A recent study by The Rithm Project offers one of the more differentiated looks at how young people engage with artificial intelligence today. Based on a survey of over 2,300 participants aged 13 to 24, it moves beyond the usual binary of opportunity versus risk and instead examines the interplay between AI use and the quality of young people’s social relationships.
The central question guiding the research is both simple and far-reaching:
Are young people’s relationships shaping how they use AI, and is AI in turn shaping those relationships?
The answer, perhaps unsurprisingly, is that both dynamics are already deeply intertwined.
The study identifies four distinct patterns of AI use among young people.
A first group, comprising roughly 28 percent of respondents, uses AI rarely or not at all. Contrary to common assumptions, this group does not represent a digitally protected or more resilient cohort. On the contrary, they report the highest levels of loneliness, anxiety and social isolation within the sample. Non-use, in this context, is often less a conscious refusal and more a reflection of limited access, lack of orientation or fragile social environments.
A second and largest group, representing 39 percent of participants, uses AI primarily as a functional tool. These young people rely on AI for information, learning and task completion. Their social lives are comparatively stable, and they demonstrate the strongest indicators of well-being across the study. Importantly, they maintain a clear boundary between AI as an instrument and human relationships as the primary space for emotional exchange.
A third group, comprising 18 percent, uses AI for personal and relational support. Here, AI becomes part of how emotions are processed and social situations are navigated. These young people are not disconnected, yet they report higher levels of anxiety and a lower sense of relational security. AI functions as an additional, sometimes more accessible channel for dealing with complex feelings.
Finally, 15 percent of respondents engage with AI in more personified ways, interacting with characters or simulated relational counterparts. This group represents the most intimate form of AI use and, correspondingly, the highest level of potential relational risk.
Across these clusters, one finding stands out with particular clarity:
AI use cannot be understood in isolation from the relational environments in which young people are embedded.
The more stable and supportive these environments are, the more likely AI is used as a tool. The more fragile or insufficient they are, the more likely AI begins to take on relational functions.
What emerges from this analysis is not a simple narrative of technological harm or benefit. Rather, the study points to a structural dynamic that is easy to overlook in current debates.
AI does not merely introduce new risks. It interacts with existing conditions and amplifies them.
This becomes particularly visible in one of the study’s key insights: the more intimate the use of AI, the greater the associated risks. These risks do not arise solely from the technology itself, but from the role it begins to play when it enters spaces that were previously shaped by human relationships.
At the same time, the study highlights a second, equally important aspect. Young people are, to a large extent, navigating these developments alone. Guidance from adults, whether in educational settings or within families, remains limited. As a result, the responsibility for making sense of AI’s role in their lives is often carried by individuals who are themselves still in the process of forming identity and relational competence.
From the perspective of international law, this constellation is not unfamiliar. It closely resembles situations in which the precautionary principle has historically been applied.
The precautionary principle addresses contexts where potential impacts are significant, where causal pathways are complex, and where waiting for full evidence may lead to irreversible consequences. It does not call for the suspension of innovation, but for the early shaping of conditions under which that innovation unfolds.
Applied to the findings of this study, the question shifts.
It is not primarily whether AI should be used by young people, since its presence in their lives is already a given.
The more relevant question is under which conditions AI becomes supportive of human development, and under which conditions it begins to substitute or displace elements of human connection that are critical for that development.
What the data suggests is that the decisive variable is not the intensity of AI use alone, but the quality of the relational environment surrounding it. Young people who experience themselves as seen, safe and supported in their relationships show a markedly lower tendency toward high-risk forms of AI engagement.
This points to a form of responsibility that extends beyond individual behaviour. It concerns the design of social, educational and institutional contexts in which young people encounter AI.
At present, there is a noticeable gap between the expansion of technological possibilities and the structures that would enable a responsible integration of these technologies into everyday life. Responsibility is, in effect, being individualised, while the corresponding frameworks remain underdeveloped.
The study does not suggest that AI will replace human relationships in any immediate or absolute sense. What it does show, however, is the beginning of a more gradual shift. Where AI offers constant availability, non-judgmental responses and low-threshold interaction, it may become an increasingly attractive option in situations where human relationships are experienced as demanding, uncertain or insufficiently supportive.
Over time, this has the potential to influence not only behaviour, but expectations. Expectations of responsiveness, of emotional safety, of what it means to be understood.
From a governance perspective, this is precisely the type of development that calls for early attention. The precautionary question, then, is not whether to allow or restrict use, but how to ensure that the environments in which this use takes place are capable of sustaining the forms of human connection on which development ultimately depends.
At present, this question remains insufficiently addressed.
And that, more than the technology itself, is where the real challenge lies.
Author
With Co-Founder
Santa is an international environmental lawyer, leadership coach, and co-founder of FindingSustainia. As a member of the Climate Innovation Fund Stuttgart, she develops systemic strategies for scalable, regenerative solutions.
Her work connects climate policy, emotional intelligence, and cyclical leadership to sustain changemakers and institutions in their effectiveness over the long term—even amid complexity and crises.
Through the Be the System Shift™ approach, she collaborates with Dr. Anna Katharina Meyer to deliver programs that enable transformation on both the individual and systemic level—including with Fortune 500 companies, ESSEC Business School, and networks of the Club of Rome and SchuleWirtschaft.