As the large language model Generative Pre-trained Transformer 4 (GPT-4) recently came into being and has attracted much attention, this study examined its efficacy in analyzing the cost of work items and estimating bid prices in construction estimating. This study utilized a rehabilitation project for the Beaver Dam Road Bridge in Pennsylvania, USA as a case study. The authors integrated ChatGPT-4 to handle bid pricing for five specific work items: concrete and formwork, reinforcement, structure backfill, membrane waterproofing system installation, and borrow excavation. Prior knowledge regarding production rates, labor hourly rates, equipment rates, and material rates was used as input. Prompts and instructions were established for interactive execution of the cost estimation. The model's outputs were compared with the ground truth and the bids from three bidders available at Pennsylvania Department of Transportation (PennDOT)’s website. The comparative analysis revealed that GPT-4 holds the potential for construction estimating with reasonable accuracy. However, it is also essential to recognize the consistency and reliability issues that may exist, which would affect ChatGPT’s performance in new scenarios.