Normalizing

Normalizing

June 20th marked the arrival of the summer solstice, and this mid-year event along with Clarivate’s release of the 2025 Journal Citation Reports (JCR) signal that the bibliometric data from 2024 is reaching a point of trustworthiness. At C&E, we are heads down into the numbers as we update and prepare new analyses for our 2025 Journal Market Trends Report (expect report release this fall).

While we’ve only had a few days with the new JCR as of this writing, one thing stood out, namely that 2024 is the first year since 2018 that looks “normal” in terms of publication volume growth. In a typical year (e.g., in the years from 2014 to 2018), the number of articles in the Web of Science (WoS) grows around 3%–4%. That growth means more citations are included in the data, and the Journal Impact Factor (JIF) for any given publication should, if things hold steady, increase a small amount each year due to inflation. However, as seen in the chart below, things got really weird from 2019 onward. The literature first expanded significantly and then contracted, with three years of outlier growth followed by two years of decline in the number of articles indexed by WoS. If your journal’s JIF went way up during the pandemic and has now fallen way down, this may be a big part of the reason why. In 2024, the data normalized, seeing a 4% year-on-year growth in the number of articles published, about 111,000 more than in 2023. 

Scale continues to be a differentiating attribute in the market (a “competitive moat” as they would say in business school), as open access (OA) revenues are directly correlated with publication output. In the megajournal sphere, being put on hold for WoS indexing in 2024 does not seem to have harmed the performance of Cureus or Heliyon, as both journals improved their JIF scores and saw enormous leaps in publication volume. Looking at article output for the top 10 largest journals in the market (well, top 11 because we include Journal of Physics: Conference Series,which operates on a different model than the rest of these author-pays OA publications), every single journal saw growth in 2024, with the exception of those from MDPI (International Journal of Molecular Sciences, Sustainability, and Applied Sciences). The highest levels of growth were all experienced by megajournals operated by Springer Nature and Elsevier, the largest publishing houses, suggesting that the ability to transfer rejected articles from a huge corpus of journals is a distinct market advantage.

So, with the literature now growing at a non-pandemic-impacted rate, and with the biggest publishers throwing their weight around, 2024 looks, dare we say it, “normal” – although with all the chaos of 2025, perhaps we’re just being nostalgic for a simpler time.

Calm Before the Storm

That sense of normalcy may evaporate quickly, as the National Institutes of Health (NIH)’s newly accelerated Nelson Memo public access policy has come into effect as of July 1, 2025. The Authors Alliance put together a superb overview of the policy and what it means for authors. One thing this guidance makes clear (a point that The Brief made back in December of last year) is that, while the NIH and authors can claim prior rights to post their papers in a repository without an embargo, publishers can require authors to pay an article processing charge (APC), or as we stated in early 2021, simply refuse to publish any articles making such an assertion.

The impacts of this NIH policy (and of all the other Nelson Memo policies going into effect in 2026) are likely to be unevenly distributed. Estimates from the Office of Science and Technology Policy (OSTP; and Dimensions data corroborate these) suggest 240,000–250,000 papers per year are based on federally funded research. This accounts for only around 4% of the current market, and the raw numbers are likely to decrease in coming years with a 35% projected reduction in the number of grants the NIH will be issuing. NIH funding is highly competitive and difficult to obtain, and the more selective journals publishing this high-profile research may be more vulnerable.

The Authors Alliance lists Sage and the American Association for the Advancement of Science (AAAS) as major scholarly publishers that will allow zero embargo availability of accepted manuscripts with no charge (our understanding is that the Institute of Electrical and Electronics Engineers (IEEE) allows for this as well). Springer Nature, by contrast, states that they will “enable compliance” of papers with zero embargo requirements in their Gold OA publications. Wiley makes it very clear to authors that there is “one route to compliance,” namely that authors choose OA publishing and pay an APC. Taylor & Francis (T&F) states that the NIH policy is “not compatible” with most of their journals’ sharing policies and authors should therefore “choose a gold OA publishing option for their article.” Other major publishers have not yet updated their compliance statements, which remain oriented to 12- to 24-month embargo requirements (e.g., Elsevier). It is not yet clear whether these will change to follow the course of Sage and AAAS or that of Springer Nature, Wiley, and T&F. As it stands today, publishers without an explicit statement are not enabling compliance with the Green OA route, and by default seem to be only supporting a Gold OA route for NIH authors (but they have not directly said so).

