Growing Up With Generative AI

Generative artificial intelligence (AI) is rapidly becoming part of everyday life for young people. From homework assistance and information searches to conversational companions and AI features embedded in social media, these tools are changing how teenagers learn, communicate and spend their time. They also play a role in shaping how young people relate to themselves and others, particularly when used for companionship or emotional support.

This growing use raises questions about young people’s emotional well-being and, in some instances, their physical integrity. In this context, debates over what this means for children’s overall well-being, whether they need additional protection, and what form that protection should take are advancing rapidly – often faster than the empirical evidence about how young people actually experience these technologies.

To shed light on how children experiment generative AI, interface commissioned YouGov to survey 970 teenagers aged 11 to 17 in France alongside their parents. The study examines access and frequency of use, reasons for turning to generative AI, perceptions of its benefits and risks, possible links with emotional well-being, and the support parents feel they need. It distinguishes between general-purpose AI tools such as ChatGPT, AI companions such as Character.AI, AI integrated into social media such as My AI on Snapchat, and voice assistants such as Siri.

Below are the main findings. They are also available in French here.

---

Generative AI use is widespread among teenagers – and starts early

Generative AI use is widespread on both sides of the generational divide. The study reports frequent use among both the surveyed parents and teenagers: 63% of both parents and children use at least one generative AI tool at least weekly.

General-purpose tools, such as ChatGPT, are the most commonly used category. Among the French minors surveyed, 78% report using general-purpose generative AI, compared with 55% for voice assistants, 50% for AI integrated into social media and 32% for AI companions. Regular use also becomes more common with age, particularly for general-purpose AI and AI integrated into social media. Still, adoption starts at an early age: 56% of 11 to 12-year-olds use at least one type of generative AI at least once a week. On a weekly basis, 40% of them use general-purpose generative AI and 33% of them use voice assistants. More strikingly, 23% of them report using AI embedded in social media platforms and 16% report using AI companions, when they should not even have access to these services – as per most of the services’ own terms and conditions.

Parents, however, do not always have a clear picture of their children’s activities. In particular, they underestimate their children’s use of general-purpose AI, with the largest gap among youngest children. Among 11-12-year-olds, 40% report using general-purpose AI regularly, while parents estimate that figure at 26%.

The younger the child, the more parents underestimate their use of AI

Source: YouGov for interface, “Parents and teenagers in the age of generative AI”, p. 17

---

Teenagers mainly see opportunities where parents also see risks

For teenagers, generative AI is above all a practical resource: a way to save time, find answers, learn and complete schoolwork. Among teenagers familiar with it, 89% identify at least one positive aspect. Parents also recognise its educational potential, but they are considerably more ambivalent. Their concerns include diminished independent thinking, misleading information and risks to personal data. 44% of parents whose children use generative AI say they don’t feel at ease when their child uses it.

Although they are outweighed by benefits, parents do have concerns

Source: YouGov for interface, “Parents and teenagers in the age of generative AI”, p. 29

The contrast is not simply between enthusiastic teenagers and sceptical adults: it points to a broader policy challenge – how to preserve the benefits young people themselves value while addressing risks that they may perceive less readily.

---

AI companions and AI embedded in social media deserve particular scrutiny

One of the study’s clearest findings concerns time spent on these services. Among teenagers who use generative AI, 92% say they spend more time than they would like to on at least one of the tools covered by the survey. While this figure is not a clinical measure of addiction, it points to a widespread feeling of spending more time with these systems than intended.

Almost every child who uses generative AI feels like they spend too much time on it

Source: YouGov for interface, “Parents and teenagers in the age of generative AI”, p. 32

This feeling is especially pronounced for AI companions and AI integrated into social media: 59% of AI companion users and 57% of users of AI integrated into social media say they spend more time with these tools than they would like, compared with 45% for general-purpose generative AI users and 43% for voice assistants users. These services are particularly relevant from a child-safety perspective because they are designed around repeated interaction: AI companions explicitly sustain ongoing conversations artificially, while social media AI is embedded in environments already optimised for frequent engagement.

The findings also suggest that the relational dimension of AI use is particularly noteworthy when looking at younger children. Although 11-12-year-olds are less likely than older teenagers to use generative AI regularly overall, 14% of 11-12-year-olds who use generative AI report using it to confide in it, compared with 6% among 13–15-year-old users. Regular use of AI companions is also already present among the youngest respondents, at 16%, compared with 13% among 13-15-year-olds and 15% among 16-17-year-olds. 

These figures suggest that policy should look beyond questions of access and pay close attention to the design of systems that encourage sustained, personalised or emotionally significant interaction, especially when they are used by younger children.

---

What’s the impact of generative AI use on teens’ well-being?

The survey presents a nuanced picture. Most teenagers report positive general well-being, regardless of their AI use. At the same time, regular generative AI users more frequently report negative emotions. Among the surveyed teenagers overall, 38% report experiencing negative emotions at least weekly. This rises to 44% among regular generative AI users and to 50% among regular users of AI companions.

Regular users of AI companions experience negative emotions more regularly

Source: YouGov for interface, “Parents and teenagers in the age of generative AI”, p. 38

Although these numbers should not be interpreted as evidence that generative AI causes emotional distress in minors, the differences are significant and do warrant attention. They indeed suggest that the relationship between certain forms of generative AI use and emotional well-being deserves closer scrutiny, particularly where systems are designed for repeated or relational interaction.

---

Parents want more support

French parents are facing these developments with limited support. Only 32% feel sufficiently supported in helping their children navigate generative AI use; another 26% feel supported, but not enough, and 42% do not feel supported at all. Those who are more vulnerable, or who use generative AI rarely or not at all, feel especially left behind.

Asked who should help them, parents most frequently identify schools and teaching staff (51%), followed by generative AI providers themselves (41%) and public authorities (30%). This suggests that parents do not see navigating AI as a matter of individual or family responsibility alone. They expect support both from institutions responsible for education and from the companies that design and deploy these systems.

Making generative AI safer for young people requires taking their real-life experiences into account – rather than simply managing their access to it. It also means looking more closely at how these systems are designed, the types of interactions they encourage, and how responsibility for mitigating potential risks should be distributed.

As generative AI becomes increasingly embedded in children’s everyday lives, policy responses must simultaneously foster opportunities for learning and participation through these tools where appropriate, better inform families, and strengthen oversight of AI systems designed to capture users’ attention over prolonged periods or foster emotional attachment. These policies must be grounded in empirical evidence rather than assumptions. It is precisely in this spirit that this study seeks to contribute to the public debate on online child safety.