What’s Happening to Reading?

What do you read, and why? A few decades ago, these weren’t urgent questions. Reading was an unremarkable activity, essentially unchanged since the advent of the modern publishing industry, in the nineteenth century. In a 2017 Shouts & Murmurs titled “Before the Internet,” the writer Emma Rathbone captured the spirit of reading as it used to be: “Before the Internet, you could laze around on a park bench in Chicago reading some Dean Koontz, and that would be a legit thing to do and no one would ever know you had done it unless you told them.” Reading was just reading, and no matter what you chose to read—the paper, Proust, “The Power Broker”—you basically did it by moving your eyes across a page, in silence, at your own pace and on your own schedule.

Today, the nature of reading has shifted. Plenty of people still enjoy traditional books and periodicals, and there are even readers for whom the networked age has enabled a kind of hyper-literacy; for them, a smartphone is a library in their pocket. For others, however, the old-fashioned, ideal sort of reading—intense, extended, beginning-to-end encounters with carefully crafted texts—has become almost anachronistic. These readers might start a book on an e-reader and then continue it on the go, via audio narration. Or they might forgo books entirely, spending evenings browsing Apple News and Substack before drifting down Reddit’s lazy river. There’s something both diffuse and concentrated about reading now; it involves a lot of random words flowing across a screen, while the lurking presence of YouTube, Fortnite, Netflix, and the like insures that, once we’ve begun to read, we must continually choose not to stop.

This shift has taken decades, and it’s been driven by technologies that have been disproportionately adopted by the young. Perhaps for these reasons, its momentousness has been obscured. In 2023, the National Endowment for the Arts reported that, over the preceding decade, the proportion of adults who read at least one book a year had fallen from fifty-five per cent to forty-eight per cent. That’s a striking change, but modest compared to what’s happened among teen-agers: the National Center for Education Statistics—which has recently been gutted by the Trump Administration—found that, over roughly the same period, the number of thirteen-year-olds who read for fun “almost every day” fell from twenty-seven per cent to fourteen per cent. Predictably, college professors have been complaining with more than usual urgency about phone-addled students who struggle to read anything of substantial length or complexity.

Some of the evidence for the drop in literacy is thin. One widely discussed study, for instance, judges students on their ability to parse the muddy and semantically tortuous opening of “Bleak House”; this is a little like assessing swimmers on their ability to cross fifty yards of molasses. And there are other reasons to be sanguine about the slide away from books, given what so many of us actually like to read. If we binge “Stranger Things” instead of reading Stephen King, or listen to self-help podcasts instead of buying self-help books, is that the end of civilization? On some level, declines in traditional reading are connected to the efflorescence of information in the digital age. Do we really want to return to a time when there was less to read, watch, hear, and learn?

Still, whatever we think of these changes, they seem likely to accelerate. Over the past few decades, many scholars have seen the decline in reading as the closing of the “Gutenberg Parenthesis”—a period of history, inaugurated by the invention of the printing press, during which a structured ecosystem of published print ruled. The internet, the theory went, closed the parenthesis by returning us to a more free-flowing, decentralized, and conversational mode of communication. Instead of reading books, we can argue in the comments. Some theorists have even proposed that we’re returning to a kind of oral culture—what the historian Walter Ong described as a “secondary orality,” in which gab and give-and-take are enhanced by the presence of text. The ascendance of podcasts, newsletters, and memes has lent credence to this view. “The Joe Rogan Experience” could be understood as a couple of guys around a campfire, passing on knowledge through conversation, like the ancient Greeks.

In retrospect, though, there’s something almost quaint about the oral-culture hypothesis. We might say that it was largely developed during the Zuckerberg Parenthesis—a period of history, inaugurated by the invention of Facebook, in which social media ruled. No one inside this parenthesis imagined how much of a threat artificial intelligence would soon pose to the conversational internet. We have already entered a world in which the people you encounter online are sometimes not actually people; instead, they are conjured using A.I. that’s been trained on unimaginably vast quantities of text. It’s as though the books have come to life, and are getting revenge by creating something new—a marriage of text, thought, and conversation that will revise the utility and value of the written word.

In January, the economist and blogger Tyler Cowen announced that he’d begun “writing for the AIs.” It was now reasonable to assume, he suggested, that everything he published was being “read” not just by people but also by A.I. systems—and he’d come to regard this second kind of readership as important. “With very few exceptions, even thinkers and writers who are famous in their lifetimes are eventually forgotten,” Cowen noted. But A.I.s might not forget; in fact, if you furnished them with enough of your text, they might extract from it “a model of how you think,” with which future readers could interact. “Your descendants, or maybe future fans, won’t have to page through a lot of dusty old books to get an inkling of your ideas,” Cowen wrote. Around this time, he began posting on his blog about mostly unremarkable periods of his life—ages four to seven, say. His human readers might not care about such posts, but the entries could make it possible “for the advanced A.I.s of the near future to write a very good Tyler Cowen biography.”

