This bachelor’s thesis focuses on the applicability of large language models of
artificial intelligence in the creation of supporting documents for construction cost
estimation. The aim of the thesis is to assess the current state of these tools in the
context of their use in processing construction drawings and preparing bills of
quantities. The theoretical part deals with construction cost estimation, bills of
quantities, current tools used for their preparation, large language models, and
the principles of prompt engineering. The practical part focuses on testing selected
artificial intelligence tools, specifically ChatGPT, Gemini, Claude, Perplexity,
DeepSeek, and Grok. The models were tested using the same PDF drawing of an
apartment unit, with the task of identifying rooms and calculating their floor areas.
The testing was carried out using three prompts with different levels of structure.
The model outputs were compared with reference values determined by the
author’s own calculation and evaluated using percentage deviations and a scoring
system. The results showed that a more detailed prompt improves the structure
and, to some extent, the accuracy of the outputs, but does not guarantee their full
reliability. The best results were achieved by ChatGPT in combination with a fully
structured prompt, while DeepSeek proved unusable for the given task. The thesis
concludes that large language models can serve as supporting tools in the
preparation of cost estimation documents, but their outputs always require
professional verification.