When used skillfully, AI prompting can democratize scholarly knowledge. A well-designed prompt can coax explanations, comparisons, or syntheses of research in language accessible not only to specialists but also to students, journalists, and the public. Library-integrated AI tools go beyond listing citations: they summarize literature, map trends, and suggest relevant studies.

Real-world examples highlight this potential. In South Korea, librarians built a generative-AI chatbot, tlooto Copilot, drawing on 300 million academic documents; users reported personalized support and more efficient research. At Northwestern University Libraries, a pilot project funded by an IMLS grant is deploying a generative‑AI chat tool to support researchers: users can ask natural-language questions about archival manuscripts, enabling deeper exploration of complex collections.

Still, AI prompting alone cannot overcome a key barrier: much scholarly content remains behind paywalls. High-quality research often sits in subscription-only journals, limiting AI’s capacity to generate accurate or deep responses. Studies show that AI-generated summaries can be prone to omission, misinterpretation, and overgeneralization, especially when evaluated against human or full-text benchmarks. When summarizing only abstracts rather than full-text articles, AI models are even more likely to miss critical methodological details and nuanced context, highlighting the limitations of relying solely on condensed summaries for accurate scholarly understanding. This is compounded by the fact that paywalled content is largely inaccessible to AI, meaning that important research remains underrepresented in AI outputs. Partial openness, therefore, creates knowledge amplification gaps: ideas in open-access materials are repeatedly summarized, translated, and circulated through AI prompting, while paywalled ideas are “under-prompted,” less visible, and risk fading from both scholarly and public discourse.

Academic libraries are uniquely positioned to address these challenges. Many operate institutional repositories (IRs) where faculty deposit preprints, postprints, and datasets. These can be optimized for AI use with enriched metadata, semantic tagging, persistent identifiers, and open licenses. Library-run publishing initiatives like D-Scribe demonstrate sustainable models for wide OA dissemination.

Libraries also negotiate transformative agreements that shift spending from subscriptions toward OA publishing. As of 2023, the Max Planck Digital Library (MPDL) has successfully established 24 active transformative agreements and 10 direct‑billing arrangements with OA publishers, enabling authors to publish openly in more than 12,000 journals at no cost. Its Microbiology Society agreement enables researchers at 86 institutes to publish OA across six journals under a “publish and read” model. Evidence shows these agreements measurably increase OA output. Libraries are not just service providers; they are policy actors shaping the future of scholarly communication.

Beyond infrastructure and policy, libraries play a critical role in building AI literacy. They teach students, faculty, and staff how to craft prompts for reliable and ethical AI outputs, critically evaluate AI summaries, and engage responsibly with scholarship. Many libraries are becoming AI competency hubs, offering workshops, online modules, and consultations. Partnerships with AI developers strengthen this work. For instance, Consensus synthesizes findings from a corpus of over 220 million research papers and always cites the original sources, enabling users to trace insights back to the academic literature. When full-text is openly available, Consensus provides direct access to PDF downloads; however, for other articles, it only links to the publisher’s page, reflecting situations where full-text access is not available.

Libraries must also build ethical frameworks guiding AI adoption, addressing transparency, privacy, equity, accessibility, and bias. With strong governance, libraries can guide generative AI to advance both scholarly rigor and equitable access.

By combining well-crafted prompting, open infrastructure, policy advocacy, AI-literacy training, and responsible partnerships, AI and libraries together can expand access meaningfully. Thoughtful prompts make scholarship understandable; libraries supply expertise, systems, and policy leverage that sustain access. Transformative agreements enlarge the share of openly published work, and enriched repositories ensure AI has high-quality, legally reusable material to reason over.

Imagine a world where anyone — a student in Nairobi, a journalist in Bogotá, or a policymaker in New Delhi — can ask an AI about the latest findings in medicine, climate science, or education and receive clear, context-rich insights linked to openly accessible research. Libraries could serve as launchpads for this future, equipping communities not only with knowledge but with the skills to use it responsibly. AI prompting, aligned with the vision and action of academic libraries, has the potential to make open access not only broader but more meaningfully usable — a future worth striving for together.