Here’s one example. A study published in August 2025 by the National Academy of Sciences, titled “The Entities Enabling Scientific Fraud at Scale Are Large, Resilient, and Growing Rapidly” (pnas.org/doi/10.1073/pnas.2420092122), identified a number of classes of systemic fraud, such as paper mills, paper brokerage, and predatory journals. The study was picked up by WIRED (wired.com/story/black-market-for-fraudulent-science-growing-faster-than-legitimate-research) and by The Los Angeles Times (latimes.com/science/story/2025-08-06/mass-fraudulent-science-is-polluting-literature), among other media outlets.
Long a concern in the library scholarly communication and academic researcher communities they serve, the jump into a more general consciousness is alarming. It threatens trust in science and engenders fear in scientific expertise. When reputable sources such as WIRED and The Los Angeles Times declare that publication of spurious content is on the rise, and growth in this area is outpacing the rate of growth in the publication of authentic studies, they reach a very large audience outside the LIS community. The implication is that the rising rate of fraud will lead to an eventual flooding of the research landscape, drowning out legitimate scientific findings.
TWO CATEGORIES OF FRAUD
Research fraud falls into two broad categories: the insider job committed by researchers in their own work and the external threat emanating from those paper mills and predatory journals cited in the National Academy of Sciences report.
Researchers may not always have the integrity we expect. Fraud perpetrated by researchers essentially constitutes acts carried out within the research process to manipulate the outcomes or the impact of the work. These could include misrepresenting or changing their own collected data, deliberate plagiarism, undisclosed conflicts of interest, or submitting the same paper simultaneously to two journals. It could also mean finding ways to bloat the perceived impact of one’s work: Extreme self-citation comes to mind. There are scores of articles from all corners of the scientific community expressing concern and a need to mitigate this behavior.
More concerning is that organized fraud is on the rise. This systemic fraud lies largely outside the control of individual researchers, although when it comes to predatory journals, scholarly communications librarians have been effective in steering researchers away from publishing in them. Fraudulent articles promoting misinformation, however, now show up in public policy spheres, influencing policy decisions, particularly in the healthcare arena. And it’s not restricted to paper mills and predatory journals.
OA: HELP OR HINDRANCE?
Has the move toward open access publishing contributed to the rise in fake research? Danny Kingsley, OA thought leader, wrote in The Conversation about why she thinks OA fosters research integrity as opposed to elevating the likelihood of fraud (theconversation.com/show-your-working-how-the-open-science-movement-tackles-scientific-misconduct-249020). She points out that making data open, registering clinical trials, and publishing research protocols help to keep scientists accountable and to provide transparency in how the research was conducted. While this is a cogent and laudable position for addressing researcher-perpetrated falsities, it does not address the coordinated or systemic fraud that is on the rise and bolstered through the ease of generating false research with AI.
To that end, the scientific community continues to debate how to address the increase in fraudulent activity in their publications. Journal publishers actively look for ways to shore up their submission and review processes and seek new ways to validate the content they make available. Every stakeholder in the research community has a part to play in mitigating this trend.
In November 2025, the Royal Swedish Academy published a set of principles by which all players in the research publishing ecosystem could foster change to improve fraud detection. Known as the Stockholm Declaration (royalsocietypublishing.org/doi/10.1098/rsos.251805), it includes four precepts: “Academia resumes control of publishing,” “Incentive systems to merit quality, not quantity,” “Independent fraud detection and prevention,” and “Legislation and policies to protect science quality and integrity.”
With regard to OA, the Stockholm Declaration is strongly supportive: “we recommend a transition from subscriptions and transformative agreements to innovative digital open access models (free submission, free reading).” In other words, the issue is not OA—it is the profit motive in academic publishing that is the true concern to address.
ENTER GENERATIVE AI
Along with OA, critics point to generative AI (gen AI) as facilitating the research fraud phenomenon. Gen AI can be problematic for three reasons. First, using open gen AI tools for information retrieval yields synthesized results that are potentially fraught not only with errors, but also with outright fabrications, aka hallucinations. The synthesis may or may not link back to the source, but how many searchers simply use that generated content without checking the provided links? Second, large language models (LLMs) do not distinguish between spurious and legitimate content found on the web. This means that when content is scraped into an LLM, fraudulent research is treated exactly the same as research from an authoritative source. Even if the searcher checks the provided links, there is zero guarantee that the linked content is valid or genuine. Finally, gen AI can be used to quickly create content that looks and reads as a true research paper.
