Gender similarity effects in human–AI interaction: Development and validation of the Behavioral AI Anthropomorphism Scale (BAAS) and implications for trust and gender attribution

As conversational artificial intelligence becomes increasingly embedded in everyday life, users routinely apply human social norms to chatbot-based systems. Yet most existing instruments capture anthropomorphism as self-reported attributions or perceptions of human-likeness, whereas enacted interactional behaviors in dialogue remain undermeasured. Across a large and diverse English sample ( N = 933), we introduce and validate the Behavioral AI Anthropomorphism Scale (BAAS), the first behavioral self-report instrument assessing enacted reciprocity and socioemotional alignment with AI chatbots. Using a random-split sample approach, exploratory and confirmatory factor analyses supported a robust two-factor structure and a hierarchical higher-order model (CFI = .966, TLI = .949, RMSEA = .079, SRMR = .044), yielding strong reliability (α = .83 - .91; ω = .85 - .92) and discriminant validity (HTMT = .85) relative to psychological anthropomorphism. Behavioral anthropomorphism emerged as a central mechanism shaping trust: it explained incremental variance beyond self-reported psychological anthropomorphism (ΔR 2 = .06–.07) and mediated associations between AI use frequency, gender attribution, and trust. Gender similarity effects further revealed that men anthropomorphized AI more when perceiving it as male, whereas women did so when perceiving it as female, with parallel patterns for trust. These findings suggest that trust in AI is more closely linked to self-reported interaction behaviors than to attribution- and perception-based anthropomorphism. By identifying behavioral anthropomorphism as a key relational process in AI chatbot interaction, this work provides a validated tool and a theoretical foundation for designing socially calibrated and user-sensitive conversational AI systems. • The BAAS captures anthropomorphism as enacted social behavior during everyday AI interaction. • The scale distinguishes behavioral anthropomorphism from psychological anthropomorphism. • Psychometric analyses support a robust two factor structure with strong reliability and validity. • Behavioral anthropomorphism predicts trust and social inferences beyond belief based measures. • Gender congruence between users and AI shapes anthropomorphizing behavior and trust formation.