Building Information Modeling (BIM) underpins digital management across the architecture, engineering, and construction (AEC) sector, yet parametric model creation remains a largely manual process. Practitioners must define component parameters, configure family properties, and specify constraint relationships within commercial BIM platforms – a workflow that is time-consuming and demands specialized expertise. This paper explores a method that uses large language models (LLMs) to automate the generation of BIM parametric designs from natural language descriptions. The proposed pipeline consists of three modules: a semantic parsing module that extracts structured design parameters from text prompts via prompt engineering, a parameter mapping module that translates these parameters into Industry Foundation Classes (IFC) entity attributes through a curated mapping dictionary, and a model generation module that produces standard IFC model files using IfcOpenShell with integrated geometric consistency checks. We evaluate the method on 30 natural language inputs covering three structural systems with varying complexity. For simple descriptions, parameter extraction accuracy reaches 87%, and the model generation success rate is 78%, reducing the requirement-to-model cycle by roughly 60% compared to traditional scripting approaches. The main sources of error are JavaScript Object Notation (JSON) format inconsistencies in large language model (LLM) output and elevation conflicts between components. We discuss limitations related to output determinism, code compliance, and domain coverage, and propose retrieval-augmented generation, domain fine-tuning, and human-Artificial Intelligence (AI) collaborative verification as directions for future improvement.
References
[1] Zou, Y., Kiviniemi, A., Jones, S. W. (2017) A review of risk management through BIM and BIM-related technologies. Safety Science, 97, 88-98.
[2] Zhang, Y., Liu, C., Liu, M., Liu, T., Lin, H., Huang, C.-B., Ning, L. (2024) Attention is all you need: Utilizing attention in AI-enabled drug discovery. Briefings in Bioinformatics, 25(1), bbad467.
[3] Sanderson, K. (2023) GPT-4 is here: what scientists think. Nature, 615(7954), 773.
[4] Aydın, Ö., Karaarslan, E., Erenay, F. S., & Bacanin, N. (2025) Generative ai in academic writing: A comparison of deepseek, qwen, chatgpt, gemini, llama, mistral, and gemma. Turkish Journal of Engineering, 10(2), 592-607.
[5] Çalışkan, E. B. (2025) Exploring possibilities and limits of ChatGPT: usage in building design studies. Turkish Journal of Engineering, 9 (3), 490-500.
[6] Wang, Y., Liao, W. (2026) Generative AI-driven architectural and structural engineering designs: a review. Journal of Intelligent Construction, 4, 9180126.
[7] Kim, S., Jeong, K., Hong, T., Lee, J., Lee, J. (2023) Deep learning-based automated generation of material data with object-space relationships for scan to BIM. Journal of Management in Engineering, 39(3), 04023004.
[8] Du, C., Esser, S., Nousias, S., Borrmann, A. (2026) Text2BIM: Generating building models using a large language model-based multiagent framework. Journal of Computing in Civil Engineering, 40(2), 04025142.
[9] Ko, J., Ajibefun, J., Yan, W. (2025) Generative AI-powered parametric modeling and BIM for architectural design and visualization. Proceedings of the Design Society, 5, 1943-1952.
[10] Gao, Y., Yiu, T. W., Shen, X., Tam, V. W. (2026) Large language models in smart construction: a systematic review of implementation strategies, applications and future directions. Engineering, Construction and Architectural Management, 33(15), 159-181.
[11] Fischer, M., Kunz, J. (2004) The scope and role of information technology in construction. Proceedings-Japan Society of Civil Engineers, 1-32.
[12] Brown, T., Mann, B., Ryder, N., Subbiah, M., Kaplan, J. D., Dhariwal, P., Amodei, D. (2020) Language models are few-shot learners. Advances in Neural Information Processing Systems, 33, 1877-1901.
[13] Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., Kiela, D. (2020) Retrieval-augmented generation for knowledge-intensive nlp tasks. Advances in Neural Information Processing Systems, 33, 9459-9474.
Share and Cite
Peng, X. (2026) Exploring Large Language Model-driven Automated Generation of BIM Parametric Designs. Scientific Research Bulletin, 3(1), 86-94. https://doi.org/10.71052/srb2024/WEHX2553