Why Reading Matters More, Not Less, in the Age of AI

An AI engineer and father on why the ability to read, question, and think independently matters more than ever.

The Question I Keep Asking as a Father

I spend most of my working day thinking about how AI can make things easier. As a software engineer I have watched it change how we write code, debug, search, and summarize. Work that took me an afternoon a few years ago now takes minutes, and I am not nostalgic about the afternoon.

But I am also a parent of a young reader, and I am building an AI product for children's reading. Those three roles pull against each other in one specific place. Every day I hand more of my own reading to a machine. Every evening I ask a child to do the opposite: to sit with a book, slowly, without help.

This September the president of Princeton put the same tension to his incoming class, in a question I think about constantly: why should we continue to read long, challenging books when AI agents can quickly summarize them for us?

The more capable AI becomes, the more certain I am of the answer. Our children need to become better readers, not less dependent on reading. And the reason goes well beyond books.

Why Adults and Children Are Not in the Same Situation

Here is the contradiction at the center of my own life, and I think it is the key to the whole question.

AI lets adults remember less, read less, and solve fewer problems unaided, and for most of us that is a fair trade. We already built the underlying abilities. We know what a good answer looks like because we spent years producing bad ones. When an AI drafts something for me, I can tell within a paragraph whether it understood the problem, because I have solved that kind of problem a hundred times without it.

A child has none of that yet. The same tool that amplifies an adult's existing ability can, for a child, quietly replace the process by which the ability was supposed to form. For adults, AI is a multiplier. For children, used carelessly, it is a substitute. That asymmetry is why I can use AI all day and still believe my child's most valuable hours are the ones spent reading without it.

Reading Is Becoming the Interface to Everything

In 2023 Andrej Karpathy, a founding member of OpenAI, wrote that "the hottest new programming language is English." He meant it as good news, and it is. But notice what it assumes.

We used to teach children to speak the language of computers. Now computers are learning to speak ours. That does not make language skills matter less; it makes them matter more. If English is the programming language, reading comprehension is the compiler. Talking to a model takes three abilities: saying precisely what you mean, understanding what comes back, and judging whether it is right. Reading comprehension carries the second and third outright and underpins the first.

AI may be able to generate the answer. Your child still needs to understand the question.

The adult evidence already points this way. When Microsoft Research and Carnegie Mellon surveyed 319 knowledge workers who use generative AI every week, they found that confidence in the AI went with less critical-thinking effort, while confidence in one's own ability went with more. The effort that remained had shifted "from information gathering to information verification." Verification is reading: slow, skeptical, line by line. Paul Graham put the broader point in four words: "writing is thinking." I would add one step before his. Nobody becomes a writer without first being a reader.

Build the Mind Before You Give It Shortcuts

In 2025 researchers at the MIT Media Lab had students write essays over four months while wearing EEG caps. One group used ChatGPT, one used a search engine, one used only their own heads. The brain-only group showed the strongest and most widespread neural engagement; the ChatGPT group showed the weakest, and it declined session after session. After the first essay, more than 80% of the ChatGPT users could not accurately quote a sentence from the essay they had just handed in. Human graders described many of the AI-assisted essays as "generic."

The part I keep returning to is the crossover. Students who had written unaided for three sessions and were then given the AI showed a spike in neural activity: they were comparing the suggestions with their own plans and integrating them. Students who had relied on the AI from the beginning showed weaker engagement when asked to work alone. It was a small study, 54 students and a preprint, and its long-term implications are uncertain. But the contrast raises exactly the question a parent should be asking: what happens when we introduce a powerful shortcut before the underlying skill has formed?

Maryanne Wolf, who has spent her career studying the reading brain, gave those Princeton students her answer: "When you farm it out, as it's so easy to do, you are literally short-circuiting the development of your own brain." A reading circuit is not installed. It is built, through thousands of hours of effortful decoding that slowly becomes effortless, and there is no later moment when it can be bolted on. The years before a child routinely uses AI are the most valuable reading years of their life.

