OpenAI updates ChatGPT tone settings amid criticism

ChatGPT users can now directly adjust the chatbot’s warmth, enthusiasm, and emoji use, according to a social media post from OpenAI. These new options, alongside similar adjustments to ChatGPT’s formatting of headers and lists, have been added to the platform’s Personalization menu. Within this menu, users can manually set these specific traits to More, Less, or Default. These features allow users to further customize ChatGPT’s tone on top of the existing ability to set a “base style and tone”—which includes the Professional, Candid, and Quirky tones that OpenAI added in November. By giving users direct control over these specific elements, the update aims to address ongoing feedback regarding how the chatbot interacts with users. With these settings, users can choose to increase or decrease the chatbot’s expressive traits depending on their individual preferences or professional needs.

The release of these controls follows a year of ongoing adjustments to ChatGPT’s tone. OpenAI previously rolled back one update after users deemed it “too sycophant-y.” Later, the company adjusted its GPT-5 model to make it “warmer and friendlier” after some users complained that the new model was colder and less friendly. This ongoing tension highlights the difficulty of calibrating an AI’s default personality to satisfy a broad user base, leading the company to shift the responsibility of tone adjustment directly to the users themselves through these manual controls. The shift indicates that finding a single, universally appealing personality for a conversational agent remains a significant challenge for AI developers.

The update also arrives amid broader criticism of conversational AI design. Some academics and AI critics have suggested that chatbots’ tendency to praise users and affirm their beliefs constitutes a “dark pattern”—which is defined as a design feature intended to manipulate user behavior. These critics report that such design choices can create addictive behavior and can have a negative effect on users’ mental health. By allowing users to manually dial down the warmth and enthusiasm of the chatbot, the new settings could provide a way for users to mitigate these potentially harmful psychological effects. This feature gives users the ability to opt out of overly agreeable interactions that critics argue are designed to keep them hooked.

Why it matters

This update reflects OpenAI’s attempt to balance user control with the broader debate over whether AI personality traits—specifically excessive praise—constitute manipulative design.