The performance of publishers with different approaches to zero embargo policy compliance will bear watching – will federally funded authors flock to the journals of publishers that allow Green OA deposit without fees, or will they embrace Gold OA publishing? How will libraries respond as they realize that increasing numbers of articles in those journals can be obtained without paying for subscriptions?

Depending on their internal policies, publishers may need to adjust the status of their PubMed Central (PMC) deposit agreements with the NIH (or, in certain cases, cease to have a deposit agreement, leaving authors to fend for themselves).

With an enormous anticipated reduction in its overall budget, thousands of employees laid off already, and reports of 40% staff reductions at the National Library of Medicine (which manages PMC), it remains unclear exactly how the NIH plans to handle the likely significant spike in support needs from authors. Many of these authors will be trying to navigate the new policies and some may be required to handle deposit on their own rather than relying on publishers to do the work for them. And then the NIH also needs to track and enforce compliance. Perhaps also of note: the NIH Frequently Asked Questions (FAQs) page about the Public Access policy makes no mention whatsoever of research data, a specific focus of the Nelson Memo.

Fair-ish Use

Two major court decisions were announced this month in cases brought against AI companies for the potential infringement involved in using copyrighted materials to train large language models (LLMs). The verdicts are somewhat contradictory, and in combination with an earlier decision, do little to clarify whether LLM training will ultimately be considered fair use or will require permission from and compensation to copyright holders.

In a case brought by authors against the company Anthropic and Claude (its AI model), Judge William Alsup ruled, essentially, that AI training is indeed fair use, at least under specific conditions, because it is “highly transformative” (one of the four factors considered in whether something is fair use). However, Alsup also noted that much of the material that Claude was trained on was pirated and obtained illegally, and that the legality of that action will be determined at a later date (with a heavy suggestion that this behavior is going to be highly problematic for Anthropic). In a separate ruling, Judge Vince Chhabria similarly agreed that AI training is transformative enough to be considered fair use, but suggested that the outputs of AI are potentially infringing and that authors could later sue separately over those (although he was hamstrung by the weak arguments made by the plaintiffs in this case and couldn’t rule directly on that issue).

These mixed verdicts continue to complicate the issue for both publishers and AI companies. The two judges seem to agree that AI training data sets must be obtained legally, through purchase or licensing, something that OpenAI has argued would be financially ruinous. However, as this is the same company that recently announced it seeks to raise $500 billion for investment in AI infrastructure, their definition of “ruinous” is unclear (perhaps they meant to say, “mildly inconvenient”?). Furthermore, Chhabria suggests that Alsup failed to focus on potential market harm for authors, and both judges left the door open for later cases to be raised, limiting their decisions to the details of the specific cases brought before them.

These two cases (along with the earlier Thomson Reuters verdict) have done little to add clarity for publishers, other than to make clear to AI companies that piracy is likely to be more costly in the long run than negotiating “ruinous” licensing deals. Hopefully things will become clearer as more cases (and the likely appeals on these cases) wind their way through the courts.

Wiley Gets Its Mojo Back

Wiley reviewed its estimated Q4 earnings for its 2025 fiscal year, and it was upbeat for the first time in a while. Highlights include:

• $1.66 billion in revenue for FY25, up 3% (on an adjusted basis, factoring out divestitures) from FY24
• Research division revenue up 3% over FY24
• 78% of research revenue comes from outside the US (a noteworthy statistic at this particular juncture)
• Article submissions up 19% and article output up 8%

On that last bullet point – while it is positive news that Wiley’s submissions and published output are growing, the market shift to OA models in which compensation is directly correlated to output, really puts a spotlight on this metric. Here, Wiley is being outpaced by Springer Nature and Elsevier, which continue to grow much faster than the rest of their commercial counterparts.

Doug Way has put together a brief analysis that is worth your time. One of his observations is that if you back out one-time AI licensing revenue, Wiley’s revenue is essentially flat:

…AI licensing revenue may be masking broader challenges. If you back out the $40M in almost entirely one-time AI licensing, Wiley’s revenue was essentially flat ($1.620B vs. $1.617B). If you back out the AI licensing revenue from the research segment, you see growth drop from 3% to 2%. Given typical inflationary increases for journal subscriptions (2–3% for packages and often 4–6% for title-by-title journals) and the high margins associated with backfiles and archives, one might expect stronger organic growth. That could signal broader “softness” in Wiley’s business, particularly as many of the largest research libraries confront uncertainty and serious budgetary constraints this year and beyond.

