AI‑Generated Fake Paper Slip Past IEEE Access: A Wake‑Up Call for Scholarly Publishing

In a striking proof‑of‑concept, Benjamin Skuse managed to generate a completely fabricated research article on optical solitons using only free AI chatbots—ChatGPT’s Mini version and Gemini 3 Flash—and submit it to IEEE Access. The manuscript, assembled through a series of prompt‑driven steps, passed the journal’s editorial screening without triggering any AI‑detection alarms, exposing a glaring vulnerability in the scholarly publishing pipeline. Skuse’s experiment unfolded in two phases. First, he instructed the language models to draft hypotheses, methods, results and even figures for a study that had never been performed. By breaking the paper into its constituent sections—abstract, introduction, methodology, results, discussion—and switching between the two bots whenever one stalled, he produced a coherent manuscript that adhered to IEEE’s formatting guidelines. The second phase involved a “war‑game” with Jenny Mahoney, managing editor of IEEE Access, who agreed to evaluate the submission under normal review conditions. Neither the editor nor the automated checks detected any tell‑tale AI‑isms, such as forced paraphrases or overly formulaic transitions. The episode arrives against a backdrop of alarming statistics. By 2025, an estimated 57 % of all published academic articles show detectable traces of large language model (نماذج اللغة الكبيرة - LLM) influence, up from just 12 % in 2023, while more than 12 % of reviews in Nature Communications were found to be entirely AI‑generated. Moreover, for every 40 papers that exhibit statistical evidence of AI usage, only one openly discloses it, suggesting that the true scale of the problem is vastly underreported. “It's not just the scientific record under threat, it's science as a whole,” warns Amanda Sulicz, IEEE’s research‑integrity manager. The ease with which Skuse fabricated a plausible study underscores how AI‑generated “slop” can erode trust in the peer‑review process, burden already overstretched reviewers, and potentially flood the literature with undetectable falsifications. Industry experts now call for robust detection tools, mandatory AI‑usage disclosures, and a re‑examination of editorial workflows. As AI continues to automate not only writing but also data analysis, the academic community faces a pivotal crossroads: adapt its safeguards or risk a cascade of misinformation that could compromise the very foundations of scientific progress.