Thus, an AI tool can generate fake papers that can be scraped by any LLM and synthesized into a result that, in and of itself, misinterprets what it retrieves from multiple results. AI can facilitate the threading of fraudulent information throughout the search process.
Back in 2010, Joeran Beel and Bela Gipp wrote about how easy it was to game Google Scholar to create “search engine spam” (doi.org/10.3998/3336451.0013.305). This situation is similar to our current experiences with gen AI in that the Google crawler could not distinguish between legitimate research and spurious PDFs. The authors proved this by creating a PDF with journal-like metadata on the back end that displayed advertisements along with a bibliography. The contrived content was meant to demonstrate that nefarious authors could use spam PDFs with fake citations to their own works, thus amping up the citation count and perceived impact. Both in this case and in the case of open gen AI tools, the tool itself cannot distinguish the quality or the authenticity of the source crawled or scraped.
Since then, efforts to mitigate Google spam have included iterations to Google Scholar’s algorithm, relevancy ranking, and result display continue to improve the likelihood that the results pushed to the top of the relevancy ranking are credible in addition to being relevant (google.com/intl/en_us/search/howsearchworks/information-quality). The Stockholm Declaration proposes essentially this kind of iteration for AI tools, stating: “AI use in scientific publishing and data tracking needs guidelines, but because the AI field is currently developing very fast without proper regulation, opportunities and risks will need continuous monitoring and adjustment.”
INFO LIT ALL OVER AGAIN
In July 2025, Forbes published an article by Ted Ladd touting the concept of skeptical intelligence as a tool to help people sort through the validity of information, particularly what is garnered from AI-generated sources. “It is a disciplined approach to questioning that combines curiosity, critical thinking, epistemic humility (knowing what you don’t know), and a toolkit for evaluating evidence” (forbes.com/sites/tedladd/2025/07/19/skeptical-
intelligence-is-crucial-in-the-age-of-ai).
This “disciplined approach” and “toolkit” that Ladd deems necessary are already well-developed in the LIS sphere. To para–phrase Yogi Berra, it is information literacy all over again. The standard definition of information literacy used in library and info pro circles comes from an ALA report in 1989: “To be information literate, a person must be able to recognize when information is needed and have the ability to locate, evaluate, and use effectively the needed information” (ala.org/acrl/publications/whitepapers/presidential).
Interestingly, the report discusses the need for information literacy in business, citing a former editor of Fortune and drawing from many examples where information literacy is key in a business context. This means that, for 36 years, library and info pros have been sounding the clarion call that information literacy is a critical business skill, yet somehow the term has not crossed domains. Ladd’s article is only the latest voice outside the library echo chamber to state a need for the type of competencies the library has already articulated in the ACRL Information Literacy Framework for Higher Education (ala.org/acrl/standards/ilframework) and other sources.
It is tempting to complain at length about the societal lack of acknowledgment of the library community’s quest for raising awareness, promoting critical thinking, and creating a well-informed society. One answer to this lack of traction is to use the language of business to speak to the business community. Jargon such as “information literacy” is an impediment to promoting the good work of librarians and info pros everywhere. It hides the toolkits, frameworks, and resources we make available to help combat the challenges.
At the 2025 Internet Librarian Connect conference, keynoter David Lankes talked extensively about the need to insert ourselves where information is consumed. Lankes drew a comparison to the Wild West days of the internet in the late 1990s as a space where librarians were providing guidance and support in navigating around the junk to locate reputable information. Discerning quality information from AI flotsam and jetsam may be more challenging due to the more rapid evolution of gen AI capabilities and the lack of transparency in what exactly those LLMs are scraping.
