Thirteen-year-olds were reading in 2023 at a level not statistically different from 1971. Only 14 percent said they read for fun almost every day, the lowest share ever recorded.¹ After computers, cable television, the internet, smartphones, tablets, educational software, search engines, and now AI, America has not produced a generation of 13-year-olds who read better than their counterparts did when Richard Nixon was president.
The latest national results deepen the warning. In 2024, NAEP reading scores fell in both fourth and eighth grade. No state recorded a reading gain at either grade compared with 2022. A third of eighth-graders were below NAEP Basic, the largest share ever reported at that level.² The 2026 NAEP reading assessment is underway, but the results will not be released until 2027.³ In the meantime, district-level evidence through 2025 has led researchers and reporters to describe the country as caught in a “reading recession.”⁴
America has heard versions of this alarm before. In 1983, A Nation at Risk warned that if a foreign power had imposed America’s educational performance on the United States, “we might well have viewed it as an act of war.” The report noted that SAT verbal scores had fallen by more than 50 points from 1963 to 1980, while mathematics scores had dropped nearly 40.⁵ The SAT record has to be read carefully because the test-taking population changed. But the warning was not imaginary.
The literacy crisis did not begin with TikTok. But TikTok, smartphones, and AI are not innocent bystanders. They arrived on ground already weakened by poor reading instruction, shallow curricula, declining leisure reading, poverty, uneven schools, and an education culture that keeps changing doctrines while neglecting the mechanics of learning.
Reading is not natural. Speech is. Children acquire spoken language by immersion. Reading must be taught. A child has to learn that marks on a page correspond to sounds, that sounds make words, that words carry meaning, and that meaning accumulates across sentences into memory, story, argument, and judgment.
If the code is not taught well, everything else becomes salvage.
That is why the “science of reading” debate matters. For decades, many American classrooms used versions of cueing theory that encouraged children to rely on context, pictures, and guesses when they met an unfamiliar word. The correction now underway is not glamorous. It does not come with a headset or an app. It begins with the simplest instruction: read the word in front of you, not the word you guessed from the picture.
That is the first literacy. It is no longer the only one.
The OECD’s PISA exam found that average reading performance across OECD countries fell by 10 points between 2018 and 2022, twice the previous record decline. Singapore remained a striking high performer, with 15-year-olds scoring 543 in reading against an OECD average of 476.⁶ The comparison matters because it breaks the habit of treating decline as inevitable. Some systems still produce students who can decode, persist, and reason at high levels. The difference is not that Singaporean children were born outside the digital century. The difference is that systems make choices.
The definition of literacy is widening. OECD says PISA 2029 will make reading the focal domain and add Media and Artificial Intelligence Literacy as an innovative domain.⁷ Schools are not waiting for theory to catch up. The text is no longer only a paragraph on a page. It is also a chart, a search result, a video, a platform ranking, a generated answer, and a number stripped of its denominator.