Caveats around AI licensing can perhaps be put to the side, as it seems reasonable to assume that Wiley will again be able to find $40 million in AI licensing next year.

Wiley appears to have emerged from the turbulence of the last couple of years – Hindawi’s collapse, followed by its clean-up, and the transition to a new CEO – with a good story to tell, even as it heads into the storm clouds that are gathering over the academic sector in the US.  

Briefly Noted

The consequences of federal policy are spreading beyond the funding and direct performance of research. Meetings programs are seeing strong declines in attendance from international attendees hesitant to come to the US. This, combined with government employees not being able to attend, has impacted the budgets of many research societies, which rely on income from meetings, as well as membership dues and publications, to fund the work they do for their communities. This in turn puts more emphasis on the importance of earnings from publication programs. Elsewhere, widely used resources are under threat, including the defunding of the Drosophila research database FlyBase, which seems to be collateral damage from the administration’s attacks on Harvard. 

In a roundup of other US government news, attacks on research journals continue; the administration’s new guidance toward what they call “Gold Standard Science” threatens to invalidate research results based on political interests; universities and research organizations are banding together to propose ways to reform the structure of federal funding (as opposed to the proposed reduction to a 15% indirect cost rate cap); and leaders of the American Academy of Family Physicians, American Academy of Pediatrics (AAP), American College of Physicians, American College of Obstetricians and Gynecologists, and the Infectious Diseases Society of America are publicly voicing their commitment to continue to recommend vaccines, with the AAP boycotting the administration’s newly revamped Advisory Committee on Immunization Practices (ACIP).

In keeping with the general chaos of the moment, Axios reports that the federal government has canceled $20 million in subscriptions to Springer Nature journals, suggesting the rationale for the cancellations is not budgetary but rather the “wokeness” of the German publisher. The magnitude of the cancellations (if any) has been countered by a representative of Springer Nature. Meanwhile, Sciencereports that some Springer Nature subscriptions have indeed been canceled but others haven’t (maybe).

Troubles continue at the National Academies of Science, Engineering, and Medicine, which is facing a 30% reduction in budget and staffing.

As the US increasingly falters as the world leader in scientific research, China continues to expand its efforts and influence, with “the number of collaborations with researchers in China increasing,” albeit at a slower pace in some areas than others.

C&E’s Michael Clarke was quoted in Science’s coverage of a recent Research Consulting white paper, “Safeguarding the future of society publishing.” This study is an expansion on work released last year, but at least to our eyes, the conclusions remain unclear. While in our experience the key result – that societies are generally seeing declining revenues from their publishing programs – rings true, it’s difficult to say much more without access to the dataset used for the analysis. Results from 66 UK research societies may not accurately reflect the situation for groups in other geographies; it’s unclear what is meant by journals that do “some outsourcing” (and if that includes those in publishing partnership agreements); and, as we have noted previously, gross revenue is not as important a measure as net earnings (a society might bring in less revenue in a partnership than it did self-publishing, but at the same time drastically reduce costs, resulting in a net gain). Nonetheless, the study raises awareness of the challenges facing societies related to publishing and data management.

A new AAAS survey suggests that researchers continue to have a poor handle on open (largely Creative Commons) licenses and what they allow or don’t allow.

The harvesting of information from the internet continues, often with devastating effects on websites and repositories while offering seemingly little value in return. The statistic that astounded us is the ratio of number of pages crawled to traffic sent. For Google this ratio has gone from 2:1 ten years ago to 18:1 today. For OpenAI it is currently 1,500:1 and for Anthropic it is a mind-boggling 60,000:1. One intriguing possible solution being floated is the creation of paywalls, or “online tollbooths,” with Cloudflare launching a new product that will charge AI bots for scraping your website.

Perhaps borrowing a strategy from Blade Runner’s Tyrell Corporation, Springer Nature was taken to task recently for their “Discovery” line of journals that the author of this piece suggests are more MDPI than MDPI.

The world of research “sleuths,” those tireless watchdogs performing post-publication peer review and finding fraud and errors in the literature, is becoming more structured and formalized. A two-year, $900,000 grant has been issued by Open Philanthropy to fund The Medical Evidence Project. The forensic metascience group will seek out “medical evidence which is bad and/or fake, especially when it contaminates treatment guidelines and kills people,” according to founder James Heathers.

In other research integrity news, Mike Rossner offers a strikingly clear argument as to why conducting replication studies is neither an effective nor efficient mechanism for combatting reproducibility problems.