Clearly, there is much to be done to stem the tide of all kinds of fake content found on the internet, from fake news and deepfakes to AI-generated garbage (AI slop) of all types. At the same time, the process by which original research is created and disseminated needs to be safeguarded even more rigorously to maintain its status and credibility as a pinnacle of reputable information. The point of scientific inquiry is to provide a standard, rational way of examining problems and questions confronting humankind. It is imperative that this essential value of conducting science and scientific research is maintained. Implementing the values expressed in the Stockholm Declaration and other recommendations for changing the research ecosystem will take time, and this needs to be a priority for the scientific and publishing communities.
Concurrently, the LIS community can bolster these efforts. In addition to the information literacy framework, a host of materials and resources is available from libraries to help create a citizenry of savvy information creators and consumers. The LIS community has a trove of tools and resources to build awareness and help people understand, evaluate, and use information.
The challenges posed by AI provide yet another opportunity for librarians and info pros to demonstrate value. We can share resources to help information users identify the validity of research papers specifically. Retraction Watch’s Hijacked Journal Checker, evolving publication screeners, and sources such as the Directory of Open Access Journals can be used to help vet sources and are openly available to the public. Traditional library resources such as curated proprietary databases from reputable vendors and tried-and-true tools such as Ulrichsweb help identify the legitimacy of journals. Upstart Scite has an AI-based approach to identifying retracted research within its platform (scite.ai/blog/2022-10-19_automated-notice-detection).
Clarivate and EBSCO provide contrasting approaches to identifying and rooting out fraudulent research, although both companies focus on publishers and journal-level fraud. Clarivate’s focus is largely two-pronged. First, it is actively identifying and delisting journals, especially from Web of Science, that do not meet best practices criteria for transparency and publishing. Journal inclusion and de-listing criteria are increasingly becoming transparent. In 2025, Journal Citation Reports excluded counts of retracted articles from the Journal Impact Factor (JIF) formula. Second, Clarivate is partnering with Retraction Watch to elevate the existence of retractions in published research by leveraging multiple screening approaches, including AI, to identify journals that do not meet their best practices criteria.
EBSCO seems to be focusing mainly on weeding out predatory and fraudulent journals. EBSCO has an evaluative, data-driven model considering a variety of factors, primarily relying on criteria from Cabells’ Predatory Reports to identify predatory publishers. It also considers the evaluative strategies of other sources, such as indexing in Scopus and Web of Science, and scholarly impact metrics. It is working on leveraging AI in tandem with Cabells content to improve the efficiency of its evaluative process. While these approaches are laudable, there is an increasing need for granular screening beyond retractions, as we begin to see AI-generated articles accepted into otherwise legitimate journal titles, possibly flooding the publishers with phony content.
THE TIME IS NOW
While the LIS community is prepared to support society in its effort to stem the tide of research fraud, it faces challenges. LIS practitioners have been technology-forward for decades, at the forefront of anti-censorship endeavors and stalwart advocates for responsible information creation and consumption. Yet it seems that in media, business, government, and even some academic circles, the solutions offered by our profession in helping with society’s thornier information conundrums are not recommended or provided. These sectors just don’t seem to think of libraries when addressing these challenges.
One way to step out of the LIS echo chamber is to drop the jargon and find the lingua franca of non-library users to communicate value. The term “information literacy” is eye-glazing at best and means nothing to most people outside of education. Learn and use the language used in media, business, government, and society at large.
Finding new and innovative ways to insert LIS values, tools, and resources into the spaces and places where people are seeking or creating information is also an important approach. Think about strategies for even higher visibility at events or locations outside of libraries and how to leverage technology to put library resources in online venues, maybe creating evaluative browser and AI add-on tools to help users navigate content at the point of discovery. Brainstorm new ways to interact with key stakeholders in business, government, technology, and online communities. When inroads are made in those sectors, librarians and info pros need to lean on and leverage personal networks to further share our tools, resources, and values.
Yes, library and info pros have been doing these things for years. To borrow from a popular quote of motivational speaker Tony Robbins, the situation now calls for “massive, determined action.” This time, the pace and volume of change mean leveling up for stronger, higher-profile partnerships; increasing our technological footprint; and using language that resonates with the sectors that can leverage our skills and amplify our signal.
