C&E Expands, Welcoming Five New Team Members
C&E is growing, and we want to share a little about why.
Over the past year, we’ve seen a material growth in client demand — driven in large part by the complexity of the environment and new opportunities created by market changes.
AI is certainly a factor, reshaping the landscape from multiple directions at once. It is creating new revenue opportunities tied to product innovation and AI licensing deals (and presenting thorny strategic questions along the way). AI is also introducing new capabilities across both the publishing workflow and enterprise-wide marketing and communications functions. At the same time, it is forcing societies and publishers to rethink brand relevance, audience strategy, and their role as stewards of trusted content — as content is increasingly discovered and consumed through AI-mediated workflows.
But the momentum extends well beyond the impact of AI. Some of our clients are launching and scaling education products and building out their product capability. Other clients are reevaluating the right partners for their magazines as the technologies and practices related to digital advertising, sponsorship, and digital marketing evolve. We are helping some societies take a hard look at whether remaining in a commercial publishing model continues to serve their long-term goals, and how they might reshape publisher relationships to better protect and advance the society’s interests over time. Nearly all of our clients are looking for ways to diversify and grow non-dues revenue — and that includes the importance of a diversified revenue strategy for publishing. And audience engagement, especially in light of new marketing technology capabilities, remains a priority across our client base.
To meet this demand, Jenn Saboe, Ginny Herbert, Marina Segel, and Neil Appleton join us as consultants.
- Jenn comes from Oxford University Press, bringing both experience working with society partners and strong digital and marketing chops. At C&E, she is part of our growing marketing and digital practice.
- Ginny comes from AIP Publishing with strong publishing expertise and particular depth in editorial and author experience. At C&E, she focuses primarily on publishing strategy projects.
- Marina spent nearly 30 years at Wolters Kluwer in operations and organizational transformation, with deep fluency in content workflows, production processes, and organizational change. At C&E, she helps clients streamline and modernize their operations, vendor partnerships, and workflows — especially in increasingly AI-enabled environments.
- Neil joins us from Elsevier, where he held progressively senior leadership roles culminating in executive responsibility for a global portfolio of more than 750 society partnerships. At C&E, he focuses on growth strategy and M&A readiness.
We’re also making a deliberate investment in our own voice. There’s a lot happening in this industry right now, and we would like to be more engaged in sharing our knowledge and perspective. Olivia Holway joins as Marketing Lead to help us do that — bringing six years of experience in marketing strategy, content, and analytics, and a strong interest in how AI is reshaping marketing practice.
We are privileged to work with some of the most ambitious organizations in our sector, and this team reflects our commitment to meeting them where they are. We are excited to welcome Jenn, Ginny, Marina, Neil, and Olivia to the firm. We are grateful for the trust our clients place in us — and excited about what we can do together with this expanded team.
All C&E team bios can be found here.
Rescinding?
1
The US Government Accountability Office (GAO) issued a report studying the potential budgetary effects of the Nelson Memo on federal agencies. As a reminder, the White House Office of Science and Technology Policy (OSTP) issued the “Nelson Memo” in 2022 under the Biden administration. That memo directed all federal agencies with research funding budgets over $100 million to develop policies that would result in immediate free public access to research papers (and data) based on research funded by that agency. While many (including us here at The Brief) expected the Trump administration to rescind the Nelson Memo shortly after (re)entering office, that did not happen (more on this in a moment). In fact, the National Institutes of Health not only embraced but accelerated implementation of the Biden-era policy.
The White House was required by Congress to provide an economic analysis of the policy. Such an analysis has never been provided. And so this GAO report is a welcome look at the budgetary implications of this policy.
OSTP under the Biden administration maintained a “have your cake and eat it too” narrative, maintaining that authors could continue to submit papers for free to hybrid journals and then deposit their papers in a federal repository contemporaneously with publication. This narrative, however, relied on an assumption that publishers would continue free publication in hybrid journals. Most of the largest publishers quickly determined that asking universities to pay for subscriptions to content they can get for free is not a viable business model and quietly began requiring open access charges for any papers (e.g., nearly all federally funded research) that require immediate public access.
