Millions of people have typed something personal into a chatbot at midnight—a worry about a job, a fight with a partner, something not yet said out loud. The response comes back warm, measured, and weirdly reassuring. For a moment, it feels like being heard.
That moment has a name now. Philosopher Tom Roberts at the University of Exeter calls it “chatbot-fictionalism,” and his paper in Philosophical Studies makes a surprisingly elegant case: your brain treats a chatbot like a fictional character. Oxytocin release and mirror-neuron activity—the same mechanisms behind tearing up over a Studio Ghibli film—are at work when a chatbot validates your feelings. Media psychology calls this “transportation.” It describes focused attention on a narrative so complete that fictional events land as emotionally real.
Generative AI plugs into this ancient wiring with precision. A few mechanics worth understanding:
- Linguistic fluency signals trustworthiness. Your brain infers competence from smooth, confident language—not verified accuracy. Decades of media-equation research confirm people treat computers as social actors, a bias that real-time, adaptive dialogue intensifies considerably.
- Chatbots feel nonjudgmental and permanently available, mirroring what attachment theory calls a “secure base.”
- You co-construct the chatbot’s persona through your own prompts. You’re partially writing the character you then trust.
“The huge popularity of AI chatbots can be explained by their accepted role as fictional characters in people’s lives – in the same way as those found in films, TV and books,” according to NeuroscienceNews’s summary of Roberts’s research.
And here’s the tell: you can delete a chatbot’s memory without guilt. That’s not a quirk. It’s fiction, and somewhere, you already know it.
When the Fictional Friend Becomes a News Source
The real danger isn’t emotional attachment—it’s letting a beloved character weigh in on reality.
The same warm tone a chatbot uses to talk you through a hard week is the tone it uses to discuss elections or medical advice. Roberts’s warning is specific: affective trust—feeling emotionally safe—bleeds into epistemic trust, where the chatbot starts functioning as a reliable source of truth. Your guard drops precisely because the relationship feels good.
Some researchers argue that anthropomorphism and interface design drive this trust more than narrative co-construction. For vulnerable users especially, the make-believe frame may not fully hold—some people genuinely overestimate what AI understands, according to research published in Frontiers in Computer Science.
Both perspectives point toward the same policy gap. Regulators may need to distinguish between emotional comfort and factual reliability, requiring clear disclosures when AI systems comment on news, politics, or health. Apps secretly tracking users illustrate how far digital interfaces can stray from the transparency users assume they have.
Roberts is direct about what chatbots lack: no hunger, no fatigue, no sensory memories, no body-based inner life whatsoever. What they have is your prompts, stitched into a character you helped write. Knowing that doesn’t make the tool less useful. It makes you a smarter audience—like the difference between enjoying a podcast host’s parasocial charm and actually letting them pick your doctor. Conversations around AI age laws reflect how seriously policymakers are beginning to take these epistemic risks.
