📊 Half the papers in two elite journals came from a few dozen people. An audit of fifteen years of the FT50, the list business schools treat as the only one that counts, found that a few dozen people wrote about half of everything in two of its titles. That is not a field. It reads like a regulars' table.

Why it matters: Across those fifteen years the audit found 946 ties between an author and an editor. Half of them carried at least one clear conflict of interest, with another 11.6% possibly conflicted. The group at the centre is also strikingly narrow: 90% did their doctorate in North America and 5% are women. It is a small room. The stakes are cash as well as standing, because some business schools pay a bonus of up to 22% of base salary for a hit in one of these titles.

What I would do: Find out whether your institution rewards a specific list in writing, because plenty do it by habit and deny it on paper. Then look at how many recent articles in your target journal come from people already inside it. If the answer is most of them, you are not competing on quality alone, and a wider journal one rung down may get your work read sooner and cited more.

Reference: Ciriello and Mindel, Journal of Business Ethics, 2 Sep 2026

Coverage: Times Higher Education, 2 Sep 2026 · 2026-09-02 · 4 min

🤖 30,000 papers each got an AI reviewer, and the authors liked it. The AAAI conference took 30,000 submissions this year, double last year's, and gave every author a machine-written report alongside the human ones. It then asked nearly 6,000 authors, reviewers and judges what they made of it. The verdict is not the rout either camp was expecting.

Why it matters: The pressure behind this is arithmetic, not enthusiasm. Science reported this week that the world now publishes over 8 million papers a year, twice as many as five years ago, and that finding people willing to read them is getting harder every year. Something has to absorb that. In a separate test, GPT-5 was turned loose on 2,500 papers that had already been accepted at leading machine learning conferences and flagged roughly a third of them as carrying at least one error that could affect interpretation or reproducibility. Human checkers agreed with 83% of those flags.

What I would do: Write for a machine reading first, because increasingly one does. The mistakes it reliably catches are the boring ones: a figure that appears one way in the text and another in the table, a total that does not add up, a script that will not execute. Those cost an hour the week before you submit and a month in revision.

Reference (preprint, not peer reviewed): Kim et al., head-to-head AI and human review, arXiv preprint

Coverage: Science, Vol 393 Issue 6815, 3 Sep 2026 · 2026-09-03 · 5 min

Open the editorial board page of the journal you are about to submit to. Take the five names at the top. Search each one inside that same journal and count what they have published there since 2023. Then find the journal's own reported time from submission to decision. You now know two things you did not know this morning: whether the board publishes in house, and what a normal wait actually looks like, so you can tell a slow decision from a stalled one. If the board is heavily represented and your paper sits in the same subfield as theirs, that is not a reason to give up. It is a reason to have your second journal chosen before the first one answers.