This essay was researched and edited by me and written by Gemini. It does not necessarily reflect my own opinions. I researched it in order to try and learn more about AI and open research (as part of my job as a librarian) and because I thought it might be useful for other people to read and reflect on.
The rise of artificial intelligence (AI) presents a transformative, yet complex, landscape for open research and academic publishing. While offering significant opportunities to enhance efficiency, accessibility, and discoverability, AI also introduces a series of challenges concerning research integrity, economic models, and ethical considerations. The intricate relationship between AI and open research is a "double-edged sword" (Lo, 2025), necessitating careful navigation to harness AI's benefits while mitigating its risks.
Open research, with its emphasis on transparency, collaboration, and the free dissemination of knowledge, provides fertile ground for the development and application of AI. Elena Giglia, a prominent voice in the open science community, highlights that "Data and text must be Open with appropriate licenses to ensure a critical mass of training material for AI" (Giglia and Tammaro, 2025). Large quantities of openly available data are crucial for training robust AI models, minimising errors and improving predictive capabilities (Giglia and Tammaro, 2025). Furthermore, the principles of open research, particularly the FAIR principles (Findable, Accessible, Interoperable, and Reusable), are essential for developing trustworthy AI. When AI itself adheres to FAIR and open principles, it moves away from opaque "black boxes" towards a more equitable and transparent research environment (Giglia and Tammaro, 2025). When the workings of AI models are made openly available, this helps to ensure reproducibility and scientific rigour, fostering a more collaborative scientific community (Rockembach, 2024).
AI's potential to bolster open access publishing is multifaceted. It can accelerate publication timelines, reduce editorial costs, and enhance the quality and consistency of published research (Nagarajan, 2024). AI tools can also significantly improve the discoverability and accessibility of research by overcoming language barriers and bridging skill gaps in interdisciplinary research (McKenna, 2023). By automating tedious tasks and streamlining data-sharing requirements, AI can lower barriers to openness, making it simpler and more appealing for researchers to deposit data, code, and protocols in easily discoverable and reusable formats (Scaplehorn and Schönenberger, 2025). AI-powered search and recommendation systems, advanced data analyses, comprehensive literature reviews, and personalised research recommendations all contribute to more efficient and impactful dissemination of knowledge (Lo, 2025).
Navigating Challenges and Ethical Concerns
Despite its promise, the integration of AI into open research is not without its perils. A significant concern revolves around the aforementioned "black box" nature of some AI models, which can hinder transparency and accountability (Nagarajan, 2024). This lack of transparency, coupled with the potential for the output of biased information stemming from biased training data, poses a serious threat to the integrity of research outcomes. Human oversight remains crucial in mitigating these risks (Nagarajan, 2024).
The intersection of AI and open access also brings to light complex issues related to copyright, authorship, and plagiarism. Open access literature, by its very nature of unrestricted content availability, has become a "preferred hunting ground for AI chatbots" (La Tunisie Médicale, 2023). While open access licenses generally permit machine learning, the question arises whether "free to read means free to train" (Decker, 2025). Academics often opt for OA publishing to maximize human readership and reuse, but arguably not to provide free training data for AI companies (Decker, 2025). The ability of AI to generate content, not just index it, complicates traditional notions of transformative use and raises concerns about "citation laundering." This is where original sources are hidden or misattributed, thereby lowering attribution standards and potentially decontextualizing research insights (Decker, 2025). The central issue, as Decker (2025) argues, is that commercial AI companies extract significant economic value from OA content without necessarily returning any value to the academic ecosystem that produced it, while simultaneously disrupting academic incentive structures.
Economically, generative AI presents a complex landscape. While it promises to streamline workflows and reduce costs, potentially making OA publishing more financially sustainable (Lo, 2025), it also opens new avenues for monetisation. AI-driven analytics and personalised content recommendations could become premium services, potentially creating a tiered system of access that undermines the inclusive ethos of OA (Lo, 2025). The significant upfront investment required for advanced AI systems may favour larger publishers, exacerbating existing inequalities in global knowledge production and dissemination (Lo, 2025). This could lead to a "technological divide" where well-resourced publishers further distance themselves from smaller or less affluent ones (Lo, 2025).
