Answers are grounded in this paper and cite their sources.
Men with prostate cancer have well‐documented unmet informational needs, yet educational materials often fail to meet recommended readability standards. Freely available generative artificial intelligence (GenAI) chatbots offer a potential solution, though the relative performance of different large language models (LLMs) and prompting strategies remains unclear. This study aims to evaluate the ability of publicly available GenAI chatbots to improve the readability of prostate cancer‐related medical information and determine whether prompt type influences readability outcomes. Eight commonly searched prostate cancer questions were identified through literature review, Google Trends analysis, and ChatGPT‐assisted topic generation. Original responses to these questions were used as source material and submitted to nine freely available GenAI chatbots (ChatGPT 5.2 Instant, DeepSeek V3.2, Grok 4.1 Auto, Qwen3‐Max, Microsoft Copilot Smart, Mistral Le Chat Research, Meta AI [Llama 4], Claude Sonnet 4.5, and Google Gemini 3 Fast). Four readability‐improvement prompts were tested: ‘Explain this in simple terms’, ‘Explain this in plain language’, ‘Explain this like I'm five’, and an engineered prompt designed to maximise readability while preserving content. Responses were analysed using the Flesch‐Kincaid Reading Ease (FKRE) score. Linear mixed‐effects models and estimated marginal means were used to compare prompt, chatbot, and content performance. A total of 285 chatbot responses were analysed. Across all models and prompts, readability improved by a mean of 19.64 points (SD 7.59) compared with baseline. All prompting strategies significantly improved readability (all p ≤ 0.005). The prompt ‘Explain this like I'm five’ produced the greatest improvement, increasing FKRE scores by a mean of 27.70 points (95% CI 15.70–39.69; p < 0.001). All chatbot models significantly improved readability compared with baseline (all p < 0.001). Grok 4.1 Auto achieved the highest overall readability score (mean FKRE 71.81) and greatest mean improvement (+23.33 points), while ChatGPT 5.2 Instant and DeepSeek V3.2 consistently improved readability across all tested prompts and content. Readability improvements varied according to subject matter, with the largest gains observed for bladder function (+29.86 points; p = 0.004) and diagnosis (+27.84 points; p = 0.006). Freely available GenAI chatbots can meaningfully improve the readability of prostate cancer‐related medical information using simple prompts. ‘Explain this like I'm five’ produced the greatest readability gains, though this prompt may reduce clinical nuance and accuracy—an important consideration for medical practitioners. When the most effective lay prompt is uncertain, a curated prompt may offer a reasonable alternative. Further research is required to evaluate the safety, accuracy, and completeness of AI‐simplified prostate cancer content, and to explore how patient eHealth literacy influences engagement with these tools.
No takes yet. Share an insight, caveat, or question.
Jackson et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: