Chatbot appearance and loneliness trajectories in women with high autistic traits

Key Findings

• — The group-by-time interaction was significant: loneliness decreased over eight weeks in the non-human-appearance condition, while the slope in the human-like condition was bounded near zero (95% CI -0.92 to 0.99).
• — Baseline loneliness did not differ between the two groups, so the divergence arose during use rather than from allocation.
• — The two chatbots were identical in dialogue strategy and usage structure, isolating visual anthropomorphism as the manipulated variable.
• — The authors conclude that appearance shapes longer-term interaction outcomes rather than only first impressions, and that for neurodiverse users it is a core design dimension that warrants longitudinal evaluation.

Methodology Notes

Two-arm randomized longitudinal experiment, n = 72 women with high autistic traits, eight weeks of use at a minimum of three sessions per week, loneliness measured at four time points, linear mixed-effects analysis with preregistered directional predictions. Published in the August 2026 issue (volume 9, article 100369); the journal gives month precision only, so the date is set to the first of the month, with the Crossref record created 2026-08-06. Open access under CC BY-NC-ND. ScienceDirect blocks automated fetching from this workspace, so the record was built from the Crossref metadata and the full abstract deposited on OpenAlex; the full text was not read, and the second author's affiliation is not confirmed. Recruitment country and the autistic-trait screening instrument are not stated in the abstract.

Sources