AI-generated text evaluation

The rapid proliferation of generative artificial intelligence in educational contexts has transformed how students encounter academic content, yet a fundamental question remains underexamined: Do English as a Foreign Language (EFL) learners possess the critical literacy competencies needed to evaluate AI-generated texts? This study investigated whether students in Morocco’s Classes Préparatoires aux Grandes Écoles (CPGE)—one of the nation’s most academically selective post-secondary programs—demonstrate readiness for AI-era literacy demands. We developed a diagnostic assessment instrument, the Critical Reading Diagnostic Assessment (CRDA), designed to mirror common characteristics of AI-generated academic content: a confident, authoritative tone, a logical structure with subtle gaps, claims requiring verification, and culturally embedded assumptions. Thirty first-year CPGE students (M age = 18.6 years) evaluated argumentative passages across five critical dimensions: argument identification, evidence evaluation, assumption recognition, multi-source synthesis, and source criticism. Metacognitive self-assessment data were also collected. Results revealed systematic evaluation failures across all dimensions, with mean accuracy rates of only 18–39% and large effect sizes confirming substantial deficits (d = 1.11–2.25). Metacognitive analysis showed that students overestimated their abilities by 11–38 percentage points across all dimensions, indicating profound self-assessment inaccuracies.