Hello fellow journalologists,
Traditionally, August was considered to be a quiet news month. Not any more. Perhaps journalists are lying on the beach and their AI agents are submitting copy for them; there’s certainly plenty to get your teeth into in today’s issue of The Jist.
If you missed the latest instalment of Journalology, which was sent on Thursday, you can read it here.
Staggering 90% of biomedical papers now show signs of AI help
The latest study also found that LLM use was more frequent in abstracts, introductions and discussion sections than in methods and results sections. An estimated 78% of discussion sections in December 2025 papers showed signs of AI, compared with 58% of results sections. AI-assisted results sections could be worrying, says Kobak, because of the propensity of LLMs to fabricate, or ‘hallucinate’, data. Using LLMs to write or edit introductions, meanwhile, could skew the leading ideas in a field. “Whatever bias the LLM may have will just suddenly permeate the literature,” says Kobak.
Nature (Kaia Glickman)
JB: Researchers are under pressure to publish and we shouldn’t be surprised by these statistics. The bottom line echoes my own thoughts on this:
Wrestling with thoughts and putting them into words is an essential part of the scientific process, she [first author Lena Holzwarth] says, and relying on LLMs to write papers eliminates this important struggle.
For a more detailed analysis of the preprint, I recommend you read Gustavo Monnerat PhD’s essay, published on Friday: Almost every biomedical paper now uses AI. Almost none of us were taught how.
Gustavo’s advice on how researchers should use AI is nicely done. He offers eight pieces of advice, the sixth of which is:
6. Disclose. “AI was used” tells an editor nothing. “An LLM was used to edit the English in the Introduction and Discussion; no text was generated de novo; no analysis was performed with AI” is a statement someone can actually assess. Journals are converging on this. Get ahead of it.
A story in Research Professional News (Universities lagging and wrongheaded on AI, report warns) covers the lack of AI training in universities:
He [Michael Zyphur] looked at the AI policies of 38 “top-tier” doctoral universities in 15 countries, as well as those of 14 funders and 18 publishers.
Zyphur found that in general the universities have “not caught up” with either developments in AI or policies adopted by other research organisations, with publishers being the earliest and most uniform movers.
Publishers may be leading the way, but we should put the horse before the cart; researchers need to learn how to use AI effectively and ethically.
Academic publishers must not get stuck in ‘war of AI against AI’, STM chief exec warns
The need for human review created a substantial staffing and financial burden, Sutton said, because each alert had to be assessed rather than automatically rejected. She described journal submission systems as being placed under particular strain by the volume of potentially problematic material. Sutton cautioned against treating all use of AI during research or manuscript preparation as misconduct. “There’s very legitimate ways to use it, and it can be helpful,” she said.
The Bookseller (Melina Spanoudi interviews Caroline Sutton)
JB: This story is a few weeks old now and ran before Anthropic announced the Claude watermark (Issue 152 of Journalology covered this).
Can Anthropic’s invisible watermarks curb ‘AI slop’? Researchers remain sceptical
The presence of the watermark reveals little about how the model was used. Detecting one “provides a signal” that content was made with Claude, says Anthropic, but is not conclusive: the model might have been used just to summarize or translate an original human idea, for example. Equally, a lack of a watermark doesn’t mean that the text was not generated by AI. Because the watermarks are based on patterns of subtle changes in a model’s word choice, passages that are very short, or that have been paraphrased or rewritten, might no longer carry a signal.
Nature (Elizabeth Gibney)
JB: The Nature podcast covered this story too.
Professors Were Singled Out for Using AI in Public Writing. Now They’re Defending Themselves.
“The issue is whether the use of AI produces better ideas, or a cleaner written text that communicates those ideas better. If so, use of AI should be encouraged rather than discouraged,” said Hausmann, a professor of the practice of international political economy. “The FT uses human editors to sharpen and shorten submitted draft op-eds, without changing the authorship of the person who sent them in. If this were done by AI, how would the outcome be different? How would it be worse? What is the harm that the policy is trying to prevent?”
The Chronicle of Higher Education (Alexandra Crosnoe)
JB: Here’s an alternate viewpoint:
But to David Perry, associate director of undergraduate studies in the history department at the University of Minnesota-Twin Cities and the author of a book on public scholarship, using AI to “edit” or “condense” is synonymous with using it to write. Cutting down an article or generalizing it for a broader audience is “what writing and thinking is about,” he added.
The Jist keeps readers up to date with news coverage of academic publishing. You can access The Jist archives here.
