Caught in the Act: Unethical AI Usage in Research

Academics have been caught copying and pasting text from ChatGPT straight into peer-reviewed papers—sometimes even leaving telltale “as an AI language model” caveats in the manuscript. The immediate outrage misses a larger point: the real risk isn’t that researchers will use AI to draft or assist writing, but that parts of the academic quality-control system—peer review, editorial screening, and publication checks—are failing to catch obvious problems.

Examples pulled from Google Scholar searches show multiple instances where AI-generated boilerplate appears verbatim in published work, including passages that explicitly include “as an AI language model” language in methods or results-adjacent text. In at least one case, the copied content includes statements about missing access to full text or inability to rewrite without original material—errors that should be obvious to anyone reading the manuscript. Similar issues are also reported in theses, where ChatGPT-like limitations and mismatched prompts appear in submitted academic writing.

The transcript argues that this kind of misconduct is likely to be relatively limited in the near term—partly because the volume of papers is large and partly because many of the flagged cases appear in lower-visibility outlets rather than top-tier journals. But attention shifts to a more consequential failure: a high-impact journal publication (Elsevier’s Resources Policy is cited, with an impact factor of 10.2) allegedly passed through peer review and editorial processes despite containing AI artifacts and text that suggests the authors did not properly integrate results. The concern is not that AI can generate prose; it’s that AI-generated drafts can be used to create narrative around data without correctly presenting tables, tests, or verifiable findings—and that such deficiencies should have been caught during review, revision, typesetting, and final editorial checks.

That leads to a broader suspicion of “publishing cabals,” where papers may be fast-tracked or accepted due to relationships rather than rigorous scrutiny. The transcript frames this as a structural problem: if journals can publish work with glaring AI remnants, then the integrity of peer review is under strain.

A parallel example comes from grant assessment. A Guardian report is cited describing Australian Research Council assessor reports allegedly containing phrases like “regenerate response,” a button-like artifact from ChatGPT that can appear when copying output. The transcript connects this to reviewer overload: researchers may process up to 20 proposals in a year, each typically 50–100 pages, on top of teaching, supervision, publishing, and other duties. Under that pressure, shortcuts become tempting—and AI artifacts become more likely.

The proposed remedy is practical rather than punitive: universities and academic services should develop clear, usable policies for AI tools, specify where AI is allowed, and invest in enough staffing and time so researchers and reviewers can do real work. The core message is that AI use in academia is likely inevitable, but the systems around it—review rigor, editorial accountability, and workload—must keep pace to prevent quality from eroding.