Literacy crises are easy to talk about as if they're waiting somewhere in the future. Kids spend less time reading. Adults scroll more. AI writes the email. Eventually, surely, this will become a problem.
It is already a problem. We have spent years building a world that makes reading optional, then acted surprised when people got worse at it. The latest national numbers are not a warning light. They are the engine making a noise.
The numbers are not subtle
In 2024, just 31% of American fourth-graders scored at or above NAEP Proficient in reading. Forty percent scored below NAEP Basic. Fourth-grade reading fell again from 2022 and sat five points below 2019. Eighth-grade scores also fell, with the losses concentrated among lower- and middle-performing students [1].
A useful caveat: NAEP Proficient is a demanding benchmark, not a synonym for "reads at grade level." But below Basic means exactly what it sounds like: a student is missing even partial mastery of the knowledge and skills the assessment expects. When two out of five fourth-graders land there, arguing over the label is a creative way to avoid the result.
This is not only an American problem. Across OECD countries, 15-year-olds' reading performance fell by 10 points between 2018 and 2022 — roughly half a year of learning. The pandemic clearly hurt. It is also not a complete explanation: international reading performance had already started sliding before 2018 [2].
We did not wake up after the pandemic to a literacy crisis. We woke up during one that was already underway.
The kids grew up
If this were just a story about disrupted elementary school, we could call it a temporary learning-loss problem and wait for a younger cohort to replace the older one. The adult data ruins that comfortable theory.
Between 2017 and 2023, the share of U.S. adults ages 16 to 65 at the lowest literacy level rose from 19% to 28%, according to the National Center for Education Statistics [3]. That assessment is not asking whether someone can recognize letters. At its lower levels, it is measuring whether adults can locate and use information in short texts, especially when the wording is not identical or distracting information gets in the way. These are ordinary tasks in an ordinary life.
Literacy is not a switch that flips on in third grade and stays on forever. It is a practiced capacity. Stop doing the work — sustained attention, decoding, inference, holding an argument in your head — and the capacity erodes. The brain is efficient like that. It gets rid of what you keep telling it you do not need.
We optimized away the practice
Most technology does not need a conspiracy to weaken literacy. It only needs to compete for attention. Dense text loses to a short video. A whole book loses to a summary. Writing a response loses to tapping one of four choices. Each substitution feels harmless because the information is still technically available. But access to information and the ability to understand it are not the same thing.
Then generative AI arrived and offered the final substitution: why read the source when a chatbot can summarize it? Why struggle through the sentence when it can simplify it? Why write your own thought when it can produce one that sounds cleaner? Used carefully, those tools can help. Used as the default, they turn every difficult act of comprehension into an optional step.
That matters because the difficult step is the exercise. You do not become literate by being near correct sentences. You become literate by making meaning from them, connecting them, questioning them, and producing your own. A machine can deliver the result of that work. It cannot quietly install the ability in your head.
The answer is more real work
There is no single fix for a decline this broad. Schools need strong reading instruction. Kids need books worth reading and adults who read with them. People need the time and expectation to wrestle with texts longer than a caption. None of this is glamorous, and none of it can be replaced by a better dashboard.
Learning software can still help, but only if it stops confusing completion with learning. It should ask a student to read, retrieve, explain, and write. It should inspect the reasoning they actually produced and tell them exactly where it broke. It should make practice more available without making practice disappear.
That is why Lune Synth™ starts with the student's own work. Read the question. Think through it. Write the answer by hand. Then let the system grade what you did and send you back to the precise skill that needs work. AI belongs after the effort, where it can improve feedback — not before it, where it can erase the reason to learn.
Declining literacy is not inevitable. But reversing it requires an unfashionable idea: people get better at reading and writing by reading and writing. We should build tools that make that work count again.
Sources
- National Center for Education Statistics, 2024 NAEP Reading: National Trends and Student Skills .
- OECD, PISA 2022 Results (Volume I): The State of Learning and Equity in Education , 2023.
- National Center for Education Statistics, PIAAC 2023 Highlights of U.S. National Results .
Want to try it? We're rolling out invites to the beta. The first 100 users get two months free and a lifetime 50% off Lune Synth™ Pro. Drop your email on the home page and we'll reach out when it's your turn. Thoughts or pushback? griffin@lunesynth.com.