This AI tool claims to pick the top 1% of preprints. Should researchers trust it?
We are not offering our products to journals or publishers. Our goal is to provide free services to authors in a private, secure environment so they can improve the work before it is published.
People want to improve their research, but they are lacking good critical judgement to help them do that. We are not going to replace their judgement; we’re going to augment it and help them to spot things that would otherwise have been missed.
Nature (Miryam Naddaf)
JB: Miryam interviewed Niv Mastboim, one of the founders of QED Science. Unfortunately, the article fails to answer the question posed in the title: should authors trust it? The interview is great publicity for QED Science, but lacks any useful insight beyond what’s already been announced.
Some researchers are starting to test AI review systems (for example, this preprint: Benchmarking Agentic Review Systems). We need more independent verification and fewer puff pieces.
Peer review is overwhelmed—can it survive in the AI era?
It’s hard to envision a wholesale shift away from peer review because the system is so firmly entrenched in most fields, both as part of career progression and as a way of garnering public trust. At the same time, if academic publishing keeps growing at the rate it has been, the system simply won’t be able to keep up. “So something’s going to break there at some point,” Neylon said. Instead of dropping peer review, he suggests that maybe research fields should find ways to encourage quality publications over quantity.
Ars Technica (Saima Sidik)
JB: This news feature provides a broad overview of some of the challenges facing scholarly communication. I agree with this quote from Cameron Neylon:
Meanwhile, scholarly communication advocate Cameron Neylon has long been on the record as questioning whether all journal articles really need peer review, given how few papers end up having an impact on the world. Maybe, he suggests, studies should have to meet a threshold of interest before they’re sent out for peer review.
Google gives publishers a new way to fight AI-driven traffic losses
As AI continues to kill traffic to websites, Google on Thursday threw a bone to those publishers negatively impacted by the change. It’s now allowing readers to push a button on a publisher’s website to indicate it’s a “favorite source” they’d like to see highlighted more often across Google Search, Discover, and Google News. The tech giant said it’s making this new, interactive “Preferred Sources” button available to online publishers to embed on their own websites.
TechCrunch (Sarah Perez)
JB: Sales teams may also be interested in this article, published by Atypon a few weeks ago: Why native advertising outperforms display ads for scholarly publishers.
These more seamless ads also improve the reading experience compared to traditional banner advertising for users who don’t feel that they’ve been “taken out” of the content. Today’s audiences demand cohesive experiences when reading a scholarly article or navigating a publisher’s website.
Journalology provides insight into the key trends in academic publishing. Please consider supporting the newsletter by upgrading your subscription either for yourself or for your team.
‘Flawed’ use of non-significant data ‘killing further research’
Academic papers are routinely incorrectly treating “non-significant” results as proof that experiments had no effect, according to new research.
A paper authored by researchers from the universities of Manchester, Oxford and Arkansas warns that there is a “widespread misinterpretation of non-significant results”, which may be killing off areas of interest in future research.
The study, published in PNAS, outlines that “non-significant results” in academic research are typically interpreted as meaning “no difference” or “no effect”, “despite long-standing recognition that this is a fundamental misinterpretation”, it says.
Times Higher Education (Juliette Rowsell)
JB: Whisper it quietly, but some (many?) researchers only have a rudimentary understanding of statistics.
Company Offering ‘100% Human-Written, Never AI’ Medical Research Is Entirely AI
Research Gold, a site that advertises services for medical researchers, including drafting peer-review ready manuscripts, systemic reviews and meta-analyses, claims that it’s “100% human-written, never AI,” and lists a number of PhD reviewers and professional methodologists on staff that carry out this meticulous, difficult work. The problem: The PhD reviewers Research Gold lists on its site are AI-generated and don’t exist.
404 (Emanuel Maiberg)
Hundreds of paper-mill papers peddled in ads were later published
Now, a new study reports that such ads appear to have plenty of takers among IEEE conference participants: Out of 4407 unique paper titles listed in ads from 2021 to ’25, nearly half, or 2062, matched papers later published. (In computer science, full papers are often published in conference proceedings.) These are just a small fraction of the 300,000 conference papers IEEE publishes each year, but some meetings may be particularly vulnerable. At IEEE conferences that published at least 25 such papers, they made up more than 10% of all the meeting’s articles, and at three of those conferences, more than 20%, according to the study, posted Tuesday on the arXiv preprint server by social scientist Anna Abalkina of the Free University of Berlin and colleagues.
