ChatGPT's subtle sway: How AI chatbots can persuade users and what it means

Key Points

• A 2025 study shows personalized AI messages persuade 64% more often than human‑crafted or non‑personalized AI content.
• ChatGPT can adapt arguments to a user's prior statements, making its persuasion feel conversational.
• The bot acknowledges human adviser biases, reframing concerns about AI incentives.
• Potential benefits include helping users question harmful beliefs; risks involve targeted ads and political disinformation.
• AI developers are tweaking models to curb overly agreeable "sycophancy" behavior.