Cyber attackers increasingly use advanced automation and machine learning to detect and evade traditional security systems, making static honeypots less effective. To address this challenge, this paper introduces a Generative AI–based honeypot trap design that creates dynamic and realistic decoy environments to better engage and analyze attackers. The aim is to develop a cost-effective and scalable solution using AI-generated files, logs, user activity, and system responses to improve deception. The research method includes a comprehensive literature review of honeypots, cyber deception, and generative AI, along with a prototype developed using Python, Flask, and large language models to generate adaptive traps in real time. Early results indicate that generative models can produce highly believable traps that reduce fingerprintability and increase attacker interaction. The system enhances threat monitoring and intelligence by dynamically adjusting to attacker behaviour. Future research will focus on optimizing real-time AI generation, expanding trap diversity, and conducting user testing through penetration simulations.