Social media rewards speed. AI supplies answers. What happens to reading?

What happens to reading when one technology rewards the quickest swipe and another can produce the answer before the pupil has wrestled with the question?

The latest international evidence does not prove that social media or artificial intelligence caused weaker reading. But it does make two links plausible.

Social media can fragment attention. AI can replace some of the mental work that builds understanding. Used differently, the same technologies can also support learning.

So my proposed view is not “ban every screen”. It is this: schools now have to protect attention and make the thinking behind an answer visible.

First, what is PISA?

PISA is the OECD’s Programme for International Student Assessment. Every few years, it gives a computer-based test to a sample of 15-year-olds across participating countries and economies.

It is not a board exam and does not measure whether pupils memorised one curriculum. It tests whether they can apply reading, mathematics and science knowledge to unfamiliar problems. In 2025, more than 760,000 pupils in 91 education systems took part.^[1]

India was not among those participants, so there is no Indian PISA score or rank for this cycle.

Across OECD countries, the average reading score fell by 28 PISA points between 2015 and 2025. Among the other participating systems, the decline was 16 points.^[1] These are changes on PISA’s scoring scale—not percentage falls, pass rates or years of schooling.

The test also found more fast-but-wrong answering. The proportion classified as “hasty” readers rose from almost 7% in 2018 to 11% in 2025 on average.^[2] The OECD sets the response-time thresholds separately by country and major language, so this is not a universal stopwatch test. But the pattern matters: speed rose without accuracy.

The plausible social-media link

A four-year US study followed 8,324 children who were about 10 years old at the start. They reported an average of 1.4 hours of social-media use a day. Higher average use was associated with a gradual increase in parent-reported inattention symptoms over time. The cumulative standardised association over four years was 0.15—small for an individual, but potentially meaningful across a population.^[3]

The researchers adjusted for socioeconomic conditions and a genetic risk score for attention-deficit/hyperactivity disorder. They did not find the same pattern for gaming or watching television and videos. Their time-order analysis also found social-media use predicting later inattention, while earlier inattention did not predict more later social-media use.^[3]

That is stronger than a one-time survey, but it is still not proof that scrolling caused the PISA reading decline. The study involved younger US children, measured symptoms rather than reading and could not randomise years of social-media exposure.

The plausible mechanism is interruption. A feed offers novelty, switching and immediate reward. Careful reading asks almost the opposite: hold a thread, tolerate difficulty, connect information across paragraphs and check whether the conclusion follows.

AI creates a different risk

PISA 2025 found that 46% of pupils across OECD countries used AI chatbots at least weekly to help them learn.^[1] It also found that pupils who reported using AI for specific schoolwork—such as summarising, drafting or research—scored about 20 PISA points lower in science than non-users on average.^[1]

That does not mean AI removed 20 points. Struggling pupils may be more likely to seek help from a chatbot. Schools, tasks and family backgrounds differ. And this was an association with science performance, not a measured effect on reading.

The counterevidence is equally important. Pupils who used AI weekly for general learning performed similarly to non-users. Those who used AI regularly and had school opportunities to build AI literacy tended to score slightly higher, although those opportunities were more common among advantaged pupils.^[1]

The distinction I take from this is between automation and augmentation.

If the pupil asks AI to produce the summary before reading the text, the tool may remove the very act—selecting the main idea, rejecting detail and connecting claims—that the assignment was meant to practise.

If the pupil first reads and writes a summary, then asks AI to challenge it, identify an omitted counterargument or generate questions, the tool can add feedback without doing all the thinking.

What this means for an Indian school

India’s PARAKH Rashtriya Sarvekshan 2024 is not comparable with PISA: the grades, instruments, samples and scales differ. But one Grade 9 language competency is directly relevant. Nationally, pupils answered 54% of questions correctly on identifying main points and summarising after careful listening or reading of news, reports or editorials.^[4]

I would respond with three operating rules.

First, create protected periods for uninterrupted reading. Phones should be away and notifications off. The purpose is not nostalgia for paper; it is to give attention enough time to settle.

Second, change the sequence of AI use: think first, prompt second. Require the pupil’s initial notes, claim or summary before the chatbot opens. Then assess the revision and the reasons for accepting or rejecting the AI’s suggestions.

Third, test accuracy under distraction. Give pupils two sources, one plausible but weak claim and a limited period. Ask them to identify the main argument, compare evidence and explain which source deserves trust. Repeat the task later and track the rubric, not merely completion.

The strongest objection is practical. Social platforms and AI are already part of pupils’ lives; schools cannot recreate an offline childhood. True. Nor should they confuse quiet compliance with deep reading.

But that is precisely why the response must go beyond prohibition. Pupils need to experience sustained attention, recognise when a feed is pulling it apart and learn when AI is extending their thought—or replacing it.

The danger is not simply that children will read on screens. It is that they will become fluent at receiving answers without staying with a difficult idea long enough to form one.