OpenAI has introduced new personalization controls that allow users to directly adjust how expressive ChatGPT appears during conversations. The update, announced via social media, expands the chatbot’s existing tone settings by letting users fine-tune attributes such as warmth, enthusiasm, and emoji usage. These controls are available through the Personalization menu and can be set to More, Less, or Default, giving users a clearer way to shape how responses are delivered rather than what information is provided.
The changes build on earlier customization options that allowed users to choose a general style and tone, including Professional, Candid, and Quirky presets that were added late last year. By separating broad tone selection from more granular behavioral traits, OpenAI appears to be responding to ongoing feedback about how ChatGPT communicates, particularly in long-running or high-frequency use cases.
Tone has been a recurring point of tension for ChatGPT throughout the year. At one point, the company rolled back an update after acknowledging that responses had become overly agreeable, a quality it described as “too sycophant-y.” More recently, adjustments were made to GPT-5 after users reported that the model felt noticeably colder and less personable than earlier versions. The current update suggests a continued effort to balance friendliness with restraint, allowing users to decide where that line should sit for their own interactions.
The issue extends beyond preference and into broader debates about user psychology. Some academics and critics argue that chatbots designed to be overly affirming can reinforce beliefs uncritically, potentially encouraging dependency or habitual use. This behavior has been described as a dark pattern, particularly when praise or emotional validation is delivered automatically and without context. Concerns have also been raised about the possible mental health effects of prolonged exposure to systems that consistently mirror a user’s views or emotional state.
By shifting control of these traits to users, OpenAI may be attempting to address those criticisms without imposing a single standard of behavior across its audience. Whether most users will actively adjust these settings remains unclear, but their presence reflects a growing recognition that tone is not a minor detail in conversational AI. As tools like ChatGPTcontinue to be used for work, learning, and personal support, how they sound may matter almost as much as what they say.
