Supporting data for: "Mapping the Role of Artificial Intelligence in Architectural Design Based on ‎Behavioural Patterns:‎ A Systematic Review with Thematic Synthesis of Methods, Applications and Impacts."

This dataset contains the complete evidence base for a systematic review of how artificial intelligence is used to bring human behavioural evidence into architectural design. **Associated article.** "Mapping the Role of Artificial Intelligence in Architectural Design Based on Behavioural Patterns: A Systematic Review with Thematic Synthesis of Methods, Applications and Impacts", currently under review at *Architectural Engineering and Design Management*. The article DOI will be added to the Related works field of this record once it is published. **What the review asked.** Existing reviews of AI in architecture are organised around a technique, a data type or an application domain. This review asked something different: how far does behavioural evidence travel inside the design process before the inference stops? A study can apply sophisticated machine learning to good behavioural data and still end at prediction, never reaching a design decision. To measure that, the review developed a six-level behaviour ladder (BL0–BL6) and a two-part stratum (A1 design-facing, A2 analytical), and coded a corpus against them. **Corpus.** 210 records were identified across three search arms: a systematic database search (January–March 2025), a targeted search of the 2025–2026 literature, and publisher browsing with citation chasing. Following PRISMA 2020, 132 were excluded at title and abstract, 77 were assessed in full, and 29 primary studies were included. Findings were triangulated against a cross-sectional survey of 46 architects, academics and construction professionals. **Files.** Master_Ledger_Recode_PRISMA.xlsx — codebook, the 29 included studies with full coding, the excluded records with reason codes, frequencies, the BL × stratum cross-tabulation, and PRISMA counts. Questionnaire_responses.xlsx — anonymised expert-survey data and analysis. Questionnaire_instrument_EN.pdf — the survey instrument in English translation. PRISMA_2020_checklist.pdf — the completed 27-item checklist. README.pdf — full documentation: inclusion and exclusion criteria, the coding framework, a data dictionary for every column, and the study's limitations. **Please read the README before reuse.** Two limitations matter in particular. Screening and coding were carried out by a single reviewer, with no second screener and no inter-rater agreement statistic. And the survey panel is small and geographically concentrated, so individual respondents should not be re-identified and the adoption figures should not be treated as generalisable. Row-level survey responses are available from the corresponding author on reasonable request. Licence: CC BY 4.0.