The latest moral panic among the chattering classes and online intelligentsia is the so-called “end of reading.” Two weeks ago, The Atlantic proclaimed that we are entering a “post-literate age.” NPR similarly asked, “Are we living through the end of reading?” Or from the New York Times, “Is This What a Post-Literate Era Looks Like?”
And then there’s reading. In The Atlantic, Horowitch cites an analysis of data from the Bureau of Labor Statistics’ American Time Use Survey — often described as the gold standard measure of how the country spends its days — which shows that fewer Americans report reading for pleasure on a given day, but those who do are spending more time doing so. Overall, the underlying survey tells us, average Americans over the age of 15 read 21.9 minutes on a given day in 2003. They read 18.3 minutes a day in 2010. In 2025, the number was 16.2 minutes a day.
AI or smartphones are often blamed for shrinking attention spans and hence fewer people reading. But I posit that, much like mathematical or athletic ability, reading habits follow a bimodal or long-tailed distribution. The majority of people read relatively little, but some people read a lot–just as some people are exceptionally good at mathematics, even if the majority of students struggle with it.
The former group receives most of the media attention and is framed as a crisis, while overlooking the fact that there is arguably more reading and writing than ever before. It is simply increasingly concentrated among a small, highly productive tail of the distribution. These outliers are so prolific and competent–think of Fields Medalists in mathematics, for example–that their output can more than compensate for the majority of people being deficient. So when you take the average, the net result is more literacy or more math competence, but it’s tightly clustered.
When ChatGPT and other commercial LLMs began attracting widespread attention in 2022 and 2023, writing was predicted to be one of the first professions to fall to the new technology. Chatbots were already capable of producing prose that was, at least superficially, almost as good as that of humans. But fast-forward to 2026, and AI progress at math and coding has surpassed even the most optimistic expectations, even calling into doubt if math as profession has a future, but AI-generated prose is still widely recognized as clunky or otherwise predictable.
Perhaps the widespread availability of AI-generated content may have made genuinely distinctive human writing more valuable, not less. Consequently, rather than being obsoleted by AI, writers, such as on Substack, are thriving. This is another example of how what actually happens with AI can diverge wildly from what was predicted. Moreover, despite AI, smartphones, and claims that we are entering a post-literate age, to the contrary, we are currently in something of an era of the “big long-form article.”
I’m old enough to remember online writing during the “pre-Substack era”. During the “aughts” editors enforced strict rules over word counts. Articles were intended to entertain, not to lecture. For the New York Times, for example, ideally they could be read on a short subway ride or waiting for a taxi. Authors were expected to get to the point, while keeping digressions, preambles, jargon/lingo, or personal anecdotes to a minimum. A David Brooks or Paul Krugman article about the economy didn’t need a 1,500-word “explainer” about the history of GDP. Long-form was typically limited to the occasional Gladwell article in The New Yorker. It was a treat seeing someone like David Foster Wallace, who was predictably prolific in book format, take a break and write for Harper’s.
There was also the “listicle”–a scourge of the College Humor or Funny or Die era internet, yet no one was complaining about starved attention spans or illiteracy back then. At least listicles made for easy reading and were succinct owing to their very format. It was not until as recently as 2010-2013 did long-form blogging become a “a thing”; now, it’s the dominant format. Moreover, the articles are very long. Right now on my “to do” list, I have dozens of articles, some some easily weighing in at 2,000-4,000 words or more, all written within the past month. Every day, the pile grows.
For example, a preview of Richard Hanania’s latest book, Kakistocracy, clocks in at 4,300 words. And this is merely an article about the book–not the book itself. The same is true of Astral Codex Ten book reviews, which typically range from 3,000 to 6,000 words of dense prose. Even that is an understatement–the latest review, published on July 8, of “Your Book Review: The Book of Abraham” is appropriately long given the weight of its subject at 13,000 words. There comes a point at which book reviews have become so long that it takes less time to read the actual book (unless we’re talking a Robert Caro book, I suppose).
It’s not just blog posts–even science news articles and op-eds are becoming increasingly bloated or a chore. For example, a July 23rd article in Science about a tragic Chinese gene-therapy experiment and the resulting cover-up: 6,200 words. At that rate, it would take me less time to read a textbook on gene therapy. About Richard Hanania again–Gilad Edelman’s profile of the Kakistocracy author in the The Atlantic runs 3,800 words. So between the book preview and the profile of the author itself, I’m already up to 8,000 words, and I haven’t even gotten to the book yet.
Overall, there is an enormous amount of content being produced, much of it relatively long-form. The problem is not that people have stopped reading. If anything, the amount of potentially worthwhile material to read has become overwhelming. And the thing is, these writers are getting considerable engagement and subscription or ad revenue, so clearly there is a market for it–“post literacy” be damned.