The Best Response to AI Is a Library Card
Why reading is the skill that makes every other skill possible, and why we need it now more than any prompt.
We are spending billions teaching people to use AI. We have not finished teaching them to read.
In this post I will:
Argue that literacy, not AI literacy, is the foundational skill we are neglecting.
Show why the books that build judgement, empathy, and critical thinking have nothing to do with AI.
Introduce The Slow AI Public Library: 44 books from 16 countries, prescribed to you based on how you use AI. Free. No signup.
You cannot be AI literate if you are not literate first
Every AI literacy framework I have seen assumes the person using AI can already read critically, assess evidence, hold ambiguity, and distinguish between a persuasive argument and a true one.
That is an enormous assumption.
These are not innate skills. They are built. They are built slowly, over years, by reading things that are difficult, uncomfortable, and longer than a ChatGPT summary. They are built by encountering people whose lives look nothing like yours and staying with them for 300 pages. They are built by finishing a book and realising you were wrong about something you were certain of at the start.
We have skipped a step. The conversation about AI literacy has leapt straight to prompt engineering and responsible use without asking whether the person at the keyboard has the critical literacy to evaluate what the machine produces. You cannot assess AI output if you have never learned to assess an argument. You cannot spot a hallucinated citation if you have never read the real thing.
Reading is the infrastructure. Everything else is built on top of it.
What reading does that AI cannot replicate
AI is good at producing text. It is poor at producing the kind of thinking that makes text worth reading.
When you read a novel, you practise sustained attention. You hold a character’s contradictions in your mind without resolving them. You sit with ambiguity for hours, not seconds. You encounter a world that was not generated for you, that does not adapt to your preferences, that asks you to meet it on its own terms.
When you read a difficult piece of non-fiction, you build the muscles of evidence assessment. You learn to distinguish between a claim and its support. You notice when an author is hedging, when they are asserting without evidence, when they are asking you to feel rather than think.
These are the skills that AI literacy requires. And they do not come from prompting. They come from reading.
I say this as someone who is guilty of the same neglect. As a teenager I inhaled books. I remember reading Post Office by Charles Bukowski and thinking: this is what reading could be. Raw, direct, unsentimental. It changed what I expected from a sentence.
Over the years, my reading narrowed. More non-fiction. More books about the thing I was working on. Ethan Mollick’s Co-Intelligence. Mustafa Suleyman’s The Coming Wave. Both excellent. Both about AI. Both reinforcing the lens I was already looking through.
The books that mattered most to me last year were different. Nan Shepherd’s The Living Mountain, a reissue of her quiet masterpiece about the Cairngorms, written in the 1940s and ignored for decades. Téa Obreht’s The Morningside, a novel about displacement and belonging in a world that has changed beyond recognition. Both reminded me of something AI never could: the fragility and resilience of our relationship to the natural world and to each other.
Neither book had anything to do with AI. Both made me better at thinking about it.
Critical literacy is not a specialised skill
There is a phrase in education: ‘reading across the curriculum.’ It means that reading is not the English department’s responsibility. It belongs everywhere. Every subject builds on the ability to read, interpret, and question a text.
The same is true for AI. Critical AI literacy is not a technology skill. It is an extension of critical literacy. The person who can read Toni Morrison closely, who can sit with the dissonance of a Claudia Rankine poem, who can follow a David Graeber argument across 500 pages and identify exactly where they disagree, that person is better equipped to evaluate AI output than someone who has taken every prompt engineering course available.
We do not need more AI courses. We need more readers.
And we need the institutions that make reading possible. Which brings me to libraries.
Why libraries are worth fighting for
Libraries are the original open-access technology. Free to use. Open to everyone. No algorithm deciding what you see. No subscription required. No data collected. No engagement metrics shaping what gets surfaced.
They are one of the few remaining public spaces where you can think slowly, borrow freely, and encounter ideas you did not ask for. A library does not optimise your reading. It lets you wander. That wandering is where the unexpected connections happen, the ones no recommendation engine would have made.
Libraries are under threat in ways that should alarm anyone who cares about literacy, equity, or democracy. The UK is losing libraries at a rate of about 40 each year. In the US, book bans are accelerating. The spaces where critical literacy is built are being dismantled while we pour investment into AI literacy programmes that assume the foundational work is already done.
It is not done. It was never done. And it will not be done by a chatbot.
The Slow AI Public Library
I built a tool. It is not an AI tool. It recommends books.
The Slow AI Public Library asks you three questions about how you use AI, what worries you about it, and what you have not made time for lately. Then it prescribes you a book. Not an AI book. A book that builds the skills AI erodes: judgement, empathy, patience, close reading, the ability to hold not-knowing.
44 books. Fiction, poetry, non-fiction. Authors from 16 countries. Every one of them chosen because it builds something that no prompt can replicate.
You get a library card you can download and share. And a link to find the book at your local library, because the best response to AI is not a better prompt. It is a library card.
If you use the library, share your card. Tell someone what you were prescribed and whether it surprised you. Better still, tell someone about the last book that changed how you think, and notice whether it had anything to do with AI.
Why literacy comes before everything
We are living through an era of sophisticated text generation. The machines produce fluent, confident, well-structured prose on any topic in seconds. The only defence against fluent nonsense is the ability to read critically. That ability is not built by technology. It is built by books, by teachers, by libraries, by the slow accumulation of encounters with writing that demands something of you.
Before AI literacy, literacy. Before digital skills, reading. Before the prompt, the page.
Go slow.
Paid membership gets you access to a fully accredited curriculum, monthly live seminars, a community of over 220 policymakers, educators, and creators, and full access to all Slow AI posts.
Excellent post. You might be interested to know that there's a 1,500-year-old tradition that does exactly what you're describing — the Talmud (Gemara). It forces you to hold two contradictory opinions simultaneously, ask what the implications of each are, check whether the claim conflicts with something stated elsewhere, and figure out why. All of it woven through ethics, trade law, marriage, human values. No passive reading allowed — every page is an argument you have to enter. I think it deserves serious consideration as a model for the kind of critical thinking the AI age demands.
Thank you for this article