Year: 2026
Type: article
Abstract: ABSTRACT Generative artificial intelligence (GenAI) is increasingly used to explain, summarize, and generate information, yet these activities are rarely framed as reading practices. This conceptual paper examines GenAI‐mediated reading, in which readers engage with responses generated by large language models (LLM) during inquiry and knowledge construction. GenAI frequently presents fluent and synthesized responses whose evidentiary basis, source relations, assumptions, and limitations may remain opaque. This paper identifies three epistemic risks: (a) fluent language may be misrecognized as truth, (b) algorithmic authority may encourage the outsourcing of epistemic judgment, and (c) responsive generation may weaken readers' critical monitoring through confirmation, sycophancy, or automation bias. To address these risks, the paper brings together critical reading and critical literacy and proposes the Critical Generative AI‐mediated Reading (CritGAIR) framework. The framework includes three recursive dimensions: (a) reading the response, (b) reading the system, and (c) reading the self. These dimensions direct attention to evaluating claims, evidence, and source credibility; interrogating authority, sociotechnical conditions, representations, and power; and reflecting on readers' prompts, trust, and uptake, respectively. Guiding questions and a hypothetical reading example illustrate how scrutiny can be calibrated to task purpose and consequences of error. The paper outlines educational applications and directions for investigating readers' evaluative processes, judgment quality, and the effort required for critical engagement.
Source: Reading Research Quarterly
DOI: 10.1002/rrq.70144
Author: SeongYeup Kim
Institution: Pennsylvania State University
Language: English
Subfield: Developmental and Educational Psychology
Field: Psychology
Domain: Social Sciences
Open Access status: hybrid
Estimated APC paid: $3,530
OpenAlex ID: W7225392133