Authentication in Internet of Things (IoT) is critical to protect devices, ensure data integrity, prevent unauthorized access, and protect against cyber threats in interconnected systems. It maintains trust and privacy in the IoT ecosystem. ChatGPT, which is a Generative Pre-trained Transformer, is a powerful tool that can provide a variety of knowledge, including the principles and common practices involved in designing network protocols. In this paper, we use interactive interrogation to let ChatGPT design the authentication protocols required for IoT and explore the security of the resulting protocol. A fatal flaw was discovered in the protocol created by ChatGPT during the security analysis process. In addition, we examine the pros and cons of the generated protocol, provide considerations necessary to protect IoT communications, and further propose an identity-based, enhanced and scalable authentication protocol for IoT devices using bilinear pairing. We also compared the performance of the improved protocol with the original. Research results can be fed back to ChatGPT to improve its knowledge of designing protocols and the improved protocol is more secure, efficient and practical. Moreover, this paper proposes a future research motif aimed at stimulating researchers to enhance artificial intelligence by proposing counterarguments to suggested perspectives on artificial intelligence.