<p>The uncontrolled and rapid adoption of generative artificial intelligence (GenAI) in higher education poses both opportunities and risks, particularly in science research courses where claims must be grounded in evidence, sound methodology, and ethical accountability. Ethical issues are of utmost importance, since GenAI poses new challenges in upholding research integrity and ethical conduct. This study employed a constructivist grounded theory to generate a theory explaining the integration and regulation of GenAI technologies in such courses. Through iterative interviews and theoretical sampling, data were collected from eight tertiary-level participants, comprising six science research instructors or advisers and two non-science comparison cases included for analytical refinement. Data were analyzed through constant comparison, memo writing, and progressive category development. Findings informed the development of the GUIDE-AI theory, a substantive instructional framework with a five-phase process comprising the Grounded thinking, User-regulated support, Instructional mediation, Delegated autonomy, and Ethical verification. This framework outlines how instructors scaffold students’ cognitive engagements with the GenAI tools, progressively redistributing responsibility and embedding ethical verification practices. Findings highlight GenAI’s role as a cognitive amplifier, the necessity of a structured developmental sequence supporting student autonomy, and that research ethical accountability must be actively taught and verified. The GUIDE-AI theory explains how instructors regulate GenAI integration through the progressive redistribution of cognitive authority, learner autonomy, instructional mediation, and verification responsibility. The framework offers a substantive explanation of responsible GenAI use within research-oriented instructional contexts.</p>