Towards Responsible Integration
Addressing these challenges requires a concerted effort from all stakeholders in the academic ecosystem. There is a pressing need for "artificial intelligence literacy," especially for researchers, to ensure the safe and ethical use of AI tools (Rockembach, 2024). The development and training of specialised professionals, such as digital ethicists, will be fundamental in these new technological environments (Rockembach, 2024).
To mitigate the issues of attribution and economic value extraction, new licensing frameworks could be considered. For example, OA licenses could mandate that AI tools trained on OA papers include citation capabilities in exchange for the free use of high-quality material, ensuring appropriate recognition for creators (Decker, 2025). Furthermore, the scholarly communications community, including librarians and publishers, should actively advocate for the inclusion of scholarly content in AI training datasets, ensuring a source of accurate information (Montague-Hellen, 2024). This may even necessitate a shift away from traditional PDF formats towards more machine-readable content to improve AI's ability to crawl academic material (Montague-Hellen, 2024).
Libraries, as stewards of access and advocates for equitable knowledge dissemination, have a crucial role in shaping the future of scholarly publishing in an AI-driven world (Lo, 2025). By critically assessing the impacts of new publishing models and actively engaging in policy development, libraries can help ensure that the benefits of AI are widely distributed and aligned with the principles of open access (Lo, 2025). Ultimately, achieving a harmonious integration of AI and open research will require continuous dialogue, proactive policy development, and a commitment to fostering a more transparent, equitable, and trustworthy research landscape.
Bibliography
Decker, S. (2025) ‘Guest Post - The Open Access – AI Conundrum: Does Free to Read Mean Free to Train?’, The Scholarly Kitchen, 15 April. Available at: https://scholarlykitchen.sspnet.org/2025/04/15/guest-post-the-open-access-ai-conundrum-does-free-to-read-mean-free-to-train/ (Accessed: 1 July 2025).
Giglia, E. and Tammaro, A.M. (2025) ‘Bridging artificial intelligence (AI) and open science: an interview with Elena Giglia’, Digital Library Perspectives, Vol. 41 No. 2, pp. 389-393. https://doi.org/10.1108/DLP-05-2025-206 (Accessed: 7 July 2025)
Gundersen, O.E. and Coakley, K. (2023) ‘Open research in artificial intelligence and the search for common ground in reproducibility: A commentary on “(Why) Are Open Research Practices the Future for the Study of Language Learning?”’, Language Learning, 73: pp. 407-413. https://doi.org/10.1111/lang.12582 (Accessed: 7 July 2025)
La Tunisie Médicale (2023) ‘Artificial intelligence and promoting open access in academic publishing’, La Tunisie Médicale, 101(6), p. 535. Available at: https://latunisiemedicale.com/index.php/tunismed/article/view/4590 (Accessed: 23 June 2025).
Lo, L.S. (2025) ‘Generative AI and open access publishing: A new economic paradigm’, Library Trends, 73(3), pp. 160–176. Available at: https://muse.jhu.edu/pub/1/article/961190 (Accessed: 1 July 2025).
McKenna, J. (2023) ‘AI is changing open access science’, MDPI Blog, 5 September. Available at: https://blog.mdpi.com/2023/09/05/ai-open-access/ (Accessed: 23 June 2025).
Montague-Hellen, B. (2024) ‘Empowering knowledge through AI: open scholarship proactively supporting well trained generative AI’, Insights, 37(1). Available at: https://doi.org/10.1629/uksg.649. (Accessed: 7 July 2025).
Nagarajan, P. (2024) ‘AI and the future of open access publishing: Revolutionizing academic research and dissemination’, Integra, 6 June. Available at: https://integranxt.com/blog/ai-and-the-future-of-open-access-publishing-revolutionizing-academic-research-and-dissemination/ (Accessed: 23 June 2025).
Rockembach, M. (2024) ‘Ciência Aberta e inteligência artificial: desafios éticos e transparência em modelos generativos’, Ciência da Informação, 53(3). Available at: https://doi.org/10.18225/ci.inf.v53i3.7227. (Accessed: 7 July 2025) [Translated by Gemini]
Scaplehorn, N. and Schönenberger, H. (2025) ‘Can AI make research more open?’, Impact of Social Sciences, 4 April. Available at: https://blogs.lse.ac.uk/impactofsocialsciences/2025/04/04/can-ai-make-research-more-open/ (Accessed: 23 June 2025).
This is great and appreciated the foreword note. Read something earlier about academic papers having prompts embedded which has agents then posting positive reviews about those papers?!