In his widely cited – and widely misunderstood – book The End of History and the Last Man, historian and author Francis Fukuyama was heavily criticized for saying that the West had ‘won’ and that there would be no meaningful conflicts to trouble the world with the fall of the Iron Curtain in the late 1980s. The following 40 years or so seem to bear this criticism out, with 9/11, the Russian invasion of Ukraine, and most recently the war in Iraq providing ample evidence that history is indeed still happening.
However, whenever we observe a phenomenon occurring, it is inevitable that someone will try to predict when that occurrence will cease. Predictions are always a fool’s errand, but that doesn’t mean thinking about them is a waste of time. And often we can logically put an argument together that leads us to the correct conclusion, even if external factors and unseen complications conspire to make a robust argument seem wholly irrelevant.
Trouble at t’mill
Recently, I was interviewed by The Economist for an upcoming news feature on the impact of AI on academic publishing, and more specifically about how increased AI use might impact predatory publishing practices. Increases in article submissions have been reported across the board by publishers in 2026 over and above the record levels already seen in 2025. The reason, it is assumed, is that some authors are submitting ‘AI slop’ to journals – taking advantage of easier AI-assisted workflows to submit more articles more often to more journals. Whether these are pure-play AI-generated articles, or merely that the use of AI’s translation capabilities has eased the workload for many non-English speaking authors, the net result has been a huge increase in article submissions to journals and repositories alike.
One consequence of this increased use of AI to produce articles is that usage of paper mills may actually decline. The reasoning is that, assuming some authors follow a path of least resistance in their publishing strategy, they will see that the benefits of paper mills (ready-made articles, supplied for a price) are less than the benefits of using AI to write something for them (ready-made articles, supplied for free). While paper mills also sometimes offer guaranteed publication, the ease with which you can create multiple articles and submit them to multiple journals simultaneously means that this advantage is also cancelled out.
Logical conclusion
In terms of predatory journals, the impact on those is less easy to predict. On the one hand, the same path of least resistance may mean that authors will take their chances in legitimate journals with their AI slop and potentially reap greater rewards with a better publication. On the other hand, there is the temptation to use predatory journals to obtain even more publications, with the only barrier to this the relatively low APC charges.
How this will all play out is not known to any of us. By following established behavioral traits and following the latest data trends, we can simply see that a drop in paper mill usage is at least logically plausible, while the impact on predatory journal use could genuinely go either way. What we can see, however, is that these trends are outstripping the contradictory and seemingly ineffective attempts institutions and publishers have so far put in place to change the trajectory of this phenomenon. Whether it means the end of paper mills or not, the overarching consequences still don’t look good.