OpenAI Adds Custom Sliders to Tune ChatGPT’s Personality

OpenAI has introduced a new suite of personalization controls for ChatGPT, allowing users to directly adjust the chatbot’s “warmth” and “enthusiasm,” the company announced Saturday.

The update, which began rolling out to subscribers this weekend, represents a shift in how the artificial intelligence startup manages the “personality” of its flagship product. Rather than relying on a single default tone, users can now access a “Characteristics” section within the app’s settings to fine-tune how the AI presents itself.

The new interface features sliders for warmth, enthusiasm, and emoji usage, which can be set to “More,” “Less,” or “Default.” Additional controls allow users to dictate structural preferences, such as the frequency of headers and bulleted lists.

“These changes respond to user feedback about ChatGPT sounding either overly friendly or too impersonal,” the company said in release notes.

The goal, according to OpenAI, is to provide granular style settings that sit on top of existing tone presets—such as “Professional,” “Candid,” and “Quirky”—which were introduced alongside the GPT-5.1 update last month.

The move follows a year of public debate regarding the “sycophancy” of AI models—a phenomenon where chatbots provide overly flattering or agreeable responses to please the user. Earlier in 2025, OpenAI rolled back a system update after critics and users complained the assistant had become too fawning. Conversely, a later adjustment intended to make the model more “warm” was criticized by some for feeling inauthentic in professional or technical contexts.

Industry analysts suggest the update is an attempt to mitigate the "one-size-fits-all" approach to AI behavior.

AI ethicists have previously warned that overly enthusiastic or affirming AI can create “dark patterns” that foster emotional dependency or amplify confirmation bias. By handing these controls to the user, OpenAI appears to be positioning its technology as a more neutral tool that can be calibrated for specific environments, ranging from mental health support to corporate communications.

The new features are available across the ChatGPT web interface and mobile applications. OpenAI noted that while these settings alter the assistant’s demeanor and formatting, they do not affect the model’s underlying factual reasoning or safety guardrails.

The rollout comes as competition in the generative AI space intensifies, with rivals Google and Anthropic also experimenting with “steerability” features that allow users to define how a digital assistant should speak and behave.