The campus's AI-literacy backbone
Libraries sit where information literacy, privacy, licensing, and access meet, the exact intersection AI disrupts. That makes the library uniquely placed to lead AI literacy, evaluate the tools flowing in through its own databases, and defend reader privacy and intellectual freedom.
A practitioner orientation to adapt to your library, not a mandate. Scope, staffing, and consortial agreements vary; use this to frame the library's role in the institution's AI work.
The library's distinct AI mandate
Six places where AI runs straight through library expertise, and where no other unit is better positioned to lead.
AI & information literacy
Evaluating sources, spotting fabrication, and citing responsibly are library competencies, now the campus's frontline defense against confident AI nonsense. No unit is better placed to teach it.
Research & discovery tools
AI search, summarization, and literature-synthesis tools are entering the research workflow through the library's own databases. You're the one who can tell a genuine aid from a hallucination engine.
Collections & licensing
Database vendors are bolting AI onto products you already license, often with new terms about training data and patron queries. Reading those terms is squarely library work.
Privacy & intellectual freedom
Reader privacy is a core library value. AI tools that log queries and build profiles collide with it directly, making the library the campus's conscience on surveillance.
Scholarly communication & integrity
Disclosure norms, authorship questions, and citation integrity for AI-assisted work run through scholarly communication, a library specialty other units lean on.
Reference & the chatbot question
When the institution wants an AI assistant to “answer questions,” the reference desk has decades of hard-won knowledge about what a good, sourced, honest answer actually requires.
The campus AI-literacy hub
Information literacy is the library's oldest job and AI's newest problem. Four ways to lead it for the whole institution.
Reframe evaluation for generative AI
The CRAAP-style questions still apply, but AI adds new ones: was this fabricated, what was it trained on, what's the tool's incentive? Update the instruction, don't discard it.
Teach verification, not prohibition
Students will use these tools. The durable skill is checking a claim against a real source, exactly what libraries have always taught. Meet them where the work is.
Reach faculty, not just students
Faculty need AI literacy as much as students, for their teaching, assessment design, and research. The library is the neutral, credible home for that development.
Model disclosure
When the library uses AI in its own services, say so, plainly. Transparent practice is the most persuasive lesson you can teach.
Licensing & the vendor question
AI is arriving through the databases you already license. Five questions to put to every vendor before an AI feature goes live.
New AI features may feed patron searches or licensed content into a vendor's model. Get an explicit answer, and a contractual limit, before you enable them.
AI features often route queries to third-party models. That can turn a private reader search into shared data, breaking a core library commitment. Know the data path.
Some vendors switch AI on by default at renewal. Decide whether it belongs in your environment, don't inherit it silently.
An AI-“enhanced” database that invents citations undermines the collection's credibility. Push liability and accuracy commitments back onto the vendor.
AI layers can create new lock-in. Confirm you keep access to the underlying content, and your patrons' privacy, if you drop the AI feature.
Reader privacy vs. AI logging
Libraries have defended reader privacy for a century. AI tooling logs, profiles, and personalizes by default, the opposite instinct. Hold the line where it matters.
- Query logs that build a searchable profile of a reader
- Personalization that quietly narrows what a patron sees
- Chatbot transcripts retained and mined without notice
- Third-party models receiving patron searches verbatim
- Minimize what's logged; retain it for as short as possible
- Keep patron-identifiable AI data out of vendor training
- Tell patrons plainly when AI is in the loop, and let them opt out
- Bring the library's privacy stance to the governance table
The library's privacy standard is often stricter than the institution's default, that's a feature. Carry it into AI governance rather than letting it erode.