They are also slipping away. The share of Americans who read for pleasure on a given day fell from 28% in 2004 to 16% in 2023. On the 2024 Nation's Report Card a third of eighth graders read below the Basic level, the highest share ever recorded. Across the OECD, fifteen-year-olds' reading has fallen 35 points since 2012, and the OECD's secretary-general described "more hasty reading, students rushing through a text and giving a quick wrong answer." Hasty reading and a quick wrong answer is also a fair description of what happens when someone skims an AI response and accepts it.

Reading Is Not Information Retrieval

There is a deeper answer to the Princeton question, and it is the one I would most like parents to take away.

When a child reads a novel, they are doing something remarkable. They are building a world inside their own mind, imagining people they have never met, feeling things they have never felt, and weighing choices they have never had to make. An AI summary can tell you what happens in a book. It cannot give a child the experience of living through it, chapter by chapter, over a week of bedtimes.

The value of reading is not just what you know when you finish the book. It is what happens to your mind while you are reading it. That is why "AI can summarize it" misses the point entirely. We were never reading to reach the ending.

The Danger of Everyone Thinking the Same Thoughts

There is one more reason, and it is about independence rather than skill.

In 2024 researchers at Exeter and UCL asked 300 people to write short stories, some with AI-generated ideas and some without, and had 600 judges rate them. AI made individual stories about 8% more novel and helped the least creative writers most. But the AI-assisted stories were also about 11% more similar to one another. Each writer got a little better; the collection got a lot more alike. The authors called it "an emerging social dilemma." A Cornell study the following year found the same drift across cultures: Indian and American writers using the same assistant converged on the same defaults.

AI models are trained on patterns across an enormous body of human writing. They can be genuinely creative, and they can also reinforce the most familiar ways of saying and thinking about things. A child who has only ever consumed those patterns has no vantage point from which to notice them. Reading widely gives children something no model can choose for them: their own stock of experiences, perspectives, and voices against which to judge what a machine produces. Wolf's version is blunter: "You can be the author of your own brains, or you can be the lazy recipient of other people's thoughts."

In a world where AI can generate almost anything, the ability to think independently and to tell true information from fluent noise becomes more valuable, not less. Reading, from early childhood, is still the best way we know to build it.

Why We Built ReadBuddy This Way

A fair reader will ask: if you believe all this, why build an AI reading product at all?

ReadBuddy was started for one thing: kids reading. Not kids using AI. We built it around a simple belief: technology should make children want to read more, not make reading unnecessary. The AI in our app helps where a tired parent at eight o'clock often cannot. It asks a child thoughtful questions about a book they have read, listens to the answers, encourages reflection, and shows parents what their child actually understood. It does not replace the experience of reading; its purpose is to help children reflect on books they have read, strengthen comprehension, and build a habit that lasts.

We have never believed that AI makes reading or serious study less necessary. We believe the opposite, and the company exists because of it: AI is there to help children grow well in the world they actually live in, not to stand in for the foundations that growing up requires. We use AI to support the reading journey, never to make the journey unnecessary. The most important part of the experience still happens away from the screen, when a child opens a book and starts to read.

What Parents Can Do Today

None of this needs a program. It needs a habit, started early and protected.

Wolf closed her Princeton talk with a six-word story: "Read deeply, long, with courage and effort." As a parent, I would add one word. Early.

Sources: Princeton University News, 3 Sept 2026; Microsoft Research & Carnegie Mellon, "The Impact of Generative AI on Critical Thinking," CHI 2025; MIT Media Lab, "Your Brain on ChatGPT," June 2025 (preprint); Bone & Sonke, iScience, Aug 2025; National Assessment Governing Board, 2024 Nation's Report Card; OECD PISA 2025; Doshi & Hauser, Science Advances, July 2024; Vashistha & Agarwal, Cornell, 2025; Paul Graham, "Writes and Write-Nots," Oct 2024.