With zero-embargo public access policies at federal agencies — and the associated policies of major publishers — now in effect, the budgetary implications for federal agencies are harder to hand wave away, as the GAO report shows clearly:
Using OSTP’s average publication charge estimates for pay-to-publish models in 2024, we estimate that the selected federal agencies spent a combined total of approximately $295 million on publication charges in 2024. Had publishers been requiring the use of pay-to-publish models for all publications at this time, like they have indicated they will begin doing for federally funded researchers, we estimate that the selected agencies would have spent between approximately $607 million and $715 million on publication charges that year. This represents a 105 to 142 percent increase in spending on publication charges.
Assuming publication counts and costs continue to follow recent historical patterns, annual expenditures on publication charges for the selected agencies could range from approximately $582 million to $937 million (in 2024 dollars) each year over the next 5 years.
This was predictable back in 2022 when the Nelson Memo was first released — and, in fact, we did predict it, writing days after the Nelson Memo was released that “A zero embargo will likely (eventually) make the subscription model unfeasible for many, if not most, journals and will necessitate increases in agency funding to support publication costs.” We went on to estimate the likely cost to federal agencies (in 2022 dollars):
Using an APC of $3,000 and a total annual paper output of 263,000 results in an (upper-bound estimate) of $789 million in annual publication costs in a scenario where 100% of federally funded research is published via Gold OA journals.
Our back-of-the-envelope calculation is right in the middle of the range estimated by GAO.
Despite the relatively straightforward math, most federal agencies have not budgeted for publication costs. The GAO reports that:
Most of the selected agencies have not conducted formal analyses or planned for the potential budgetary and other effects of shifting towards pay-to-publish models. This could be explained at least in part by agencies not fully anticipating the expected shift towards pay-to-publish models, as discussed above. NIH was the only agency that had begun conducting such an analysis during our review.
In other words, most federal agencies have continued under the premise that there would be no costs associated with the implementation of this policy, even as it has become clear there are very real costs coming due. NIH, trying to get out in front of the bill, is considering the idea of publication charge caps.
Meanwhile, the American Institute of Physics’ FYI reports that OSTP appears to be in the process of rescinding the Nelson Memo:
OSTP is now apparently in the process of repealing the Nelson memo, though what that will mean for existing federal open-access publishing policies is unclear. OSTP did not respond to a request for comment prior to publication. When Congress passed its 2026 appropriations minibus in January, a joint explanatory statement accompanying the budget asked OSTP to report on the status of its “process of repealing” the Nelson memo. A report accompanying the House Appropriations Committee’s recently published Commerce-Justice-Science appropriations bill for fiscal year 2027 says that the committee “recognizes that OSTP is in the process of repealing” the memo, and requests a briefing on this action. The committee also requested that the National Science Foundation “pause implementation of new public access policies” until OSTP has had time to repeal the memo, citing concern that NSF is continuing to “implement public access policies without administration guidance and coordination.”
It is unclear what this means for agencies or their budgets. Does rescinding the Nelson Memo leave in place the prior Holdren Memo that required public access to federally funded research but allowed a 12-month embargo window? Or does rescinding the Nelson Memo merely remove the requirement for agencies to mandate a zero-embargo public access policy, while still allowing them to do so if they wish? Or does rescinding the Nelson Memo mean replacing it with something else entirely?
It is becoming increasingly clear that the current administration is developing an allergy to the costs associated with research publication. As we reported last month, the administration’s proposed budget calls for a federal prohibition on “expensive subscriptions to academic journals and prohibitively high publishing costs unless required by federal statute or approved in advance by a federal agency.” FYI observes that budget requests by NASA and NIST echo this stance, both saying they will no longer use federal funds to pay for academic journals or publish research results (they may presumably use non-federal funds for these purposes).
