However, we must avoid drawing a false conclusion: the screen itself is not harmful. Watching a film, learning a language via an app, chatting with friends or using artificial intelligence to understand a problem are not the same as spending five hours compulsively scrolling through short videos. What matters is the duration, the content, the context and, above all, what digital use is replacing in real life.
One of the best-established effects relates to sleep. Digital technologies can delay bedtime, disrupt sleep through notifications and keep the brain in a state of arousal. The American Academy of Paediatrics emphasises that excessive digital use is associated with disruptions to the duration and quality of sleep, which in turn are linked, in young people, to poorer attention spans, difficulties at school and emotional problems.
A study involving nearly 2,000 adolescents, reported in September 2026 by the American Academy of Paediatrics, also found that acquiring a smartphone around the age of 13 was associated with more insufficient sleep the following year. In this cohort, higher screen time was also associated with a greater likelihood of depression and obesity. A statistical association does not mean that the phone is necessarily the sole cause, but the evidence is hard to ignore.
Social media is another major issue. According to WHO Europe, the proportion of adolescents with problematic social media use rose from 7 per cent in 2018 to 11 per cent in 2022 in a large-scale survey involving nearly 280,000 young people aged 11, 13 and 15 years of age across 44 countries and regions. The WHO defines this problematic use as characterised by a loss of control, neglecting other activities and negative consequences in daily life.
The link with mental health must, however, be presented with caution. Studies have found links between certain forms of excessive digital use and anxiety, depression, loneliness or reduced well-being. But causality often works both ways: a teenager who is struggling with their self-esteem may be more likely to seek refuge in their phone, whilst compulsive use may in turn exacerbate their difficulties. The WHO specifically emphasises this complex, bidirectional relationship.
There is also an issue with attention. Every notification, fifteen-second video or switch between apps places a fresh demand on the brain. For children aged 6 to 12, the American Academy of Paediatrics reports that excessive digital use is associated with poorer academic performance and weaker attention span. However, it also points out that well-designed educational digital content, used in moderation, can promote certain types of learning.
And now we have artificial intelligence.
With ChatGPT and other generative AIs, a problem distinct from that of mere screen time has emerged: the possibility of delegating part of the intellectual effort itself.
A pupil can ask an AI to summarise a book they haven’t read, solve a problem they haven’t worked out, write an essay they haven’t thought through, or find an answer immediately without having to consult multiple sources. The result may be excellent, whilst masking a lack of learning.
The OECD summarises this paradox quite well in its Digital Education Outlook 2026. Initial research shows that general-purpose generative AIs can improve the results produced by pupils whilst they have the tool at their disposal, without necessarily improving what they actually learn. In particular, the organisation highlights the risk of ‘cognitive offloading’ and metacognitive disengagement: when the machine systematically carries out the reasoning process, certain skills may simply be practised less.
This is probably where the real educational challenge of AI lies.
The brain also learns through effort, error, hesitation, exploration and memorisation. Systematically obtaining the answer before having even tried can short-circuit part of this process. It is a bit like constantly using a calculator before you have even learnt to count: the tool is extraordinarily powerful, but its usefulness depends on when you use it.
This does not in any way mean that AI would make young people less intelligent. We do not have the evidence to support such a simplistic claim. And when used differently, it can have the opposite effect.
The OECD notes precisely that when artificial intelligence is integrated into a genuine educational approach — for example, to challenge pupils, encourage them to argue their case, provide personalised feedback or help them explore different solutions — it can bring about lasting improvements in learning and develop skills such as critical thinking, creativity and reasoning.
In other words: asking AI to ‘do my homework’ and asking it to ‘critique my reasoning and explain my mistakes’ are two radically different uses of the same technology.
Another consequence could be more cultural in nature: a decline in tolerance for boredom. Today, smartphones allow us to eliminate moments of inactivity almost instantly: whilst queuing, on the underground, before going to sleep or simply when nothing is happening.
Yet boredom is not necessarily useless. These are also the periods during which the mind wanders, imagines, organises its memories or comes up with new ideas. We still lack sufficient perspective to gauge exactly the impact of a childhood virtually devoid of such moments.
We must also consider human relationships. Video games, social media and messaging apps can, of course, create genuine communities and sometimes enable isolated teenagers to meet people who share their interests. But when digital relationships largely replace physical interactions rather than complementing them, there is a risk of loneliness. As recently as 2026, the WHO emphasised that social media, games and AI can, in certain contexts, deepen isolation and displace offline relationships.
Artificial intelligence adds a new dimension: young people can now have extremely realistic conversations with machines capable of simulating listening, affection, understanding or friendship. We still have very little understanding of the psychological consequences of daily interaction with these artificial interlocutors throughout a child’s upbringing.
Finally, there are also issues such as disinformation, deepfakes, cyberbullying, unrealistic body standards and content that is algorithmically optimised to grab attention. The WHO has issued a particular warning about young people’s exposure to cyberbullying, unrealistic body images and dangerous content, as well as the lack of transparency in algorithmic systems.
But condemning screens or banning AI would probably be just as simplistic a response as viewing these technologies as harmless.
The current generation will have to live with artificial intelligence. The fundamental question is therefore probably no longer how to prevent young people from using it, but how to teach them to remain intellectually independent when a machine can think, write and create in their place.
We will probably need to adopt a new digital lifestyle: ensuring we get enough sleep without our phones, taking part in sport, having face-to-face conversations, engaging in extended reading, taking breaks from stimulation and, above all, setting aside time to search for answers ourselves before asking a machine.
For the main danger of artificial intelligence for young people may not be that it becomes too intelligent.
It is that we become too accustomed to no longer making the effort to think without it.
