The Library That Vanished
I remember the first time I Googled something and got an answer. Not a library card, not a dusty encyclopedia, not a librarian’s patient smile—just a screen. That was the moment the internet stopped being a library and became a mirror. A mirror that didn’t reflect truth—it reflected what we wanted to believe, amplified by machines that didn’t care if it was real.
Back then, the internet felt like a promise. All the world’s knowledge, open, free, and waiting. I’d spend hours in the early 2000s, following links from Wikipedia to academic papers to obscure forums, building understanding like a child stacking blocks. I didn’t memorize every fact, but I learned how to find them. That was the skill. That was the gift.
Now? I ask a question and get a polished, plausible lie. It sounds like a professor. It cites sources that don’t exist. It looks like a TED Talk. And I—like most of us—don’t pause. I just click. I trust the smoothness. I’ve outsourced my skepticism.
Cognitive offloading isn’t just convenient. It’s seductive. Why strain your brain when an algorithm will do the heavy lifting? But when the algorithm’s answer is garbage, you don’t just get bad information—you get corrupted thinking. You stop asking, "Is this true?" and start asking, "Does this feel right?" And that’s when the real damage begins.
The Slop Is Everywhere
Look at TikTok. Sixty percent of videos served to new users are AI-generated. Sixty percent. And it’s not just memes. It’s "health advice." It’s "science explainers." It’s children’s cartoons with cats wearing lab coats and melting into puddles of neon goo. The kids’ category? 97 out of 100 videos were AI. Three were real. Three.
And once you engage—once you watch, like, or even pause—your feed doubles down. The algorithm doesn’t care if it’s nonsense. It only cares if it’s sticky. And AI slop? It’s sticky as hell. It’s absurd. It’s grotesque. It’s unforgettable. Hany Farid, the world’s leading deepfake expert, says he stopped trusting his own eyes. That’s not a joke. That’s a warning.
YouTube’s not much better. One in five videos? AI-generated. ScienceDirect’s AI summaries misdefine "domain-general cognition" as "applying knowledge across all domains." It’s wrong. The real definition? Mental processes like attention, working memory—things that actually happen inside your brain. The AI didn’t understand. It just stitched together words that sounded right.
And the citations? Half of them don’t exist. I’ve seen it. Students ask ChatGPT for papers on neuroplasticity. It gives them titles, authors, journals. They cite them in their essays. Weeks later, they find out: no such paper. No such author. No such journal. It was a hallucination dressed in academic robes.
We’re not just drowning in misinformation. We’re drowning in manufactured meaning.
The Paywall Paradox
Here’s the cruel twist: the truth is still out there. But it’s behind a paywall.
Want to read a peer-reviewed study on AI’s impact on cognition? You need a university login. A Nature subscription. A JSTOR account. Want to read the New York Times’ investigation into AI-generated misinformation? You’ll need to pay $15 a month.
Meanwhile, the AI-generated blog post that says "5G causes autism"? Free. The TikTok video that "proves" the moon landing was faked? Free. The YouTube summary that claims "ChatGPT can cure depression"? Free.
The market doesn’t reward truth. It rewards volume. Speed. Virality. The most reliable sources—the ones that fact-check, edit, verify—are the ones that can’t compete with the bots. They’re too slow. Too expensive. Too careful.
And here’s the kicker: the tools that could help us navigate this mess? Also paywalled. Mike Caulfield’s four-step prompting strategy? It works. But it requires GPT-4, Claude 3, or another premium LLM. The people who need this skill most—the students, the elderly, the under-resourced—are the ones who can’t afford it. We’ve built a knowledge economy where truth is a luxury good.
The New Literacy
We’re not living in a post-truth era. We’re living in a post-verification era.
The old literacy was knowing how to use a library. The new one? Knowing how to interrogate a search result.
Mike Caulfield’s advice isn’t a hack. It’s a lifeline.
- Treat the first response as a draft. Don’t read it. Skim it. Then close it.
- Ask: "Give me evidence for and against this—with links to reputable sources." Not "Is this true?"—"Show me your work."
- Demand format: "Summarize this in bullet points." "Explain it like I’m in high school." "Put it in a table."
- Verify every link. Click it. Read the original. Check the author. Is this from a university? A journal? Or a blog that’s been scraping AI outputs since 2023?
It’s not about being cynical. It’s about being curious. It’s about refusing to outsource your critical thinking.
And here’s the uncomfortable truth: no AI can counsel you. No chatbot can diagnose your anxiety. No language model can tell you whether your grief is normal. I’ve seen students take mental health advice from ChatGPT. One girl followed its suggestion to stop taking her antidepressants. She ended up in the ER.
AI doesn’t understand suffering. It predicts words. That’s all.
The Last Library
I still use the internet. I still use AI. But I don’t trust it.
I ask it to challenge me. I ask it to find flaws in my own arguments. I use it to draft, not to decide. I let it do the typing. I do the thinking.
The early internet didn’t fail us. We failed it. We stopped being librarians. We became spectators. We let the machines curate our reality.
But we can still reclaim it.
Not by rejecting technology. By reclaiming our attention. Our skepticism. Our curiosity.
The library isn’t gone. It’s just harder to find now.
And the key? It’s still in your hands.