Exploring Chatgpt as a digital social support tool for bulimia nervosa
PLOS ONE, vol.21, no.3, 2026 (SCI-Expanded, Scopus)
- Publication Type: Article / Article
- Volume: 21 Issue: 3
- Publication Date: 2026
- Doi Number: 10.1371/journal.pone.0345010
- Journal Name: PLOS ONE
- Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, L'Année philologique, Aerospace Database, BIOSIS, Chemical Abstracts Core, EMBASE, Index Islamicus, Linguistic Bibliography, MEDLINE, Psycinfo, zbMATH, Directory of Open Access Journals, Zoological Record, Academic Search Ultimate (EBSCO), Natural Science Collection (ProQuest), Biological Science Database (ProQuest), Biomedical Reference Collection: Corporate Edition (EBSCO), Health Research Premium Collection (ProQuest), Materials Science & Engineering Collection (ProQuest), Pharma Collection (ProQuest), Technology Collection (ProQuest)
- Ondokuz Mayıs University Affiliated: Yes
Abstract
Social support is essential in eating disorder care. Beyond traditional sources, ChatGPT offers a novel and underexplored digital approach that may complement but also potentially challenge existing support systems during treatment and recovery. This study aims to evaluate the potential of ChatGPT as a digital social support tool for individuals with bulimia nervosa. This qualitative and descriptive study used 32 bulimia-related scenarios based on the Edinburgh Bulimia Test. ChatGPT's responses were evaluated via structured content analysis using a scenario-based approach. Each response was independently rated by a psychiatrist and a clinical psychologist across four social support categories: emotional, informational, appraisal, and instrumental. Ratings were made using a 5-point Likert scale (1 = very poor, 5 = excellent). Inter-rater reliability was assessed using Cohen's Kappa coefficients. Friedman and Mann-Whitney U tests were applied to compare support types. Cohen's Kappa indicated agreement between raters (kappa = 0.58-1.00). ChatGPT demonstrated the highest performance in the informational support category across all scenarios, receiving the highest possible scores from both raters. The Friedman test revealed statistically significant differences among the support categories (p < 0.001). Post-hoc pairwise comparisons showed that informational support was rated significantly higher than all other support types (p < 0.001), while instrumental support received the lowest ratings. No statistically significant difference was observed between emotional and appraisal support categories (p = 1.000). This study highlights ChatGPT's emerging potential as a digital provider of social support in bulimia nervosa, particularly in delivering informational support.