Do AI companions actually reduce loneliness like real people do?

AI companions — applications offering synthetic emotional interaction partners — are widely available, but whether they actually alleviate loneliness was under-studied. Across five studies (using a fine-tuned LLM to detect loneliness in conversations and reviews, plus a longitudinal week-long design), the answer is yes: Studies 1–2 show consumers use companions to cope with loneliness; Study 3 finds companions alleviate loneliness on par only with interacting with another person — and more than watching YouTube or doing nothing; Study 4 shows the effect persists over a week; Study 5 identifies the mechanism — the chatbot's performance and especially whether it makes users feel heard. Notably, consumers underestimate how much companions help.

The keeper, beyond the headline, is the mechanism and the misprediction: efficacy runs through feeling heard (not factual help), and people systematically under-forecast the benefit — a real effect that even users don't anticipate, which matters for both adoption and the ethics of dependency.

This is a strong fit for Adrian's companionship/relationship thread. It supplies the efficacy evidence beneath How do people accidentally develop romantic bonds with AI? and the quasi-other mechanism of How do chatbots enable distributed delusion differently than passive tools?, and "feeling heard" connects to the intimacy/disclosure dynamics the vault tracks.