The most likely result of this uncertainty — and possible federal prohibition or cap on paying for either subscriptions or author fees — is shifting the cost of publication to universities. We wrote about this scenario when caps were first mooted by NIH, writing in the July 2025 issue of The Brief:
In the end, even if implemented, price caps may simply fail to make any difference in what journals charge and what authors pay for publication. A Plan S parallel comes to mind: when cOAlition S ended financial support for APCs in hybrid journals, it did not have the desired effect on researchers, and in fact, publication in hybrid journals increased. This is due to the spread of transformative agreements, in which libraries strike deals with publishers that include both subscription access to journals and payment of APCs for campus researchers. Since the author is no longer using research funds to pay directly for the APC, cOAlition S has no control over their spending. Here we could see the same thing — increased uptake of transformative agreements by US institutions that would allow NIH-funded researchers to escape from restrictive caps, as the money being spent would not be coming directly from the NIH and would be out of their control.
In this sense, a prohibition (or cap) on the use of federal funds for publication fees is a haircut on university grant overheads. The administration has tried to reduce overheads directly, though such attempts were blocked by both the courts and Congress. Shifting around a billion dollars in publication fees from the federal government to universities is an indirect method of accomplishing the same thing (albeit to a lesser degree). However, given that universities have long paid the bulk of journal costs (via subscriptions), it is at least a shift that has some precedent.
Fake It Till You Break It
An audit published in The Lancet (Topaz et al.) of 2.5 million biomedical papers claims to have found 2,810 papers containing fabricated citations — references that could not be traced to any known publication. The study, which examined 97 million references from papers published between January 2023 and February 2026, found that the rate of fabricated citations in 2025 was more than 12 times greater than in 2023. The authors describe their findings as “conservative underestimates” — a lower bound on a problem they believe is far larger than what the audit could detect.
The steep increase in (reportedly) fake citations tracks almost precisely with the mainstream adoption of generative AI writing tools. Generative AI models are known to “hallucinate” — producing plausible-sounding but nonexistent citations with accurate-looking titles, authors, and DOIs. When researchers use these tools to assist with literature reviews or drafting, fabricated references can make it into submitted manuscripts, and from there into the published record.
It is worth digging a little deeper into what exactly Topaz et al. mean by a fake citation and how such citations present in an article. The examples provided in the paper and on the exceptionally polished marketing website for the paper show instances where a paper contains one or more citations that appear normal upon a cursory review. They often include real journal names and real authors. They have a full citation and even include a DOI and links to a PubMed and/or Google Scholar abstract. Clicking on the DOI resolves to a real journal article — just not the journal article listed in the citation. This is what makes them tricky to spot.
One of the papers highlighted by Topaz et al. is “Comparative analysis of ureteroileal anastomotic stricture rates: Bricker versus Wallace techniques in ileal conduit urinary diversion—a single-surgeon study with BMI-matched design and long-term follow-up excluding cancer recurrence bias,” published in Frontiers in Oncology. There are, according to Topaz et al., 18 (18!) fake citations in this article. The second one in the reference list is the following paper:
Hautmann RE, de Petriconi RC, Volkmer BG. Radical cystectomy for bladder cancer: morbidity, mortality, and oncological outcomes in 1,000+ patients. Eur Urol. (2023) 84:e45–50.
European Urology (Eur Urol.) is a real journal (a prestigious one, in fact). But there is no paper with this title that appears in the journal. It is made entirely up. The DOI that appears with this reference links not to European Urology but to a different paper (“Urinary Diversion: How Experts Divert”) with the same first author that appears in a different journal (Urology). The paper that it links to is a real paper, just not the one cited. There are 17 other references like this in this paper.
