantial rates of fabricated citations and substantive bibliographic errors in both GPT-3.5 and GPT-4 outputs. (LLM, conf=0.93)
- Supportedcites [8] source
The Results (‘Citation Existence and Accuracy’) report that only 76 of 102 generated citations were real. The Discussion (‘Principal Findings’) reports inaccurate citations and frequent DOI hallucination in both the natural sciences and humanities, supporting the claim across disciplines. (LLM, conf=0.91)
- Supportedcites [9] source
The Discussion (p. 7) reports that only 6% of ChatGPT-generated citations matched actual manuscripts and that none of Bard’s citations were accurate. The Discussion (p. 8) also reports similar fabrication and bibliographic errors in physical geography education and internal medicine, supporting the claim that the problem has been observed across disciplines. (LLM, conf=0.91)
- Partly supportedcites [14] sourceExperimental — accuracy not yet measured
Study 1’s Results (Overall Hallucinations and Errors) report fabricated references and smaller bibliographic errors, including incorrect years or journals, in GPT-3.5 and GPT-4 outputs. But the study’s Design section says its reference-generation tasks covered topics in psychology; this source alone does not establish that the pattern occurs across various disciplines. (LLM, conf=0.84)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Specifically, accuracy rates within generic reviews were significantly lower for body dysmorphic disorder compared to both major depressive disorder (P<.001) and binge eating disorder (P=.001).”
§ Influence of Topic Familiarity and Prompt Specificity on Citation Fabrication in Mental Health Research Using Large Language Models: Experimental Study > Results > Accuracy Rates
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Within generic reviews, fabrication rates were significantly lower for major depressive disorder than for body dysmorphic disorder (P=.001).”
§ Influence of Topic Familiarity and Prompt Specificity on Citation Fabrication in Mental Health Research Using Large Language Models: Experimental Study > Results > Fabrication Rates
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
47 more claims
Create a free account to see all 81 claims and every source verdict.
- It may state two separate claims as one. 18
- Probably not a substantial claim. 18
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?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“First, prior studies have examined at most 4 models simultaneously, and none have compared current-generation systems from all major commercial providers within a single experimental framework [8 -10].”
From How LLMs Cite and Why It Matters: A Cross-Model Audit of Reference Fabrication in AI-Assisted Academic Writing and Methods to Detect Phantom Citations (2026) · cites this paper as [10]
Experimental — accuracy not yet measuredThe cited source says what the claim attributes to it.
In the citing paper
“This interpretation is consistent with Linardon et al. [10] finding that topic familiarity modulates hallucination, but our result extends the mechanism from domain familiarity to temporal familiarity within the same domain.”
From How LLMs Cite and Why It Matters: A Cross-Model Audit of Reference Fabrication in AI-Assisted Academic Writing and Methods to Detect Phantom Citations (2026) · cites this paper as Linardon et al. [10]
The cited source says what the claim attributes to it.
In the citing paper
“suggested that models are more likely to fabricate when their training data is sparse for a given subject.”
From How LLMs Cite and Why It Matters: A Cross-Model Audit of Reference Fabrication in AI-Assisted Academic Writing and Methods to Detect Phantom Citations (2026) · cites this paper as [10]
The cited source says what the claim attributes to it.
In the citing paper
“fabrication rates reached 28 to 29% for less visible disorders (e.g., body dysmorphic disorder), compared to 6% for major depressive disorder.”
From How LLMs Cite and Why It Matters: A Cross-Model Audit of Reference Fabrication in AI-Assisted Academic Writing and Methods to Detect Phantom Citations (2026) · cites this paper as [10]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated major: it would touch a result — a value a result rests on, reported and recomputed p-values on opposite sides of the significance level, two passages that conflict with no sign which is right, or a cited work retracted before this paper was published.
In the paper
§ Influence of Topic Familiarity and Prompt Specificity on Citation Fabrication in Mental Health Research Using Large Language Models: Experimental Study > Discussion > Principal Findings
“among the 141 nonfabricated citations, 77 (54.6%) contained bibliographic errors”
§ Influence of Topic Familiarity and Prompt Specificity on Citation Fabrication in Mental Health Research Using Large Language Models: Experimental Study > Abstract > Results
“Among the 141 real citations, 64 (45.4%) contained errors”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
“| | DOIc | 6/33 (18.2) | 11/30 (37.7) |”
§ Influence of Topic Familiarity and Prompt Specificity on Citation Fabrication in Mental Health Research Using Large Language Models: Experimental Study > Results > Overview
Reported 37.7 · recomputed 36.666667
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
“| 16/52 (30.2) | 39/131 (29.8) |”
§ Influence of Topic Familiarity and Prompt Specificity on Citation Fabrication in Mental Health Research Using Large Language Models: Experimental Study > Results > Overview
Reported 30.2 · recomputed 30.769231
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Influence of Topic Familiarity and Prompt Specificity on Citation Fabrication in Mental Health Research Using Large Language Models: Experimental Study > Abstract > Results
“being lowest for body dysmorphic disorder (20/34, 59%)”
§ Influence of Topic Familiarity and Prompt Specificity on Citation Fabrication in Mental Health Research Using Large Language Models: Experimental Study > Results > Accuracy Rates
“body dysmorphic disorder: 10/34, 29%”
Citations
Cites
Huang, L., Yu, W., Ma, W., Zhong, W., Feng, Z., Wang, H., Chen, Q., Peng, W., Feng, X., Qin, B., et al. (2025) · ACM Transactions on Information Systems · doi:10.1145/3703155 · Cited by 4,688 · cited 1×
“These concerns are not unfounded, as one of the most well-documented limitations of LLMs is their tendency to “hallucinate,” which occurs when the model generates false, fabricated, or unverifiable information that appears coherent and credible [6].”
