Academic journals routinely reject AI-generated submissions. But if an article meets rigorous standards and its origins are disclosed, what exactly is lost by publishing it?

Peter Singer

I am a co-editor of the Journal of Controversial Ideas, an open-access, peer-reviewed journal. Five years ago, my co-editors and I launched the journal to combat the narrowing of the range of ‘acceptable’ issues that can be raised in the humanities and social sciences. We do not reject well-argued articles because their topics are too sensitive, and we allow authors to publish under a pseudonym if they fear that publishing their controversial ideas will harm their careers, or lead to personal abuse and death threats.

Like many academic journals, we now receive submissions that have been substantially or entirely written by AI, or more specifically, large language models. We began using Pangram, a commercial AI detection tool. A 2025 study by Brian Jabarian and Alex Imas tested 1,992 human-written passages with matching AI-generated ones. The texts were drawn from six different genres: news, blogs, consumer reviews, novels, restaurant reviews, and résumés.

The authors report that with medium to longer texts – the category into which all our submissions fall – Pangram had false positive and false negative rates of “essentially zero”; that is, there were virtually no passages written by humans that Pangram reported to be generated by AI, and virtually no passages generated by AI reported as written by humans. Even when texts were re-processed by AI “humanisers” that claim to remove traces of AI generation, Pangram correctly reported nearly all longer texts produced by AI.

Admittedly, none of the passages tested was an academic article. It is difficult to see why the results of the study would not carry over to academic writing, but that would be worth checking.

AI’s abilities are increasing very rapidly, but assuming that the Jabarian and Imas study is still accurate, we face the further question: what should we do with these papers? Among the organisations and publishers involved in setting standards for academic journals, the question is now considered settled: AI cannot be a co-author.

Because my co-editors and I are philosophers, probing widely accepted assumptions is what makes us tick. So, we have started to discuss whether it is sound to assume that academic journals should not publish articles known to be substantially AI-generated.

My own view, currently under discussion with my co-editors (and not the view of the Journal of Controversial Ideas), comes with a caveat: Because contributions to our journal come primarily from philosophers, as well as others in the humanities or social sciences, I address the question as it applies to such a publication. Those publishing in other fields and formats can assess how relevant my view is to their own.

Peer-reviewed academic journals exist, in my view, to advance our knowledge and understanding by publishing the best work being done in a given field. To that end, we ask reviewers to filter out articles that are poorly written, insufficiently precise in their claims, say little of importance that is new, use weak arguments, draw on disputable evidence, overlook or fail to respond adequately to objections, or misrepresent positions that they are arguing against. In this way, we strive to publish articles that will advance our knowledge and understanding of an important topic, and hence be worth the time we spend selecting and publishing them, and the time our readers spend reading them.

If that is the goal, and the best reviewers we can find – both human and AI – agree that an article generated by AI meets that standard, why should we reject it?

Obviously, anyone who claims to have written an article that was generated by AI is misrepresenting their achievement. But this is an objection to publishing an AI-generated article without disclosing its provenance, not to the publication itself.

The objection can be addressed by asking contributors to state their role in creating the article. They might choose from: author; co-author with other humans or with AI; prompter, where a specific prompt led the AI to generate the article; or corresponding agent, where the human role was limited to submitting the article and engaging with the editors about any revisions that might be required.

Editors could then check that the stated human role matches the best available evidence (currently, as mentioned, by using Pangram). If the article is accepted, the contributor’s role and the detector’s estimate of the extent of AI generation would be published alongside it.

This approach will remain viable only as long as the best AI detector remains ahead of the best AI humaniser. Perhaps rewriting a paper over and over again will reduce the risk of detection; or submitting many different rewritten papers may lead to at least one escaping detection. Unfortunately, if rewriting AI-generated papers to avoid detection becomes a profitable business, we are likely to see an arms race in which those with larger financial resources have an advantage.

But the current standard of simply rejecting AI-generated papers faces the same problem, and, importantly, misses out on the benefits of publishing AI-generated papers that make an important contribution to knowledge. As AI continues to increase its power, that loss will continue to grow.

Perhaps there are good reasons for rejecting my view, and if so, I hope to learn what they are. But I suspect that resistance to it is also an emotional response, rooted in the persistent human pride that once sustained the belief that the universe revolves around our planet, and refused to accept that humans descended from other animals.

https://www.project-syndicate.org/commentary/legitimacy-of-ai-generated-academic-articles-by-peter-singer-2026-11

PLEASE VISIT:

YOURDEMOCRACY.NET RECORDS HISTORY AS IT SHOULD BE — NOT AS THE WESTERN MEDIA WRONGLY REPORTS IT — SINCE 2005.

         Gus Leonisky

         POLITICAL CARTOONIST SINCE 1951.

         RABID ATHEIST.

         WELCOME TO THIS INSANE WORLD….