The implications extend well beyond academic publishing. Scientific literature is itself training data for the next generation of AI models. If that literature is increasingly contaminated with hallucinated citations, the models trained on it will inherit and potentially amplify those errors — a feedback loop with serious consequences for research integrity, clinical decision-making, and public trust in science.
The tools to catch this already exist — the audit itself used large language models to flag citation mismatches at scale. But burning through AI model tokens is not required to eliminate fake citations; all that is needed is basic text and metadata matching. Does the title in the citation match the one the DOI resolves to? Does the metadata provided in the citation (authors, journal, volume, issue, date) match? Is the journal cited real and indexed in relevant databases? This is not hard stuff, and the technologies involved have been available for decades. (If you are a publisher and do not have this metadata check in place, please let us know and we’d be happy to help).
Topaz et al. note on their marketing page that fake citations cluster among publishers. They claim that “More than a third of all fabricated citations come from just two publishers.” Frustratingly, they do not say which two publishers. In a statement to Retraction Watch, lead author Maxim Topaz said a “raw comparison of publisher-level rates would be misleading without adjusting for the volume and type of papers each publisher indexes in PubMed. What I can say is that the concentration is disproportionately among large open access journals and publishers, which is consistent with what others have observed about where papermill activity and less rigorous peer review tend to cluster.”
This statement of concern about potentially misleading readers is unconvincing. Surely someone with the sophistication and resources required to develop such a polished marketing website for a single paper is capable of making a figure that adjusts for overall publisher output. The statement underscores the fact that the data reported by this study are not publicly available. Other than a handful of examples provided, we have no way to validate the authors’ reported findings or learn more about the patterns involved. It is hard to know exactly the extent of the actions required to remedy this problem without knowing more about the details.
Briefly Noted
BioOne and Johns Hopkins University Press are joining forces in a merger. BioOne Complete and its eBook collections will join Hopkins Press alongside Project MUSE, bringing together two respected nonprofit organizations with complementary strengths across the sciences, humanities, and social sciences. At a moment when consolidation in publishing is usually associated with large commercial players, this merger stands out. Writing in The Scholarly Kitchen and a joint video announcement, the organizations described the partnership as an effort to strengthen nonprofit scholarly infrastructure while preserving continuity for publishers and libraries.
Elsevier has joined a class-action lawsuit against Meta — alongside Hachette, Macmillan, McGraw-Hill, Cengage, and author Scott Turow — alleging that Meta used copyrighted academic papers and literary works scraped from pirate sites including Sci-Hub and LibGen to train its Llama model. The complaint alleges Mark Zuckerberg personally authorized the piracy strategy after being advised that licensing content would undermine a fair use defense. Notably, Elsevier also alleges that Llama has generated hallucination-riddled summaries of scholarly articles — a harm that goes beyond economics, potentially damaging the professional credibility of individual researchers and authors. The case builds on a legal theory flagged in last year’s Kadrey v. Meta ruling — that AI models trained on copyrighted works can flood markets with competing content, constituting indirect market harm even if the model can’t reproduce text verbatim. The outcome could set a significant precedent for how AI companies source the content underpinning their models.
Wiley has appointed Jessica Kowalski as Executive Vice President and General Manager, Research, succeeding Jay Flynn, who is leaving the company. Kowalski comes from Microsoft, with prior senior roles at AWS and RELX (Elsevier and LexisNexis). The hire signals that Wiley is moving deliberately to put AI and data leadership at the center of its strategy. Elsevier has been repositioning itself as a data company for years. Wolters Kluwer has focused on data and software that builds on its content assets and subject expertise. While this positioning has not helped the equity price of either publisher recently (as we discussed last month in The Brief), that may say more about the fickleness of markets during times of technology upheaval than the strategy of either company. Judging from the earnings calls of the major publishers, AI is moving from a side hustle to a growth story (whether the story gains traction with investors remains to be seen). Nonetheless, one must play the cards in one’s hand, and Wiley, despite some preliminary moves toward AI services, is still (mostly) a publisher (and that is a fact that investors have appreciated recently).
