English teachers' readiness, perceptions and use of AI tools in resource-constrained public high schools of Sindh, Pakistan

Generative artificial intelligence (AI) tools reach teachers through free smartphone applications, yet evidence on whether English teachers in under-resourced government schools of Pakistan are ready to use them remains thin. Drawing on the Technology Acceptance Model (TAM), AI-TPACK and self-efficacy, this study examined High School English teachers' perceptions, readiness and frequency of AI use, and the contextual conditions that shape them in Sindh. An exploratory cross-sectional survey was completed by 64 English teachers from 13 districts, recruited through convenience and snowball sampling. The instrument measured perceptions, four readiness components (AI literacy, AI pedagogical competence, self-efficacy and ethical awareness); institutional support; and resource challenges (α = .77–.95). Data were analysed using correlations, multiple regression, group comparisons, and thematic coding of open-ended responses. Teachers reported high readiness (M = 3.88), rated lowest institutional support (M = 2.99); only 20.3% had received AI training, and 85.9% relied on personal smartphones. Perceptions, benefits, institutional support and challenges together explained 62% of the variance in overall readiness. However, AI pedagogical competence was predicted by AI literacy (β = .41) and self-efficacy (β = .36), not by favourable perceptions (ΔR² = .27, R² = .64). Frequency of AI use was linked to perceptions (r = .48– .49) but not to competence or self-efficacy. Urban teachers use AI more often than rural teachers (d = 0.53; p = .054), though readiness did not differ. Acceptance drives use, whereas literacy and confidence drive the ability to teach with AI.