Science (Jeffrey Brainard)
JB: In my recent assessment of IEEE’s publishing programme I highlighted the newly enhanced research integrity team. They have a lot of work ahead of them:
So far IEEE has retracted 66 of the 2062 articles Abalkina identified as having a paper mill link. Abalkina says IEEE should move faster. IEEE has picked up the pace of retractions of conference proceedings overall: 1700 in 2025 and 1200 so far this year, Longobardi says. He did not have data on how many of these showed evidence of paper mill origin.
You can read the preprint here: Opening Pandora’s box: Paper mills in conference proceedings.
Authorship-for-Sale: From Fake Papers to Forensic Scientometrics
Alongside these efforts, systemic reform might also be necessary. Advocates of the Slow Science movement, for example, have called for key changes to the current publish-or-perish fast-science culture that incentivizes and enables paper mills. The group’s chief call is for a focus on quality over quantity when it comes to research output—fast science, like fast food, isn’t healthy; good science needs time to cook.
Journal of Medical Internet Research (Cliff Dominy)
Sleuth identifies dozens of studies that used the wrong antibody
More than 50 studies on cell ageing have apparently used the wrong antibody to identify a key protein in experiments, according to a science sleuth.
The latest case comes two months after Sholto David, a UK-based independent molecular biologist, identified hundreds of studies with a similar error and three months after he and another researcher spotted problems with antibody-validation images in the catalogue of one of the world’s largest suppliers, Thermo Fisher Scientific, headquartered in Waltham, Massachusetts. David reported the latest antibody mix-up in a 21 July post on the research-integrity blog For Better Science. It’s not clear whether the flaws compromise the studies, but David says they might in some cases.
Nature (Holly Else)
Sage issues dozens of retractions from two journals it acquired
In recent weeks Sage has retracted batches of articles published in two journals it acquired from IOS Press in 2023. In each case, the retractions were for manipulation of, and concerns about, peer review.
…
Since acquiring IOS Press in November 2023, Sage set the record for most retractions for a single journal when it retracted over 1,500 papers from JIFS. The move came after Clarivate put the journal’s indexing on hold for concerns about article quality. Its entry on the Clarivate website still shows the “on hold” flag. Chris Burnage, public affairs manager at Sage, told us the latest retractions were a result of a separate investigation from the one in 2024-2025.
Retraction Watch (Avery Orrall)
JB: Sage was not the only publisher to announce mass retractions over the past few weeks. Elsevier also made it into the Retraction Watch pages with Ceramics journal pulls dozens of papers flagged for integrity concerns.
Ceramics International began retracting the papers in July. More than 50 of the retractions were requested by an “impartial field expert acting in the role of an independent Publishing Ethics advisor,” according to the notices.
Most of the retraction notices – 43 – reference PubPeer comments calling out issues with the work. In emails we have seen, scientific sleuth Mu Yang flagged 80 papers from Ceramics International to the journal’s publisher, Elsevier, on April 23, 2025.
Elsevier also featured in another Retraction Watch story: Biology journal pulls 100+ articles and counting from special issues for peer review manipulation:
Elsevier’s International Journal of Biological Macromolecules has issued more than 100 retractions and counting for manipulated peer review in guest-edited special issues. The publisher has said one more retraction is coming.
Science or fiction? Shadowy ‘paper mills’ let you pay to be a published author – of fraudulent research
In a recent paper, which has not yet been peer reviewed, the trio of scientists led by Richardson, formerly at Northwestern University in the US, detailed more than 18,000 advertisements from seven paper mills across seven countries between 2020 and 2026.
“We wanted to get an idea, across a wide range of paper mills, countries, topics and product types how much these things actually go for,” he says. “What we put together is the largest ever dataset of paper mill advertisements.”
They found many groups advertising authorship of scientific manuscripts across Telegram, Facebook and various websites for as little as US$30.
The Guardian (Jackson Ryan)
JB: The preprint (BuyTheBy: A dataset of 18,710 text-based paper mill advertisements with 51,812 timestamped prices) was published back in April and it’s not clear to me why The Guardian chose to cover it now. Anna Abalkina is a co-author of this preprint and the one from the previous story.
Maybe scientific progress isn’t slowing, after all
That science’s best days are behind it, and its rate of progress is slowing, is an old claim. And, since science itself is a perfectly good subject for scientists to investigate, many have looked into it. One notable contribution came in 2023, when Michael Park, then a phd student at the University of Minnesota, and his colleagues published a paper in Nature. It analysed citation patterns in 45m scientific papers and 3.9m patents and concluded that the “disruptiveness” of both had fallen off a cliff since the 1950s. This was widely reported (including in The Economist). Its findings found their way into Congressional hearings and speeches by White House officials.