A Comprehensive Overview of Large Language Models
Naveed, H., Khan, A. U., Qiu, S., Saqib, M., Anwar, S., Usman, M., Akhtar, N., Barnes, N., & Mian, A. (2023) · arXiv · doi:10.48550/arxiv.2307.06435 · Cited by 1,259 · cited 3×
“LLMs are advanced artificial intelligence systems trained on vast amounts of textual data to generate coherent and contextually relevant natural language responses [2].”
InfluentialA Comprehensive Overview of Large Language Models
Naveed, H., Khan, A. U., Qiu, S., Saqib, M., Anwar, S., Usman, M., Akhtar, N., Barnes, N., & Mian, A. (2025) · ACM Transactions on Intelligent Systems and Technology · doi:10.1145/3744746 · Cited by 1,257 · cited 3×
“LLMs are advanced artificial intelligence systems trained on vast amounts of textual data to generate coherent and contextually relevant natural language responses [2].”
InfluentialFabrication and errors in the bibliographic citations generated by ChatGPT
Walters, W. H., & Wilder, E. I. (2023) · Scientific Reports · doi:10.1038/s41598-023-41032-5 · Cited by 499 · cited 3×
“First, we closely mirrored the style of prompts used in previous studies of LLM-generated literature reviews in other disciplines [7].”
AWARE: Mobile Context Instrumentation Framework
Ferreira, D., Kostakos, V., & Dey, A. K. (2015) · Frontiers in ICT · doi:10.3389/fict.2015.00006 · Cited by 409
AI and science: what 1,600 researchers think
Van Noorden, R., & Perkel, J. M. (2023) · Nature · doi:10.1038/d41586-023-02980-0 · Cited by 383 · cited 2×
“As these tools become more deeply integrated into research workflows [5], it is essential to equip users with the knowledge and skills to critically assess model-generated outputs.”
Publish or perish: academic life as management faculty live it
Miller, A. N., Taylor, S. G., & Bedeian, A. G. (2011) · Career Development International · doi:10.1108/13620431111167751 · Cited by 326 · cited 1×
“Researchers today face growing pressure to maintain high levels of productivity to remain competitive for funding, tenure-track positions, academic promotions, and international recognition in their fields [1].”
Rosenbloom, L. (2019) · The Charleston Advisor · doi:10.5260/chara.21.2.8 · Cited by 284
Gravel, J., D’Amours-Gravel, M., & Osmanlliu, E. (2023) · Mayo Clinic Proceedings: Digital Health · doi:10.1016/j.mcpdig.2023.05.004 · Cited by 222
Gravel, J., D’Amours-Gravel, M., & Osmanlliu, E. (2023) · doi:10.1101/2023.03.16.23286914 · Cited by 220
Cited by
Hallucinated citations are polluting the scientific literature. What can be done?
Naddaf, M., & Quill, E. (2026) · Nature · doi:10.1038/d41586-026-00969-z · Cited by 36
Resnik, D. B., & Hosseini, M. (2026) · Accountability in Research · doi:10.1080/08989621.2026.2645390 · Cited by 26
“For instance, in a paper published in the journal Academic Ethics, 19 out of 29 citations were hallucinated (Retraction Watch 2025).”
Detecting and Correcting Reference Hallucinations in Commercial LLMs and Deep Research Agents
Rao, D., Wong, E., & Callison-Burch, C. (2026) · arXiv · doi:10.48550/arxiv.2604.03173 · Cited by 16
Wang, X., Yin, C., He, H., Guo, J., Fu, X., & 白飞虎 (2026) · Frontiers in Public Health · doi:10.3389/fpubh.2026.1810358 · Cited by 10
Ülkir, M., & Paslı, B. (2026) · Clinical Anatomy · doi:10.1002/ca.70187 · Cited by 7
Metadata
- Authors
- Linardon, J., Jarman, H. K., McClure, Z., Anderson, C., Liu, C., & Messer, M.
- Affiliations
- Deakin University (AU)
- Corresponding author
- Linardon, Jake
- Venue
- JMIR Mental Health
- Publisher
- JMIR Publications Inc.
- Published
- 2025-11-12
- Volume
- 12
- Pages
- e80371
- ISSN
- 2368-7959 (ISSN-L)
- PMC
- PMC12658395
- OpenAlex
- W4416154989
- Cited by
- 40
- References
- 15
- Citations by year
- 2026: 39 · 2025: 1
- FWCI
- 41.17
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- cc-by
- Locations
- JMIR Mental Health (publishedVersion, cc-by)
- PubMed (publishedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Figshare (submittedVersion)
- Funders
- Medical Research Council
- National Health and Medical Research Council (1196948)
- Topics
- Social Sciences › Education › Academic integrity and plagiarism
- Social Sciences › Library and Information Sciences › Academic Writing and Publishing
- Social Sciences › Library and Information Sciences › scientometrics and bibliometrics research
- Keywords
- hallucinations, large language models, GPT-4o, bibliographic errors, prompt specificity, topic familiarity, mental health research, binge eating disorder, body dysmorphic disorder, major depressive disorder, digital interventions
- MeSH terms
- Large Language Models, Humans, Language, Mental Health, Biomedical Research
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.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
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
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.