Muck Rack’s Generative Pulse report — based on analysis of 25 million links from AI responses across ChatGPT, Claude, and Gemini — offers a picture of how AI models actually cite sources. 99% of citations come from non-paid sources, with journalism (27%), corporate blogs (24%), and aggregators like Wikipedia (17%) dominating. Academic and research content accounts for just 4% overall — but that figure is averaged across all query types, most of which are consumer queries with no research intent. When you look at where models actually go for authoritative answers, the picture shifts. PubMed Central is Claude’s single most cited domain across the entire study, ScienceDirect appears in Claude’s top five, and NIH rounds out Gemini’s top five. ChatGPT has no authoritative scientific sources in its top five. For marketing and communications teams, the study confirms that earned media and thought leadership content play an important role in driving AI visibility.
Adobe is acquiring Semrush — a leading SEO platform for marketers and digital professionals — for $1.9 billion, adding to its already extensive marketing technology portfolio. The deal reflects a broader shift in how brands think about discoverability as platforms like ChatGPT, Claude, and Gemini increasingly shape how users find information. While framed as an SEO acquisition, the larger play may be Adobe strengthening its position in the race to become the central platform for AI-driven content performance, visibility, and marketing intelligence.
Wondering how much of your traffic comes from AI assistants, which ones, and how it compares to organic search and other referral channels? Tools like Semrush and Profound have had this capability for a while. Google Analytics has now caught up, adding a dedicated AI Assistant channel that tracks visits from ChatGPT, Gemini, Claude, and others automatically — no custom filters required.
The journal Organizational Science (published by INFORMS) has received a surge of papers roughly double the size of the surge they saw during the COVID pandemic. Based on scoring from an AI detection tool, the editors believe nearly all of this growth is due to AI. “Submissions judged to be human-only actually declined.” The problem is not limited to Organizational Science. To combat “AI slop” on arXiv, the preprint platform has instituted a substantive penalty, according to reporting in Research Information. Authors who submit papers with “unchecked” AI (which might include “inappropriate language, plagiarised content, biased content, errors, mistakes, incorrect references, or misleading content,” among other indicators) will face a 1-year submission ban.
Silverchair has released its 2026 Future of Peer Review report based on “eight years of ScholarOne Manuscripts submission and review data, a literature review, expert interviews, and a survey of thousands of reviewers and authors.” Nearly a quarter of all published manuscripts flow through ScholarOne Manuscripts, providing Silverchair with a large dataset for analysis. What they found is what most journal editors live daily: declining acceptance rates for reviewers (meaning more reviewers declining review invitations). “Reviewer acceptance rates (the frequency with which reviewers accept an invitation to review) fell every year from 2018 to 2024 (43% to 22.3%). Our study shows that it takes 1.5x longer for reviewers to decline invitations than it does to accept them.” This is one of the many insights from this excellent and timely report.
We reported on the explosive growth of MCPs in the March 2025 issue of The Brief. The growth continues. Wiley has added Scholar Gateway to the ChatGPT App Directory, joining existing integrations with Claude, Mistral, and other AI platforms, grounding AI responses in peer-reviewed literature with verified citations and DOI links. The same week, EBSCO announced a partnership bringing EBSCOhost’s peer-reviewed databases into Perplexity’s Premium Sources focusing on the BYOL (bring your own license) business model. Google launched Deep Research Max, an autonomous research agent that connects to proprietary data sources via MCP, positioning those sources as a trusted source layer AI agents draw on at the moment of research.
Scopus AI has jettisoned its well-known brand, changing its name to the immediately forgettable “AI Discovery.” We challenge anyone to recall this name in a week.
***
If a submission contains incontrovertible evidence that the authors did not check the results of LLM generation, this means we can’t trust anything in the paper. —Thomas Dietterich, chair of the Computer Science Section of arXiv