AI chatbot usage may increase mental health help-seeking among college-aged individuals

Key Findings

• — 417 participants aged 18 to 28 were recruited through CloudResearch in July 2025 and randomised to an AI-source or human-counsellor-source label on identical support text.
• — Desire to seek professional help increased after the task in both conditions, so the AI label did not suppress help-seeking intention.
• — Rated message quality and perceived emotional supportiveness were similar across the two source conditions.
• — Internet addiction may moderate the effect of message source on perceived emotional support.
• — The authors conclude that AI tools may effectively encourage help-seeking regardless of participants' hikikomori and internet addiction levels.

Methodology Notes

Between-subjects vignette experiment: participants completed hikikomori, internet-addiction and help-seeking-intention measures, read one fixed ChatGPT conversation, were randomised to a source label, then rated message quality, perceived emotional supportiveness and help-seeking intentions again. Single stimulus transcript, online panel sample, intentions rather than behaviour, and no clinical crisis exposure, so it speaks to labelling effects and not to crisis-response safety. Published online ahead of print 2026-09-08, DOI 10.1080/07448481.2026.2727947; tandfonline 403s from this box, so the record was verified from PubMed (PMID 42711684) and the Crossref and OpenAlex records.

Sources