Recent advancements in large language models (LLMs) such as ChatGPT have shown promise in enhancing persuasive communication. We investigated whether employing direct personality trait labels in GPT prompts can generate persuasive messages for individuals with matching traits. An experiment with 419 participants in four topics—healthcare (Flu Vaccine), consumer products (Smartwatch), financial products (Annuity Insurance), and environmental sustainability (Green Shopping)—revealed that while personality-tailored messages did not consistently result in higher ratings from trait-matched individuals, significant topic influences were observed. For instance, messages emphasizing high Consciousness were more effective for certain topics, regardless of participants’ Conscientiousness levels. Similarly, messages for low Neuroticism regarding some topics were more persuasive for all participants. These results suggest that simple personality trait labels are usually insufficient for generating reliably personalized content, and message effectiveness depends on topic-related stereotypes. Such limitations of basic trait labels in prompts provide clues for sophisticated prompt designs to improve personalized persuasion for different topics with LLMs.