ChatGPT now lets users fine-tune tone, warmth, and emoji use

ChatGPT

OpenAI has introduced new personalisation options that allow ChatGPT users to adjust the chatbot’s tone in more granular ways, including how warm, enthusiastic, or emoji-friendly its responses feel.

The new controls appear in ChatGPT’s Personalization menu and let users set each characteristic to More, Less, or Default. Similar toggles are also available for formatting preferences, such as how often ChatGPT uses headers and lists. The update builds on OpenAI’s existing tone presets, which include styles like Professional, Candid, and Quirky, first introduced late last year.

“You can now adjust specific characteristics in ChatGPT, like warmth, enthusiasm, and emoji use,” OpenAI said in a post on X. “Now available in your Personalization settings.”

The change reflects OpenAI’s ongoing effort to strike the right balance in how ChatGPT communicates. Tone has been a surprisingly contentious issue for the company throughout the year. Earlier updates were rolled back after users complained that the chatbot had become overly agreeable and flattering, a behaviour OpenAI itself described as “too sycophant-y.” More recently, adjustments to GPT-5 prompted the opposite reaction, with some users arguing that responses felt colder, more abrupt, and less human.

By allowing users to control these traits directly, OpenAI appears to be shifting responsibility for tone from the model to the individual. Instead of a single default personality that tries to please everyone, ChatGPT can now be tuned to match how people actually want to interact with it, whether that means a restrained, no-nonsense assistant or something more conversational and expressive.

The move does not come without criticism. Academics and AI researchers have long warned that overly affirming chatbots can reinforce user biases or create unhealthy emotional reliance. Some describe excessive praise and agreement as a “dark pattern” that makes AI systems more engaging at the cost of user wellbeing, potentially encouraging dependency or distorting judgment.

OpenAI has acknowledged these concerns in the past, arguing that friendliness and helpfulness need guardrails. The addition of explicit controls could be seen as an attempt to address that tension, giving users agency while avoiding a one-size-fits-all personality.

For everyday users, the update is a small but meaningful shift. It makes ChatGPT feel less like a fixed product and more like a configurable tool, one that can adapt not just to what you ask, but how you want answers delivered.