Academic Publishers Build Screening Tools and Guidelines as AI-Generated Submissions Flood Journals

Academic publishers are building screening tools and drafting author guidelines to confront a surge of AI-generated paper submissions, according to a report published on 21 August 2026 by Publishers Marketplace.

Medical journal editor Itamar Ashkenazi, a member of the Committee on Publication Ethics (COPE) Council — the body that sets ethical standards for academic publishers — told a COPE presentation that submissions to academic journals have ballooned. Many papers now show telltale signs of AI generation, he said: factual mistakes, oddly fluent prose, and authors with improbably long publication lists compiled in just a few years.

The volume is the problem. When a small journal that once handled a few hundred submissions a year suddenly receives thousands, the peer-review system — where other scholars volunteer to read and vet manuscripts before publication — strains or breaks. Editors cannot easily tell which papers are genuine research and which are text produced by a chatbot and lightly edited.

The Committee on Publication Ethics maintains that artificial intelligence cannot be listed as an author of an academic paper. That position dates to a February 2023 statement on "Authorship and AI tools," which frames AI as a tool to speed up editorial decisions and lighten the workload for editors — not as a creator of scholarship.

But the line between assistance and authorship has only blurred. COPE has since published a topic discussion titled "When editors suspect AI," covering undisclosed AI use in submissions, warning signs for editors, and cases where AI may have been used to write peer reviews themselves — the very quality-control step meant to catch problems.

Individual publishers have moved at different speeds. Wiley, a major academic publisher, has published AI guidelines for researchers covering what authors, editors, and peer reviewers need to consider before using AI in research and writing. The American Psychological Association requires authors who use generative AI while drafting a manuscript to disclose it in the methods section and cite the tool. The Association for Computing Machinery permits authors to use generative AI tools to prepare manuscripts, subject to stated requirements.

A 2026 article by J. Chigwada in Frontiers in Research Metrics and Analytics notes that guidelines from STM — the international trade body for scientific, technical, and medical publishers — emphasise that AI can help refine, correct, and edit manuscripts but should not extend to certain other functions, a line the article does not fully specify. A 2025 article by M. Hosseini in a SAGE journal argues that publishers using AI detection tools should screen all relevant submissions, not just randomly or selectively chosen ones.

The Authors Guild, which represents writers across trade and academic markets, published "AI Best Practices for Authors" in May 2026 — guidelines rather than rules, designed to help members make their own informed decisions.

What makes this stand out is the mismatch of speed and scale. Generative AI tools became widely available in late 2022; within four years, they had reshaped the submission pipeline for an industry built on slow, careful human review. The tools now being developed to detect AI-generated text are racing to keep up with the tools that produce it — and the publishers setting the rules are the same ones running the review process. For readers who rely on academic journals for medical findings, social research, or technology reporting, the integrity of that pipeline matters beyond the trade.

COPE's own policies frame the issue around four questions: how the AI tool was used, how the output was verified, how transparent the authors were, and how editors assessed the result. The guidelines now taking shape across publishers are, in effect, attempts to turn those questions into checkable requirements — one imprint, one journal, one submission at a time.