Nature announced that going forward, all articles published will be accompanied by their (anonymized) peer reviews. While most reaction has been positive, it has been noted that transparent peer reviews is limited strictly to Nature’s flagship journal and that a similar policy would go a long way toward increasing trust for lower-prestige journals such as Scientific Reports. In response, Samuel Moore offers a thoughtful essay on the difference between transparency and governance, which includes this quote: “So much of open advocacy hinges on the idea that making materials openly available will improve the scientific process, although there is much less written on how specifically it will improve science and even less that directly shows the benefits of transparency.”

ASAPbio suggests that, despite huge gains made during the Covid pandemic, preprinting remains a minority activity in the life sciences, involving only an estimated 13%–14% of articles.

PubPub appears to be the latest victim of the “Red Queen’s Race,” as Knowledge Futures (KF) is shutting down the platform at the end of 2025. As KF notes, “Over the course of two years working with that model, we came to an unfortunate realization: the communities we served with PubPub Legacy, an easy-to-use, free academic site builder, largely did not have the resources to support its active development. The communities who did have resources required degrees of customization and bespoke features that the platform, as it existed, was not suited to provide.” As we noted back in 2019, community-led infrastructure that hopes to last for the long-term must find ways of self-sustaining, because eventually, the grants will run out.

Further evidence of “The Great Fragmentation” came this month as the Royal Society of Chemistry has revisited their 2022 plan to transition all of their journals to OA within five years, as it became clear that moving to a singular, author-pays business model would disenfranchise large segments of the chemistry community. We expect to see more organizations similarly putting the needs of their communities over idealistic, but self-limiting, one-size-fits-all strategies.

Beyond court cases determining licensing rights, AI remained at the forefront of the scholarly communications conversation, particularly as multiple reports coming out of the US Department of Health and Human Services were found to contain what appear to be AI-hallucinated citations to non-existent papers. Two really interesting articles this month discussed why AI peer review is a bad idea – the first looking at the ways AI algorithms will bias peer review and the adjustments authors will maketo win their favor (including hiding instructions to AI reviewers within the paper itself), and the second noting that AIs, as they are currently developing, lack the reasoning capability to do proper peer review, and allowing them to take it over would “relinquish our agency to improve the scientific literature.”

In what is perhaps a too-apt metaphor for the current moment, it has been revealed that Anthropic has destroyed millions of print books in order to scan them. Destructive scanning was faster than using a non-destructive scanning method such as Google has previously done to build its books database.

JSTOR announces JSTOR Seeklight, a human-in-the-loop AI-supported tool to help digital archivists.

Meanwhile, a new McKinsey report is bullish on the application of AI technologies to research and development (or at least on the money they can make by advising organizations about the application of AI technologies to R&D). The report notes R&D productivity has declined across many sectors (e.g., drug discovery, semiconductors, and agriculture). They posit that AI will help “bend the curve” of productivity in R&D, theoretically leading to more research outputs.

If you haven’t yet found the time, we highly recommend this piece from Karin Wulf, on the necessities of teaching the humanities in a STEM world.

June was a busy month for mergers and acquisitions, as Taylor & Francis acquired the books program at Amsterdam University Press, MIT Press acquired University Science Books from American Institute of Physics Publishing, and Knowledge Unlatched has once again been sold, moving from its initial unlikely acquirer, Wiley, to yet another seemingly odd fit in Annual Reviews.

“Networking Networks — A Festschrift in Honor of Clifford Lynch” has been published as a special issue of portal portal: Libraries and the Academy.

John Inglis has announced he is stepping down as Executive Director and Publisher at Cold Spring Harbor Laboratory Press. Although it is hard to imagine scientific publishing without John’s wisdom, we wish him many happy days ahead.

Congratulations are due to Dennis Lloyd, who takes office as the President of the Association of University Presses.

Change is afoot at cOAlition S as Executive Directory Johan Rooryck has announced his departure from the group and Robert Kiley ends his seven-year tenure as the organization’s Head of Strategy.

Richard Baxter, newly appointed as Managing Director of Kudos, has sadly passed away. Our condolences to his family, friends, and colleagues.

Finally, how much you trust pie and bar charts may depend on how much you know about their inventor, William Playfair, described here as, “a spy, a scoundrel, and a scholar.”

***
…it takes all the running you can do, to keep in the same place. —Red Queen to Alice, in Through the Looking Glass, quoted as a tragicomic description of maintaining scholarly communication infrastructure in this piece by Katherine Skinner