Envisioning a More Standardized Approach to AI in Scholarly Publishing

Letter from the Executive Director, May 2025

For the past two years, the impact of artificial intelligence (AI) systems on scholarly publishing has been the focus of many conversations. Whether the topic is the use of unlicensed journal content in large language model outputs, how one successfully negotiates AI licensing deals, or the impact of AI bots crawling your site and slowing down services for actual patrons, the disruptions are real. As I discussed last December, for the past year the board of NISO has been having internal conversations regarding the need for community guidance around several issues related to AI systems. NISO will soon engage the broader community on the subject as well.

In May, NISO is hosting two workshops on AI in scholarly publishing, focused on the real-world problems—and prioritizing possible solutions for addressing them—that artificial intelligence is creating for the scholarly communications ecosystem. These conversations are specifically designed for our community of publishers, content providers, and technologists, who are at the confusing intersection of AI and academic publishing. The invitation-only meeting will involve senior leadership at large organizations among NISO’s membership as well as a few other invited guests. 

The purpose of this meeting is to brainstorm key issues at play, synthesize these problems into achievable solutions, and then prioritize a plan of work for NISO to improve efficiency working with AI systems across the network of publishers. The goal is to define a small set of priority projects that NISO can advance collective action around at a network level, not necessarily at a product level. We will also discuss and prioritize concrete next steps to advance the ideas generated.

In speaking with several executives about these issues in recent months, many ideas have emerged: 

• Attribution. One key question is whether semantics can be leveraged more effectively in AI systems to facilitate attribution and improve credit in LLM outputs. In other words, can we build smarter, more structured systems that “understand” who wrote what, and make sure that credit is recognized and accrues back to the author? This could be through better use of persistent identifiers, enhanced markup, or semantic enrichment, or simply ensuring that this work—already done by the publishers—is preserved. 
• Licensing. To support innovation and protect rights, the publishing community needs systems to effectively and efficiently license copyrighted content to LLM and AI systems developers. That could mean new kinds of machine-readable licenses, API integrations, or even standardized terms that work across platforms. Can we establish consistent terminology that might be used in licensing content for AI? Alternatively, could we develop model licenses that serve as a basis for negotiation with AI vendors?
• AI agent access and usage. Looking forward, we can envision an ecosystem in which AI agents are used to read, summarize, and report back to users on the latest content. This poses a lot of questions: What does a publisher system that provides AI access look like, and how is it different from existing systems? And how will  usage be tracked and reported in such a system?
• Search and discovery. Last fall, the NISO Open Discovery Initiative launched a survey on expectations and concerns about generative AI in web-scale library discovery. How might transparency be built into systems information so that one can understand what is included in an AI tool beyond “everything”?
• Accessibility. There are potentially many ways that AI can be used to improve the accessibility of born-accessible or back-file content. As AI becomes more mature and accurate, publishers could potentially harness this technology to advance these efforts at scale. Deploying these tools could benefit from guidelines or best practices for using AI tools for alt text generation.
• Provisioning of content for AI bots and agents. AI bots are crawling content hosting sites, particularly those hosting open access content, at scale—scraping content, hammering servers, and soaking up bandwidth. This is resulting in slower load times and degraded service for researchers, students, and readers. Can we develop responsible crawling policies, access protocols for bots, or technical defenses to prevent reading from being degraded?

These bullets are not meant to be a roadmap of NISO’s upcoming work, but rather a sample of the issues our community is facing and a starting point for a conversation and brainstorming exercise. I’m certain many other ideas will emerge from our conversations. In the weeks following the workshops, NISO will circulate a report of the event and how things will be moving forward based on the group’s guidance. Obviously, the membership of NISO will have a critical role in approving whatever project proposals are advanced.

As with many NISO initiatives, our goal for these workshops is to foster a space for open dialogue and creative problem-solving as well as propel impactful community guidance work. Some of these ideas might find homes in other organizations, which is certainly helpful, since NISO is unlikely to have the resources to address all of them. That is why prioritization will be critical, with input from the participants and NISO’s board, leadership committees, and members. Hopefully, you all, like me, will be excited to hear the outcomes and see the new work on AI systems advance.

Sincerely,

Todd Carpenter,
Executive Director, NISO