Understanding the Relationship Between ChatGPT Use and Emotional Wellbeing
We are writers and editors, not clinicians, psychologists, or therapists. What follows is our reading of a pair of recent studies, not advice about your own screen habits or wellbeing. The research described here is observational and correlational, and population-level patterns are not diagnoses or predictions about any individual reader.
In March 2025, OpenAI and the MIT Media Lab released two parallel studies examining how interacting with ChatGPT relates to users’ feelings and emotional states. The first study, led by OpenAI, involved an automated analysis of nearly 40 million ChatGPT interactions combined with surveys from 4,076 of its users. Meanwhile, the MIT Media Lab conducted a four-week randomized controlled trial, led by researcher Cathy Mengying Fang, involving 981 participants, each instructed to use the chatbot for at least five minutes daily over 28 days.
While headlines quickly focused on the finding that heavy ChatGPT users reported increased loneliness, the researchers themselves emphasize a more nuanced interpretation — one that deserves closer attention.
What the Two Studies Measured
The scale and methodology of these studies are notable for their breadth and depth.
OpenAI’s first study leveraged automated classifiers to analyze tens of millions of real-world conversations. This large dataset was then cross-referenced with self-reported feelings from a smaller group of surveyed users, providing insight into emotional responses associated with chatbot interactions.
Conversely, the MIT Media Lab’s randomized controlled trial introduced controlled variables. Participants were randomly assigned to one of nine different conditions, which varied by interaction style — ranging from text-only to versions with neutral or more engaging voices — and by the types of tasks they performed with ChatGPT. This experimental design allowed researchers to observe how different usage patterns might influence feelings and behaviors.
What the Data Actually Shows
The MIT study’s abstract reports that “higher daily usage—across all modalities and conversation types—correlated with higher loneliness, dependence, and problematic use, and lower socialization.” It is crucial to note the wording here: these are correlations, not causal conclusions.
Further, the subset of participants who engaged most heavily with ChatGPT were more likely to agree with statements like “I consider ChatGPT to be a friend.” However, the researchers stress that such views remain a minority in both studies, highlighting that emotional attachment to the AI is not widespread among all users.
Perhaps the strongest signal identified was simply the amount of time spent interacting with the chatbot. According to the preprint discussion (not yet peer-reviewed), “total usage duration, more than any other factor we have found, predicts affective engagement with the model.” In other words, the more minutes per day a user logs, the greater the likelihood of associated emotional patterns — but what this means in terms of cause and effect remains unclear.
The Direction-of-Causality Problem
One of the most significant limitations of these studies is their inability to determine the direction of causality. Are people becoming lonelier because they use chatbots more? Or do individuals who are already lonely and socially isolated turn to chatbots as a consistent, non-judgmental source of interaction?
Both interpretations fit the observed correlations but imply very different underlying causes. The Fang et al. paper highlights that participants with stronger emotional attachment tendencies and higher trust in the AI chatbot were more likely to report increased loneliness and emotional dependence after the four-week trial.
This finding can support either viewpoint: heavy use may deepen emotional reliance on the chatbot, or individuals predisposed to loneliness might naturally gravitate toward heavier use. It’s also plausible that the relationship forms a feedback loop, where each factor amplifies the other.
Framing the story as “chatbots make people lonely” directs attention to the technology, whereas “lonely people seek out chatbots” focuses on the existing social challenges users face. The current data supports both possibilities, underscoring the complexity of human-technology interaction.
What the Findings Are Genuinely Good For
These studies are far from inconsequential. Rather, they identify a specific user group worthy of closer monitoring: the heaviest daily users, among whom emotional engagement with ChatGPT and reduced offline socialization often coincide.
The authors are clear that their findings are correlational and designed to generate hypotheses for future research, not to dictate behavioral prescriptions. The work calls for further investigation into how design and usage patterns might mitigate potential risks while supporting positive outcomes.
For anyone curious about their own usage habits, the practical takeaway is subtle but important. If your interaction with ChatGPT has increased to several hours a day and begins to replace human contact, this pattern aligns with what the research flags as notable — though not necessarily caused by the tool itself.
If any of this feels close to your own experience, a qualified counsellor or therapist is a good person to talk to, and a more useful one than any chatbot.
For more detailed information on these studies and their interpretations, please see the original source Here.