The Economist (unsigned)
JB: Nature occasionally publishes “Matters Arising” articles; independent researchers critique a previously published Nature paper.
The Economist story is pegged to a recent critique of a high profile 2023 Nature paper (Papers and patents are becoming less disruptive over time).
The rebuttal to the critique, written by the authors of the original paper, made me smile:
Twenty per cent of their sample is non-research content that almost by definition lacks references. Simple keyword searches highlight the problem’s severity, identifying among others 456 For Dummies guides, 50 Dr. Seuss and Curious George books, and the Captain Underpants series, all without references, in their sample.
In case you’re wondering, this is the second time a reference to the cartoon “Captain Underpants” has appeared on nature.com.
Large language models reduce originality of research proposals
The use of large language models (LLMs) has increased dramatically in US research funding proposals since 2023 and has led to less original and more generic ideas that often replicate prior projects, new analysis finds.
A paper published in Proceedings of the National Academy of Sciences on 11 August examined 5,700 confidential grant proposal submissions and 131,000 publicly released awards for grants from the US National Science Foundation (NSF) and National Institutes of Health (NIH) between 2021 and 2025.
Times Higher Education (Rosalind Skillen)
JB: You can read the PNAS paper here: The rise of large language models and the direction and impact of US federal research funding.
When the best decision is no decision: the rise of randomization in grant funding
More evidence is on the way. [Tom] Stafford and Dan Penny, director of market intelligence at Springer Nature in London, recently surveyed some 4,000 researchers worldwide on how funding should be allocated, with partial randomization among the options. (Nature’s news and careers teams are independent from Nature’s publisher, Springer Nature.) The results, expected later this year, could show whether the scientific community is ready to formalize the luck of the draw.
Nature (Max Bennett)
JB: See also: NIH proposes major revamp of how it scores research grant proposals.
AI Helps Researchers Win NIH Grants. Will Science Suffer?
New research shows that scientists who rely heavily on artificial intelligence to write grant applications are more likely to get funding from the National Institutes of Health. But that tactic could come at the expense of exploring more novel scientific ideas.
That’s one of the central takeaways of a study published this month in the Proceedings of the National Academy of Sciences, which analyzed more than 125,000 grant applications—including funded, unfunded and pending proposals—submitted to the NIH and National Science Foundation from 2021 to 2025. The study’s time frame encompasses the rapid rise of generative AI tools that first became widely available in late 2022. Using word-distribution modeling to identify proposals suggestive of high levels of large language model involvement, the researchers identified a surge in AI use between 2023 and 2025.
Inside Higher Ed (Kathryn Palmer)
AI isn’t ready to research itself
But the AI system mostly failed at its two assigned tasks, earning overall scores of 2/6 and 1/6 from the original papers’ authors. A typical way in which it would fail was to select a few hypotheses to explore, but settle too early on one and not backtrack sufficiently when its approach wasn’t working. Subsequent self-review wasn’t sufficiently negative, so the system persisted on its initial choices, whittling down its claims until it said little of interest. Another reason for its poor marks was that the system didn’t fully follow instructions or appreciate context, and failed to present its work well: it left a lot of time and compute unused, and it produced poorly written and poorly formatted papers.
Nature (Matthew Hutson)
Join the Journalology coaching programme
Most business coaches work across multiple industries and are unable to provide useful strategic insight relevant to scholarly publishing. The Journalology coaching programme is different.
If you would like to work with me one-to-one, please click the button below to learn how I could help you.
And finally…
Some journals encourage authors to send presubmission enquiries. I suspect many have turned off that option because AI makes it so easy to send enquiries to multiple journals. One news outlet has adopted an even more stringent policy: Facing a “deluge” from AI, this publication will only take pitches by phone.
Raskin told me that she’s noticed an increase of AI-generated story pitches since The Food Section upped its pay rates for freelancers on July 17. Raskin estimates that a quarter of recent pitches were “wholly AI, meaning neither the writer nor the proposed story appears to exist in the real world.”
I’m always grateful when journalologists alert me to interesting stories. I even appreciate receiving press releases. Please don’t call me, though.
Until next time,
James
P.S. If you think your colleagues would benefit from reading The Jist, please share this newsletter with them.