A fake quotation once required effort. Now it requires a prompt. A misleading chart once required technical skill. Now it can be generated, captioned, and circulated before anyone asks where the numbers came from. The old problem was that people might not know enough to argue well. The new problem is that they may be able to argue fluently without knowing much at all.
A serious literacy agenda for the AI age begins before AI. Teach reading directly in the early grades: phonemic awareness, phonics, fluency, vocabulary, comprehension, and background knowledge. Stop pretending children can infer the alphabetic code by immersion alone. Screen early. Intervene quickly. Treat the inability to read by third grade as an emergency, not a statistic.
Then build knowledge. Reading comprehension is not a free-floating skill. A student understands a passage about the Civil War, photosynthesis, mortgage rates, or Ukraine partly because she already knows enough about history, biology, finance, or geography to make the sentences attach to something. Skills without knowledge become strategies in search of content.
By high school, literacy should include media and AI verification. Students should compare an AI summary with the original, trace a viral claim to its source, learn lateral reading, and understand percentages, base rates, medians, margins of error, and absolute risk. The modern lie often arrives as a number.
Families, libraries, and news organizations have work to do as well. Children are more likely to become readers when they see adults read. Journalism should make the same bargain with its readers: clear ledes, visible evidence, summaries that point into the record rather than replacing it, charts that explain their denominators, and links to primary documents.
Return to the boy at the page. He has already learned the old escape routes: look at the picture, read the room, guess from the first mark, move on before the difficulty catches him. The adult version is not so different. Look at the headline, read the tribe, guess from the tone, ask the machine, move on before the evidence slows you down.
The teacher taps the word again.
“Look at the letters.”
That is still the test.
The machine can write. That part is settled. It can produce the memo, the lesson plan, the summary, the campaign post, the apology, the sermon, the essay, the fake quotation, the plausible chart, the soothing answer.
The human question is harder. Can the student still read well enough to know whether the machine is right? Can the worker understand the system that writes the memo? Can the citizen follow evidence past the headline? Can the journalist, teacher, lawyer, doctor, engineer, pastor, parent, and voter still do what literacy has always demanded: slow down long enough for language to become thought?
The future will not belong to people who refuse AI. Nor will it belong safely to people who surrender to it. It will belong to people who can use machines without becoming machine-readers themselves, scanning, accepting, forwarding, forgetting.
A literate society is not one where everyone can produce words. It is one where enough people can still receive them deeply, test them honestly, and answer back from a mind that has not gone silent.