Beyond Better Essays: Generative AI, Evidence Evaluation, and Critical Thinking in L2 English Writing: A Source-Audited Critical Integrative Review

Abstract

Generative artificial intelligence (GenAI) can support second-language (L2) writers in planning, drafting, revising, and receiving feedback. But the fact that it can write flowing prose does not prove it can reason independently or rely on dependable sources. This critical integrative review explores what source-audited empirical evidence supports about GenAI, L2 English writing, critical thinking, and evidence evaluation. The evidence set comprises 20 bibliographically verified publications. The researcher confirmed that six were tertiary EFL/ESL English academic-writing studies; 13 were closely related L2/EFL writing studies but lacked one strict scope detail in the accessible primary record; and one reading-based critical-thinking study was retained only as contextual evidence. She extracted only facts supported by publisher records, full texts where available, and Crossref metadata. This review is not systematic, and we make no claims about reproducible database-search logs, dual screening, or pooled effect estimation. Across the audited source record, evidence more often concerned immediate assisted revision, feedback, engagement, and process outcomes than independent transfer. Results for critical thinking were mixed and depended on the instructional design and outcome assessed. No study in the evidence set directly assessed citation accuracy, lateral reading, or a similar objective outcome of evidence evaluation. The review thus distinguishes between textual quality, argumentative reasoning, and evidential quality. It argues that GenAI-supported writing pedagogy should require source verification, claim-evidence justification, and AI-free transfer assessment, rather than using polished, AI-assisted text as a proxy for epistemic learning.