Cowen can think this way because large language models, such as OpenAI’s ChatGPT or Anthropic’s Claude, are, among other things, reading machines. It’s not exactly right to say that they “read,” in the human sense: an L.L.M. can’t be moved by what it reads, because it has no emotions, and its heart can’t race in suspense. But it’s also undeniable that there are aspects of reading at which A.I.s excel at a superhuman level. During its training, an L.L.M. will “read” and “understand” an unimaginably large quantity of text. Later, it will be able to recall the substance of that text instantaneously (if not always perfectly), and to draw connections, make comparisons, and extract insights, which it can bring to bear on new pieces of text, on which it hasn’t been trained, at outrageous speed. The systems are like college graduates who, while they were at school, literally did all the reading. And they can read more, if you give them assignments.

I’ve known a few people who seem to have read everything, and learning from them has been life-changing. A.I. can’t substitute for those individuals because it’s essentially generic and consensus-driven; you won’t look to ChatGPT as a role model for the life of the mind, or thrill to Gemini’s grand theories or idiosyncratic insights. But A.I. has readerly strengths that lie precisely in its impersonality. On David Perell’s “How I Write” podcast, Cowen explains that, as he reads, he peppers a chatbot with questions about whatever he doesn’t understand; the A.I. never tires of such questions and, in answering them, draws on a range of knowledge that no human being could access so quickly. This turns any text into a kind of springboard or syllabus. But A.I. can also simplify: if you’re struggling with the opening of “Bleak House,” you can ask for it to be rewritten using easier, more modern English. “Gas looming through the fog in divers places in the streets, much as the sun may, from the spongey fields, be seen to loom by husbandman and ploughboy,” Dickens wrote. Claude takes a more direct path: “Gas lamps glow dimly through the fog at various spots throughout the streets, much like how the sun might appear to farmers working in misty fields.”

In this way, readers who are armed with A.I. may find themselves blurring the line between primary and secondary sources—especially if they read material for which they believe it’s possible to separate form from content. Many people are already comfortable doing this: since 2012, the Berlin-based company Blinkist, which touts itself as “the future of reading,” has been offering fifteen-minute summaries of popular nonfiction books, in both text and audio format. (In a “blink” lasting a quarter of an hour, you may be able to come to grips with Ryan Holiday’s exploration of Stoic and Buddhist philosophy, “Stillness is the Key.”) Or consider Reader’s Digest Condensed Books, a subscription-based anthology which published, on a seasonal basis, handsome hardcover volumes containing four or five novels that had been trimmed to roughly half their original size. The books were popular—in 1987, the Times reported that one and a half million readers bought ten million volumes annually—and, when I was growing up, my parents kept a shelf of them in our house; without really thinking about it, I read a few “condensed” thrillers by Dick Francis and Nora Roberts. (The series is still offered today, as Reader’s Digest Fiction Favorites.) If I were writing an academic paper on Francis’s novel “Whip Hand,” from 1979, I’d get in big trouble for relying on the condensed version. But if what I’m after is the story, the vibe, the suspense, I might be justified in feeling that I’d “read” the book. Certainly, I’d be unlikely to seek out the unabridged version.

In our current reading regime, summarized or altered texts are the exception, not the rule. But over the next decade or so, that polarity may well reverse: we may routinely start with alternative texts and only later decide to seek out originals, in roughly the same way that we now download samples of new books to our Kindles before committing to them. Because A.I. can generate abridgments, summaries, and other condensed editions on demand, we may even switch between versions as circumstances dictate—the way that, today, you might decide to listen to a podcast at “2x” speed, or quit a boring TV show and turn to Wikipedia to find out how it ended. Pop songs often come in different edits—the clean edit, and various E.D.M. remixes. As a writer, I may not want to see my text refracted in this way. But the power of refraction won’t be mine to control; it will lay with readers and their A.I.s. Together, they will collapse the space between reading and editing.

It’s reasonable to argue that some kinds of writing shouldn’t, or perhaps can’t, be summarized. If you read a summary of Elena Ferrante’s Neapolitan novels—Lila did this, Lenù did that—you cheat yourself. Perhaps Douglas R. Hofstadter’s “Gödel, Escher, Bach: An Eternal Golden Braid” could be boiled down to its key concepts, and maybe a chatbot could explain them to you more clearly than Hofstadter does—but length and difficulty are part of the point of that book. And surely readers will continue to value the authentic voices of their fellow human beings. Recently, I’ve been reading Tolstoy’s “Childhood, Boyhood, Youth.” It’s full of German phrases, odd historical details, and Russian cultural nuances that I don’t understand. Even so, I like to skip the footnotes; I want to stay in the flow of the story, and under Tolstoy’s spell. The proportion of people who simply love old-fashioned reading may be shrinking, but it won’t shrink to zero, or anywhere near it.