The construction industry persistently underperforms in hazard recognition, often leading to severe workplace injuries due to unrecognized hazards. With the recent advancements in Artificial Intelligence (AI) and the emergence of Large Language Models (LLM), the construction sector has begun exploring these technologies for various applications. However, a systematic comparison of popular LLMs to evaluate their effectiveness in identifying construction hazards remains unexplored. Additionally, previous studies have primarily focused on assessing LLMs using textual input and output, leaving their performance with visual inputs underexplored. This study addresses this gap by systematically assessing and comparing the hazard recognition performance of five widely used LLMs using construction case images. The findings establish a baseline standard for LLMs in construction hazard identification through zero-shot learning and reveal that LLMs do not perform significantly well in this context. Additionally, the study provides valuable insights into the reliability and potential applications of LLMs for enhancing hazard recognition in the construction industry.

The recent emergence of multimodal large language models (LLMs) has introduced new opportunities for improving visual hazard recognition on construction sites. Unlike traditional computer vision models that rely on domain-specific training and extensive datasets, modern LLMs can interpret and describe complex visual scenes using simple natural language prompts. However, despite growing interest in their applications, there has been limited investigation into how different LLMs perform in safety-critical visual tasks within the construction domain. To address this gap, this study conducts a comparative evaluation of five state-of-the-art LLMs: Claude-3 Opus, GPT-4.5, GPT-4o, GPT-o3, and Gemini 2.0 Pro, to assess their ability to identify potential hazards from real-world construction images. Each model was tested under three prompting strategies: zero-shot, few-shot, and chain-of-thought (CoT). Zero-shot prompting involved minimal instruction, few-shot incorporated basic safety context and a hazard source mnemonic, and CoT provided step-by-step reasoning examples to scaffold model thinking. Quantitative analysis was performed using precision, recall, and F1-score metrics across all conditions. Results reveal that prompting strategy significantly influenced performance, with CoT prompting consistently producing higher accuracy across models. Additionally, LLM performance varied under different conditions, with GPT-4.5 and GPT-o3 outperforming others in most settings. The findings also demonstrate the critical role of prompt design in enhancing the accuracy and consistency of multimodal LLMs for construction safety applications. This study offers actionable insights into the integration of prompt engineering and LLMs for practical hazard recognition, contributing to the development of more reliable AI-assisted safety systems.