AI in Libraries: The Question Isn't Whether. It's How.

I've seen librarians comment on EdTech posts: "Stop trying to replace my job with AI."

I understand that reaction completely. The EdTech industry has a long history of promising transformation and delivering software that creates more work than it saves. The "AI will do it for you" framing is everywhere right now, and it's understandable that librarians — professionals whose value lies precisely in their judgment, relationships, and expertise — would be wary of it.

But I think the framing is wrong. And the research is starting to show why.

What the Research Actually Says

Two recent studies caught my attention, both from the library science field.

Moukhliss & McCowan (2024) conducted a case study on pedagogically redesigned library guides — the resource pages librarians create to help students navigate research. When guides were redesigned using a structured assessment rubric, students navigated them significantly faster: mean task completion time dropped from 39 seconds to 22.2 seconds. They also rated the redesigned guides substantially higher for personal use and likelihood to recommend.

The finding that matters isn't the time saving. It's what drove it: thoughtful design, informed by an understanding of how students actually use resources. The tool supported the librarian's professional judgment. It didn't replace it.

Sheffield & Butler (2026) took this further, demonstrating that AI tools can now handle the assessment work that makes this kind of redesign possible at scale — work that was previously too time-intensive to be sustainable with reduced budgets and staffing. Their finding: AI accuracy compared favourably to human assessment for this task, meaning librarians could redirect their expertise toward the work only they can do.

What I Built — and Why

I'm a Year 3 primary teacher, not a librarian. But I spent a lot of time in our school library watching a problem that felt solvable.

A few years ago I ordered hundreds of carefully chosen books — award winners, highly rated titles across every genre, books that reading specialists consistently flagged as the ones children couldn't put down. Then I watched most of them sit untouched on the shelf. Not because the books were wrong. Because there was no systematic way to connect each child to the titles most likely to excite them.

The librarian's job — hand-matching students to books based on their reading level, interests, favourite authors, mood, and what they'd recently finished — is genuinely skilled work. The problem is scale. With thirty students and a full curriculum, a teacher or librarian can do this brilliantly for a handful of children. The rest get generic suggestions, reading lists, or nothing at all.

That's the gap LibraryAid was built to close. Not to replace the librarian's expertise — but to make it available to every student simultaneously.

The system analyses over 30 factors per student: reading level, interests, favourite authors, curriculum topics, series progression, and literary quality. It generates personalised recommendations from the books already in your catalogue. The librarian's collection, their curation choices, their knowledge of their students — all of that remains central. LibraryAid just makes it accessible at scale.

The Distinction That Matters

There's a meaningful difference between AI that bypasses professional judgment and AI that scales it.

A librarian who knows a child well can make a recommendation that lands perfectly — the right book at the right moment, chosen with understanding of that child's emotional state, reading history, and what they need right now. No algorithm replaces that.

What an algorithm can do is ensure that every student gets a personalised starting point — a list of books matched to their actual interests and reading level — so that the librarian's time is freed for the conversations, the relationships, and the moments of insight that only humans can provide.

The research from Moukhliss & McCowan and Sheffield & Butler points in the same direction. AI that supports thoughtful design improves student outcomes. AI that supports librarian expertise extends its reach. The question for any EdTech tool isn't whether it uses AI. It's whether the AI is in service of professional judgment or in place of it.

The Honest Version

I'll be direct about what LibraryAid does and doesn't do.

It doesn't know that a particular student needs a book about resilience right now. It doesn't notice that a child who was engaged last week seems withdrawn today. It doesn't have the conversation that turns a reluctant reader into a confident one.

What it does is handle the matching work systematically — so that librarians and teachers can spend their time on everything else. In practice, in my own classroom, this meant less time trying to remember which student had read what, and more time having genuine conversations about what they thought of their books.

The question isn't whether AI belongs in libraries. It's whether it's being used thoughtfully. Used thoughtfully, it gives librarians' expertise somewhere better to land.