Simulation-Based Assessment of Citation Inflation in AI Literature Reviews

Abstract

An extensive bibliography can appear persuasive even when many references do not directly support the statements beside them. An evaluation that overlooks these constraints may reach conclusions that cannot be reproduced. The planned investigation seeks to evaluate whether structured source-claim auditing improves recognition of inflated or weakly related citations. Relevant methodological and data-quality literature informs the design of the sampling and validation strategy [3–7]. The protocol specifies an examination simulation with short AI-literature-review passages and preclassified citation-to-claim relationships. Comparators will include unguided citation review, a year-check checklist and evidence-matching assessment rubrics. Two storage-systems references are included only for the technology-governance background [1] and resilience context [2], not as evidence of human responses. The design prioritises precise operational definitions and a straightforward baseline before introducing more complex alternatives. Statistical and computational sources help define the reproducibility checks and comparison procedures [8–12]. The primary outcomes are citation relevance accuracy, verification time, false acceptance of sources and retention of review skills. Potential biases include chronological consistency, simulated counts, assessor agreement and the distinction between genuine publication and evidential fit. Primary analyses would be reported by the main deployment or participant subgroup so that averaged results do not conceal systematic failures. Additional research informs the prespecified stress tests and limits on interpretation [13–17]. The intended contribution is an auditable research protocol rather than an unsupported claim of effectiveness; related work frames its transferability limits [18–23].