AI evidence extraction

When References Mislead: Verification, AI Attribution, and Academic Bullying in Scholarly Evaluation

Carlos Heredia Chimeno · AI & Antiquity · 2026

AI-generated evidence extraction, verified across multiple analytical personas. Not a substitute for the peer-reviewed original.

9/10

Relevance

2/4

Quality (LMQS)

Evidence

Citations

0.00

FWCI

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.64946/aiantiquity.v2i1.editorial

Methodology & findings

Study design

Narrative editorial analysis synthesizing theoretical argument, documented case examples from scholarly communication, and thematic review of pedagogical and methodological approaches to AI in humanities teaching and research.

Main result

The paper finds that "generative AI models are not databases. They do not 'know' whether a reference exists; they predict text" and that "A single fabricated reference, once cited, reformatted, exported to reference managers, or reused in teaching materials, can circulate widely before anyone notices the absence of an original source." The authors conclude that bibliographic fabrication by AI represents a systemic challenge requiring institutional response rather than individual sanction, and that "scholarship must now actively defend itself against the production of plausible but unverified knowledge."

Reports effect sizes.

Research paradigm

Critical humanities scholarship; reflexive epistemology emphasizing procedural ethics and verification in AI-mediated scholarly communication

Author conclusions

The authors conclude that "Scholarly rigour is best protected not through suspicion and punishment, but through education, awareness, and procedural adaptation" and that "the task is not to reject AI, but to embed verification as a non-negotiable scholarly standard, ensuring that the ease of generating academically styled text does not weaken the chain of traceability on which humanistic knowledge depends." They further conclude that addressing this requires "shared protocols that distinguish clearly between documented irregularity, correctable error, and demonstrable misconduct."

Risk of bias

Potential selection bias in case examples cited (cases discussed may not be representative); confirmation bias in framing AI bibliography errors as systemic rather than sporadic; editorial perspective may privilege certain disciplinary approaches; Confirmation bias in attributing AI authorship based on bibliographic anomalies; Selection bias in documented cases presented; Reputation bias in academic networks where allegations circulate informally; As a narrative editorial essay rather than empirical research, systematic bias assessment is not applicable. However, the paper's reliance on informal documentation of cases ("Documented episodes within recent scholarly communication") without systematic review represents a potential selection bias. The framing emphasizes procedural fairness and resists punitive approaches, which may reflect editorial positioning rather than neutral evidence synthesis.

Open questions raised

The paper identifies the need for: clearer institutional guidelines on acceptable AI use in scholarship; explicit verification protocols at editorial level; practical pedagogical exercises exposing how AI bibliography errors occur; refined editorial procedures incorporating systematic reference verification; shared protocols distinguishing between error, negligence, and fraud in scholarly assessment
Need for explicit verification protocols and clearer guidelines on acceptable AI use in academic publishing
Requirement for constructive communication procedures between editors, reviewers, and authors when bibliographic issues arise
Development of institutional norms and editorial procedures that systematically verify references
Integration of AI literacy and critical evaluation into teaching practices across disciplines
The paper identifies the need for: (1) explicit teaching of verification practices and digital literacy regarding AI-generated bibliographies; (2) adaptation of editorial workflows to incorporate verification checks; (3) development of shared protocols distinguishing between documented irregularity, correctable error, and misconduct; (4) refinement of institutional norms and everyday research habits to embed verification as non-negotiable; (5) methodological strategies for working with AI in humanities scholarship without relinquishing scholarly rigor.

Extracted from: pdfAgreement 71%

Explore related topics

Hallucination Control Research Integrity Position Paper