AI and Academic Publication: The Good, the Bad, and the Ugly
TL;DR
The task for journals is not to pretend that AI can be excluded from scholarly writing, but to decide how it should be used, disclosed, governed, and judged.
Abstract
Artificial intelligence has entered academic publication. It is already present in drafting, editing, translation, formatting, reviewing, and manuscript preparation. It is used by students, early-career researchers, senior scholars, reviewers, editors, and publishers. Some uses are minor. Some are substantial. Some are legitimate. Some are not. The task for journals is therefore not to pretend that AI can be excluded from scholarly writing. The task is to decide how it should be used, disclosed, governed, and judged. Academic writing has always involved tools. Authors use word processors, grammar checkers, reference managers, statistical software, translation programmes, plagiarism-detection systems, and journal submission platforms. These tools have changed the mechanics of writing and publication. Generative AI changes more than mechanics. It can produce fluent prose. It can summarise complex material. It can imitate disciplinary styles. It can create abstracts, titles, responses to reviewers, tables, outlines, and plausible literature overviews. It can also invent references, distort arguments, conceal weak reasoning, and produce text that appears scholarly without being scholarly. This editorial offers a simple position. AI is not an author. AI is not a scholar. AI is not a source of academic responsibility. It is a tool. Used well, it can support clarity, accessibility, and editorial preparation. Used poorly, it can damage quality. Used dishonestly, it can threaten trust in academic publication. The distinction matters. A journal should not treat all AI use as misconduct. Nor should it treat all AI use as harmless assistance. The proper standard is human accountability.