PromptAD is a novel method for few-shot anomaly detection that learns prompts using only normal samples. The method addresses the challenge of one-class anomaly detection by proposing semantic concatenation (SC) and explicit anomaly margin (EAM). SC constructs anomaly prompts by concatenating normal prompts with anomaly suffixes, providing sufficient negative samples for prompt learning. EAM introduces a hyper-parameter to explicitly control the margin between normal and anomaly prompt features, enhancing the model's ability to distinguish between normal and anomalous samples. PromptAD achieves state-of-the-art results on MVTec and VisA benchmarks, outperforming existing methods in both image-level and pixel-level anomaly detection. The method is effective in scenarios where only normal samples are available, making it suitable for industrial applications. The results show that PromptAD significantly improves performance in few-shot settings, demonstrating its effectiveness in anomaly detection tasks.PromptAD is a novel method for few-shot anomaly detection that learns prompts using only normal samples. The method addresses the challenge of one-class anomaly detection by proposing semantic concatenation (SC) and explicit anomaly margin (EAM). SC constructs anomaly prompts by concatenating normal prompts with anomaly suffixes, providing sufficient negative samples for prompt learning. EAM introduces a hyper-parameter to explicitly control the margin between normal and anomaly prompt features, enhancing the model's ability to distinguish between normal and anomalous samples. PromptAD achieves state-of-the-art results on MVTec and VisA benchmarks, outperforming existing methods in both image-level and pixel-level anomaly detection. The method is effective in scenarios where only normal samples are available, making it suitable for industrial applications. The results show that PromptAD significantly improves performance in few-shot settings, demonstrating its effectiveness in anomaly detection tasks.