The integration of artificial intelligence (AI) in generating and managing police-related public opinion content has introduced both opportunities and challenges. While AI enhances efficiency in monitoring and analyzing public opinion, it also risks producing distorted or superficial content that can exacerbate misunderstandings, biases, and public distrust. This paper explores the challenges in managing AI-generated public opinion, the root causes of distortion, and potential strategies to mitigate these risks. Key issues include insufficient preventive mechanisms, limitations in AI-based monitoring and early warning systems, and ineffective communication strategies. Additionally, the paper examines the overemphasis on crisis response, inadequate tools for assessing sensitive content, and poor cross-platform coordination as underlying factors of distortion. To address these challenges, this study proposes strategies such as developing AI-enhanced risk assessment mechanisms, improving real-time public opinion analysis tools, and fostering interdepartmental collaboration. These approaches aim to build a robust governance framework for AI-driven public opinion management, ensuring fairness, accuracy, and public trust.