skip to Main Content
Being A Journal Editor In The Age Of AI

Being a journal editor in the age of AI

I am an editor for a number of academic journals, and I am also the Editor-in-Chief of Structural Engineering International. In that role, I read every paper that makes it through the initial checklist by the editorial assistant. I assess whether the paper fits the journal and whether it should be passed on to one of the editors to handle.

That means I see a lot of papers.

And, in the age of AI, I seem to be seeing even more papers.

Of course, generative AI can be useful. It can help authors improve language, brainstorm, structure ideas, or overcome some of the barriers that come with writing in a second or third language. I don’t think the answer is to pretend that these tools do not exist.

But from my side of the editorial desk, I am also seeing some developments that worry me.

Here are some observations from the past months:

  • Some papers look like the first thing ChatGPT spit out. I don’t mean that the English is polished. There is nothing wrong with using tools to improve your writing when journal policies permit it. I mean manuscripts that feel like a first generated draft that nobody has critically read, revised, questioned, or made their own. Academic writing still requires thinking. If you wouldn’t submit your first human-written draft, why would you submit your first AI-generated one?
  • The volume has increased enormously. Generative AI has lowered the barrier to producing something that looks like an academic paper. Unfortunately, it hasn’t lowered the barrier to producing good research. For editors and reviewers, this creates an additional workload in a system that was already relying heavily on voluntary academic labor.
  • What the fuck, authors? Check the scope of the journal. Yes, I said it. I edit a journal in structural and bridge engineering. I have received manuscripts about medicine. I don’t know whether authors are asking AI for a list of journals, whether papers are being submitted automatically or semi-automatically, or whether people are simply firing manuscripts at every journal they can find. Whatever the cause: please read the aims and scope before submitting. This was basic publishing hygiene before ChatGPT, and it still is.
  • The writing is becoming strangely uniform. Certain phrases, sentence structures, introductions, and ways of presenting conclusions appear again and again. Everything is polished, balanced, comprehensive, and rather beige (and I only like beige for nude stilettos, for the rest, I hate beige and beige sofas with a vengeance). A paper doesn’t need literary fireworks, but I do want to hear the researchers’ thinking. Scientific writing should communicate clearly, but it should also communicate what the authors actually did, found, and understood – and I still want junior academics to learn how to find their own voice.
  • I worry increasingly about what sits underneath the words. A polished manuscript can hide weak methodology, questionable references, invented details, or data that cannot be reproduced. That means I find myself paying more attention to whether the research itself hangs together rather than being impressed by a smooth-looking manuscript. This is probably a good development, but it also makes editorial assessment more demanding. And a fabricated dataset is very hard to spot, but apparently more common in the age of AI.
  • The policies are developing, but the practical implementation is less obvious. Taylor & Francis currently requires transparency around generative AI use and does not permit generative AI to create or manipulate images, figures, or original research data for publication. It also explicitly tells editors not to upload unpublished manuscripts into generative AI systems because of confidentiality and intellectual-property concerns. These rules make sense to me. What is less clear in everyday editorial work is how we reliably check compliance.
  • We need to avoid turning editors into AI detectives. I don’t think my job should become guessing whether paragraph 17 was written by a human or a language model. Detection is difficult, and suspicion based on writing style alone can create its own problems. The more useful questions are: Is the research sound? Are the methods transparent? Are the references real and appropriate? Can the results be traced back to the reported data and analysis? Have the authors taken responsibility for everything in the manuscript?

Perhaps that last point is where I currently land.

AI does not remove author responsibility.

A manuscript is not good because it sounds academic. It is good because it contains rigorous research, communicates that research clearly, engages properly with previous scholarship, and allows readers to understand how the authors arrived at their conclusions.

As editors, I think we will need better tools, clearer procedures, and continued guidance from publishers. We may also need to rethink editorial workflows if submission volumes continue to grow. But authors have a role here too. Before you press submit, read your manuscript. Check every reference. Check every figure. Check your data. Check the scope of the journal. And ask yourself whether you are willing to take responsibility for every sentence carrying your name.

AI can assist scholarship.

It cannot take responsibility for scholarship.

And in the meantime, I’m hitting that reject button A LOT.

If you are an author, reviewer, or editor, what changes have you noticed in academic publishing since generative AI became widely available?

Share with your peers!
This Post Has 0 Comments

Leave a Reply

Your email address will not be published. Required fields are marked *

Back To Top

Free Templates for your Research

Sign up here to get access to worksheets for your research that help you have more efficient meetings, reflect on your work, and plan your month. Suitable for anyone from Master’s thesis students to full professors!