Wired is reporting on a new Gallup study that finds more than a third of Generation Z adults now turn to artificial intelligence as a form of social companionship. The same research reveals a troubling paradox at the heart of that trend: young people who rely on AI for friendship, or who maintain exclusively online social connections, report higher rates of loneliness than those who do not.
To understand why that finding lands with such force, it helps to place it inside the longer arc of how this generation came of age. Generation Z — broadly, those born in the late 1990s and into the early 2010s — were the first cohort to pass through adolescence with smartphones already normalized, social media already mature, and algorithmic feeds already shaping their sense of what social life looks like. Then the pandemic arrived and compressed years of potential social development into a period of enforced isolation. Schools, clubs, part-time jobs, the low-stakes physical proximity that quietly builds relationship skills — all of it was disrupted at a formative moment. Researchers studying adolescent development had been raising alarms about rising loneliness and anxiety among young people well before 2020. The pandemic did not create those trends, but it accelerated them sharply.
Into that environment arrived a new generation of large language model chatbots — systems capable of holding what feel, at least momentarily, like warm, attentive, responsive conversations. Companies building these tools have not been shy about the companionship use case. Platforms explicitly designed around AI friendship and AI romance have attracted millions of users, and mainstream assistants have been updated with more conversational, emotionally responsive personas. The commercial logic is straightforward: loneliness is widespread, and a product that appears to address it can find an enormous market.
The Gallup data, as reported by Wired, complicates that logic considerably. The finding that AI social companions correlate with higher loneliness rather than lower loneliness is consistent with what social scientists have long argued about the difference between connection and the simulation of connection. Human relationships generate a sense of belonging partly through their unpredictability, their cost, their reciprocity — the fact that another person is also taking a risk, also showing up, also changed by the interaction. A system that is infinitely patient, never tired, never distracted, and algorithmically optimized to be pleasant may satisfy certain surface needs while leaving the deeper architecture of belonging untouched. The likely reading of the Gallup numbers is not that AI causes loneliness, but that the same underlying conditions — social anxiety, underdeveloped relationship skills, geographic or economic isolation — that drive someone toward an AI companion also make genuine human connection harder to build and sustain.
The consequences of this pattern, if it holds, spread across several domains. For the individuals concerned, there is a reasonable worry about a feedback loop: AI interaction may feel like social exercise while actually substituting for the more demanding work of human relationship, leaving underlying deficits in place or deepening them. For the technology companies developing these tools, the Gallup findings are the kind of evidence that tends to precede regulatory and reputational pressure. Policymakers in several jurisdictions have already begun scrutinizing AI companion platforms, particularly their effects on minors and vulnerable users. A credible large-scale study connecting AI companionship to worse loneliness outcomes gives those conversations sharper teeth.
For the broader mental health and public health communities, the data points toward a more urgent conversation about what digital social infrastructure actually delivers versus what it promises. Online-only friendships showing up alongside AI companions in the same risk category is significant — this suggests the problem is not unique to artificial interlocutors but speaks to something about the quality and structure of connection that screen-mediated relationships, however sincere, may struggle to provide.
There are limits to what any summary of a Gallup report can establish. Correlation studies of this kind cannot easily distinguish cause from effect, and the population of people currently choosing AI companions is self-selected in ways that complicate clean conclusions. More longitudinal work, tracking individuals over time, will be needed before anyone can say with confidence what AI companionship does to social development rather than merely who tends to seek it.
What to watch for next is precisely that kind of research, as well as how the companies most invested in AI companionship respond to findings like these. Whether they treat the evidence as a product design problem — an invitation to make their tools less isolating — or as a communications challenge will say a great deal about how seriously the industry takes its own social footprint. And among policymakers, the question of whether AI companion platforms targeting young users should carry any kind of disclosure requirement or usage guidance seems, after this report, considerably harder to dismiss.




