The Reading Crisis Before AI

Part two of a series. The first piece made an argument. This one shows the receipts.

In the first piece in this series, we argued that the real danger of AI is what it does to people whose reading and attention are already weak. That argument rests on a claim that has to be true for the rest to matter: that the weakening is real, measurable, and older than the technology. So here is the evidence. None of what follows is a guess. It is in the test scores.

Are students really losing the ability to read?

In June 2026, a writing instructor named Tyler Jagt described assigning his class a 20-page article, the same length he had assigned for five years and read without complaint as an undergraduate a decade earlier. Not one student finished it. One told him it was too long, and that she kept losing track of what it was about. His essay in The Chronicle of Higher Education described what he sees as a structural shift, not a bad semester.

The generational collapse in literacy is measurable, persistent, and likely to get worse.

Tyler Jagt, The Chronicle of Higher Education

One instructor’s classroom is an anecdote, and on its own you should not build much on it. The reason to take it seriously is that the national numbers say the same thing he does.

What do the test scores actually show?

The most consistent long-run measure of American learning is the National Assessment of Educational Progress, the federal test often called the Nation’s Report Card. It has tracked reading since 1992, which makes its trend lines hard to wave away. Both ends of the school system are sliding at once.

Measure (2024)	Result
12th-grade average reading score	Lowest since the test began in 1992
12th graders below “Basic”	32%, the largest share ever recorded
4th graders below “Proficient”	70% (up from 68% in 2022 and 66% in 2019)
Last time 4th grade hit 70% below proficient	2005

The 12th-grade figures come from the 2024 Nation’s Report Card. The 4th-grade figures were compiled by the Annie E. Casey Foundation from the same federal data. Children entering the system and teenagers leaving it are both reading worse than the students before them.

Which students are hit hardest?

Unevenly, and that matters. In 2024, the Casey Foundation found that at least eight in ten American Indian or Alaska Native (85%), Black (84%), and Latino (80%) fourth graders were not proficient in reading, alongside 65% of multiracial and 61% of white fourth graders. Between 2019 and 2024, 46 states saw the share of below-proficient fourth graders rise. A decline this broad is not about one kind of student or one kind of school. It falls hardest on the children with the least support to begin with, which is the opposite of what a healthy system would do.

Did this start before AI?

This is the question that decides how to read everything else, and the answer is yes. ChatGPT launched in November 2022. The reading declines were years deep by then. Federal officials noted that scores for the lowest-performing students began falling more than a decade ago and kept falling before the pandemic. The timeline is not subtle.

What does the smartphone have to do with it?

More than the chatbot, most likely. In a 2017 study of nearly 800 people, the University of Texas researcher Adrian Ward found that the mere presence of your own phone reduces your available mental capacity, even when the phone is off and face down. Move it to another room and people perform better on attention-heavy tasks. As Ward put it, the effort of not thinking about the phone is itself using up the resources you would otherwise spend on the task.

Attention has measurably fragmented in the same window. The informatics researcher Gloria Mark has tracked how long people stay on a single screen before switching. In 2004 it averaged about two and a half minutes. By the mid-2010s it had fallen to roughly 47 seconds, a figure other labs have since replicated. Sustained reading is precisely the muscle a device built to interrupt you keeps pulling you off of. The reading brain needs long, unbroken attention, and that is the thing in shortest supply.

So where does AI fit in?

It arrives late, into a population already reading less and concentrating less, and it removes the last bit of friction. Before, a distracted student still had to skim the article to fake having read it. Now the machine reads it for them and hands back a summary in seconds. The work that used to build the skill, however grudgingly, can be skipped entirely.

That is the danger the first piece described. A tool magnifies whoever holds it, and a reader who has stopped reading is exactly who it magnifies in the wrong direction. The technical name for handing the mental task away is cognitive offloading, and AI makes it frictionless in a way nothing before it did. AI did not start this fire. It poured accelerant on one already burning.

Can the trend be reversed?

There is real reason to think so. Reading is a built skill, not a fixed trait, which is why it responds to practice and why the brain builds more from slow, effortful work than from shortcuts. Scores that moved down can move back up. A trend is not a destiny. What it takes in practice, for you, for your kids, and at work, is the subject of the next pieces in this series. The evidence here only settles the first question: the decline is real, it is measured, and it was well underway before anyone typed a prompt.

Common questions

Did AI cause the reading decline?

No. The declines were measurable years before ChatGPT launched in late 2022. AI is best understood as an accelerant of a trend that smartphones and changes in schooling set in motion, not as the cause.

Are test scores a fair way to measure this?

No single test is perfect, but the Nation’s Report Card is the most consistent measure the United States has, running since 1992. When the same downward trend appears across different grades, different years, and most states at once, it is hard to explain away as a testing quirk.

Is this only an American problem?

The school data here is from the United States. The attention research is not country-specific: smartphones fragment focus the same way everywhere they are used, which is nearly everywhere.

What is the most telling single number?

The climb from 66% to 70% of fourth graders below proficient between 2019 and 2024. It is recent, it is large, and most of it happened before AI was in widespread use.

Sources

• The Nation’s Report Card (NAEP): Grade 12 Reading, 2024
• National Assessment Governing Board: declines in 12th-grade reading (2025)
• Annie E. Casey Foundation: 7 in 10 fourth graders below proficient (2024)
• Tyler Jagt, “My Students Can’t Read,” The Chronicle of Higher Education (2026)
• University of Texas: “Brain Drain” smartphone study (Ward et al., 2017)
• Gloria Mark: research on the shrinking attention span

You may also like

• The Tool That Uses You (part one)
• What AI Literacy Means
• Physical Books, Handwriting, and Learning
• Glossary: What Is Cognitive Offloading?
• AI Memory: Augment vs Replace
• How to Use Claude AI
• The daily Beginners in AI newsletter