Anthropomorphism in text-based chatbots has become a central construct in research on human–AI interaction, yet findings remain dispersed across studies of trust, social presence, satisfaction, adoption, and purchase intention. This article reports a systematic review and meta-analytic structural equation modelling (MASEM) synthesis of 20 independent samples from 19 peer-reviewed quantitative studies (N = 8,249) restricted exclusively to text-based chatbots across Scopus, Web of Science, PsycINFO, ABI/Inform, and Business Source Complete. Two-stage MASEM yielded robust positive pooled correlations between anthropomorphism and satisfaction (r = .696), trust (r = .633), social presence (r = .599), adoption/use intention (r = .595), and purchase intention (r = .565). Trust emerged as the strongest mechanism linking anthropomorphism to adoption and purchase intention, while satisfaction fully mediated the anthropomorphism–purchase pathway. Evidence for social-presence mediation was directionally positive but statistically inconclusive given the small number of studies reporting this construct pair. The review advances a conservative ACMM framework — antecedents, chatbot anthropomorphism, consequences, mediators, and moderators — grounded in the Computers Are Social Actors paradigm, media equation theory, and the three-factor theory of anthropomorphism. Cultural context moderated the anthropomorphism–trust path (East Asian r̄ = .680 vs non-East Asian r̄ = .577; Δr = .103). The review identifies full Fornell-Larcker matrix reporting as the single most important practice for enabling future integrated MASEM and offers specific reporting recommendations for primary researchers.