Regional Differences in Self-Reported Composite Scores Related to Integrating Artificial Intelligence into Education in a Sample of Teachers Across Educational Levels in Ecuador

Integrating artificial intelligence (AI) into education requires coordination of pedagogical decisions, data use, and inclusive responses. This exploratory study examined regional differences in observed self-reported composite scores for pedagogical use of AI, educational data management, and adaptation and inclusion among 334 teachers in Ecuador with prior AI experience. Using a cross-sectional design with convenience and chain-referral sampling, data were collected from 2 June to 1 July 2026. Each score averaged three five-point Likert-type items (Cronbach’s α = 0.861–0.890). The ordinal CFA was technically unstable and measurement invariance across regions was not established, so comparisons concern observed scores rather than latent competencies. MANOVA indicated regional differences in the joint vector (Pillai’s trace = 0.050, p = 0.010); Welch ANOVAs remained significant after Holm adjustment, with small effects (ω² = 0.026–0.031). Coast–Amazon standardized differences were moderate, but only adaptation and inclusion remained significant when nine pairwise contrasts were Holm-adjusted. In HC3 regressions, Amazon scores were 0.458–0.483 points lower than Coast scores after covariate adjustment; Highlands–Coast differences were nonsignificant. Models explained 5.8–6.3% of variance, and the pattern persisted in a restricted Basic Education/Unified General Baccalaureate analysis. Findings are limited to this non-probability sample and do not establish population-level regional differences or causal mechanisms.