After ChatGPT launched, an embarrassing pattern began showing up in the formal scientific literature: published, peer-reviewed papers that contained the conversational boilerplate a chatbot emits when it cannot or will not answer. Phrases such as “as an AI language model,” “Certainly, here is a possible introduction for your topic,” and “I don’t have access to real-time information” turned up verbatim in journal articles - clear evidence that an author had pasted LLM output straight into a manuscript without disclosure, and that peer reviewers and editors had not caught it.
A 2024 study by Alex Glynn, “Suspected Undeclared Use of Artificial Intelligence in the Academic Literature,” collected and analyzed the first 768 such examples. Its findings were unflattering to publishers. The problem was “widespread, penetrating the journals, conference proceedings, and textbooks of highly respected publishers,” and the undeclared AI text appeared disproportionately “in journals with higher citation metrics and higher article processing charges.” Worse, “an extremely small minority of cases are corrected post publication, and the corrections are often insufficient.” The author stressed that the detectable cases “likely represent a small fraction” of the real total, since most machine text leaves no obvious tell.
These cases are notable precisely because they are the detectable ones - the failures so blatant they survived editing only because nobody read closely. They became the public face of a deeper anxiety about whether peer review can verify what it is reviewing in the age of generative text, and they fueled the venue-level policies (such as those at ICML and NeurIPS) demanding disclosure and author accountability.
Why business readers should care: the leftover-phrase cases are a vivid reminder that a quality-control process is only as good as the human attention behind it. When a workflow assumes someone is reading the output and no one is, generative AI does not just introduce errors - it exposes that the checks were already hollow.