Generative Artificial Intelligence in Architectural Research: Limits of Regulation

Purpose : The aim is to outline the problem area, define the boundaries of what is subject to regulation, distinguish ineffective regulatory strategies from promising ones, and identify likely trajectories as they apply to master’s and doctoral dissertations, primarily in architectural disciplines. Design/methodology/approach: A comparative analysis of Russian and international regulatory, publishing, and academic documents from 2023–2026 and empirical studies on detectors and the authenticity of generated text, experience in drafting regulations for a dissertation council in architectural disciplines. Research findings: Five areas of uncertainty are identified: authorship and responsibility, text as evidence of competence, reliability and plausibility, detectability and the flip side of metrics, and confidentiality and symmetry of expert evaluation. Three boundaries of regulation are defined, the ineffectiveness of a complete ban, total control, detectors as evidence, percentage quotas, regulation of specific products, and non-interference is demonstrated, transferable solutions from local legislation are described, i.e., functional classification, disclosure instead of prohibition, presumption of good faith and procedural safeguards, symmetry of requirements, industry standards, and technological neutrality. Practical implications : Transferable solutions of a local regulation include a function-based classification of use, disclosure instead of prohibition, a presumption of good faith with procedural guarantees, symmetrical requirements for candidates and reviewers, discipline-specific rules for images and reconstructions, and technological neutrality. Originality/value: The problem is shown to be epistemic and institutional rather than technological. The centre of gravity of assessment shifts from the manuscript to the oral defense of a thesis and verifiable artefacts. The restoration-theory principle of distinguishability of original and supplemented material is applied to the digital artefacts of a dissertation.