Earlier this month, the OECD released the 2025 results of PISA — the test given to fifteen-year-olds in 91 countries every three years to measure how well they actually read, do math, and reason through problems. The reading numbers are the worst ever recorded since the test began in 2000. And here's the detail that should stop you: the decline didn't start with COVID. Reading scores in OECD countries peaked around 2012 and have been sliding ever since. Whatever's happening here, we can't blame it entirely on lockdowns and Zoom school anymore.
The OECD's own report doesn't dance around a likely cause. It names "increased digitalisation, longer screen time, and falling rates of reading for enjoyment" as coinciding directly with the drop. That's not a talk-radio theory. That's the testing agency's own analysts, in their own published findings.
I caught part of a Clay Travis and Buck Sexton segment recently where Travis made an argument he admitted he couldn't fully back up on the spot — that there's "a depth of knowledge that comes from reading" that video and social feeds don't build the same way, and that he worried science hadn't caught up to prove what he suspected was true.
It turns out the science already has. Cognitive researcher Maryanne Wolf, who has spent two decades mapping what she calls "deep reading circuitry," has published research for years showing that sustained, linear reading engages different neural processes than skimming does — the kind that build inference-making, analogical reasoning, and the ability to hold someone else's perspective in your head long enough to actually consider it. Her line on the alternative is blunt: "When we skim, we literally, physiologically, don't have time to think." Separate research she cites found more than 80% of college educators now see a measurable "shallowing" effect on their students' reading comprehension.
Travis also mentioned, correctly, that more than half of new articles published online are now AI-generated. Independent analyses of the web this year put that figure right around 50% — it briefly crossed the halfway mark in late 2024 and has held steady there since. He's worried about a future full of people with "surface-level ability to analyze things" and no depth underneath it. Between the reading scores and the flood of machine-written text nobody's reading closely enough to evaluate, that's not a fringe worry anymore.
I don't think kids are going back to book reports and card catalogs, and I'm not sure they should. I use AI every day to research and draft pieces like this one — you're reading the result of that collaboration right now. There's no putting that back in the box, and I wouldn't want to.
But here's the part I keep landing on: the less time young people spend with real books, the more they'll need exactly the skills that deep reading builds — not less. A world flooded with AI-generated text, some of it excellent and some of it filler dressed up to look like insight, is a world that rewards people who can actually read closely enough to tell the difference. Skimming a feed doesn't build that muscle. Neither does a highlight reel or a fifteen-second video, however entertaining it is.
My grandkids are lucky — good schools, parents who care whether they're actually reading. Most kids aren't going to trade their Switch or PlayStation for a paperback most nights, and I'm not naive enough to think a blog post changes that. But the occasional hour spent with an actual book, cover to cover, still seems to be building something a scrolling feed can't replicate — and if the research is right, that gap is only going to matter more as AI keeps generating more of what we all read every day.