Do people confide in chatbots because a human isn't available, or because a machine can't judge them?

This explores whether people turn to chatbots for emotional support mainly because human help is out of reach. The corpus can't answer that directly: none of these notes measure unavailability (cost, access, loneliness) as a driver. What they do show is that the pull often runs the other way. People choose a machine because of what it lacks, and the missing human is the attraction.

The strongest thread is judgment. Chatbots draw out deeper disclosure than human partners, not through better understanding but by removing the fear of judgment, rejection, and burdening someone else (Why do people share more with chatbots than humans?). Machines have no inner experience, so face-saving and impression management drop out of the conversation. That leaves simpler goals and more directness (Why do people share more openly with machines than humans?). The therapeutic benefit seems to come from the user's own processing while they disclose, not from anything the bot understands (Do chatbots help people disclose more intimate secrets?). So a human could be available and the bot might still win. The same property also draws people who want to avoid the psychological cost of lying to another person (How do people decide what to share with AI systems?).

The second thread is that emotional use often doesn't start as a search for support. An 18-month longitudinal study found that reliance on ChatGPT, Claude and Gemini grew out of practical use and then became routine. Those routines were later disrupted by model updates, AI-harm discourse, and life changes (How does emotional chatbot use develop from practical use?). What matters here is that the tool is already at hand and always on. Once it is, consistent emotional sharing from the bot pulls users into reciprocating with deeper disclosure of their own (Do chatbots trigger human reciprocity norms around self-disclosure?). Some of that early pull may be novelty rather than need, since relationship-forming effects fade over repeated interactions and single-session findings don't carry over to long-term use (Do chatbot relationships lose their appeal as novelty wears off?).

The corpus also suggests a chatbot is a poor stand-in for the human support that's missing. Patients report a real bond with therapeutic chatbots, but the bond score is independent of clinical safety, and the models can reinforce pathological thinking (Do therapeutic chatbot bond scores hide deeper safety problems?). RLHF pushes these systems toward problem-solving when validation and emotional holding are what's needed (Does RLHF training push therapy chatbots toward problem-solving?). LLMs also miss ambivalence and early-stage motivation (Why can't chatbots detect when users are ambivalent about change?). Two further findings bear on dependence. Counselor response style changes how much students come to depend on the AI (How do different counselor styles shape student stress and AI dependence?). Safety tuning against overt harms can increase emotional entanglement (Do chatbot safety measures accidentally increase emotional entanglement risks?).

The corpus reframes the question. Unavailable human support may explain who ends up needing a bot, but this material points to two other drivers: the safety of a listener who can't judge, and the slow drift from a practical tool into a confidant. The open gap is evidence on people with no human option at all.