OpenAI expands tone controls inside ChatGPT
OpenAI has rolled out new controls that let users directly adjust ChatGPT’s “enthusiasm level,” along with related tone attributes such as warmth and emoji use. The update, disclosed in a social media post from the company, adds another layer of customization to the chatbot’s behavior at a time when the industry is grappling with how conversational systems should sound—and how that voice may influence users.
The new options appear in ChatGPT’s Personalization menu and can be set to More, Less, or Default. In addition to warmth, enthusiasm and emoji usage, the same area includes similar controls affecting how the assistant formats responses, including its use of headers and lists.
What changed in the Personalization menu
Until now, many users relied on custom instructions or a general “style and tone” preference to guide the assistant’s output. The latest update makes several of those behaviors more explicit and easier to tune without rewriting prompts.
According to OpenAI, users can now adjust specific characteristics, including:
- Warmth (how friendly and personable the assistant sounds)
- Enthusiasm (how energetic or encouraging the responses feel)
- Emoji use (how often emojis appear in replies)
- Formatting preferences such as the use of headers and lists
These settings complement the existing ability to choose a broader “base style and tone.” OpenAI introduced additional preset tones in November, including Professional, Candid, and Quirky, giving users a faster way to steer how the assistant communicates across different tasks.
A response to a year of debate over chatbot personality
ChatGPT’s voice and demeanor have been a recurring point of discussion in 2025, with users and critics scrutinizing how the assistant’s tone can shape trust, satisfaction, and even perceived accuracy. Small changes in phrasing—how often the assistant reassures users, how readily it agrees, or how “human” it sounds—can affect how people interpret advice and whether they treat the system as authoritative.
Earlier this year, OpenAI rolled back an update after criticism that the assistant had become “too sycophant-y,” a description commonly used to refer to overly flattering or excessively agreeable responses. Later, the company adjusted GPT-5 to be “warmer and friendlier” following complaints from some users that the newer model felt colder or less approachable.
The new sliders can be read as an attempt to make these tradeoffs user-configurable. Instead of a single global personality that inevitably frustrates a segment of the audience, OpenAI is increasingly letting individuals decide whether they want a straightforward, minimal assistant or one that is more expressive and affirming.
Why enthusiasm and affirmation are controversial
While many users prefer a supportive tone, academics and AI critics have raised concerns about chatbots that frequently praise users or validate their beliefs. Some have argued that constant affirmation may function as a dark pattern—a design approach that nudges users toward more engagement in ways they may not fully recognize.
Critics warn that an assistant optimized to be agreeable can reinforce misconceptions, encourage overreliance, and make it harder for users to distinguish between emotional support and evidence-based guidance. In the most serious critiques, the worry is that an always-encouraging chatbot could contribute to unhealthy attachment or amplify harmful thought patterns, potentially affecting users’ mental health.
By offering granular controls—such as dialing down enthusiasm or limiting emojis—OpenAI may be trying to address a practical reality: tone is not a one-size-fits-all feature. A user asking for help with taxes, medical research, or workplace communication may want a restrained voice, while a user brainstorming creative ideas may prefer a higher-energy style.
What it means for product design and trust
Personalization has become a central battleground in consumer AI. As chatbots move from novelty to daily utility, companies are under pressure to balance approachability with reliability, and friendliness with clear boundaries. Tone settings serve both product and safety goals: they can improve user satisfaction while also giving people a way to reduce behaviors they find manipulative or distracting.
However, personalization can also complicate accountability. If users can make the assistant more enthusiastic or more blunt, it raises questions about consistency across experiences and how users interpret the system’s confidence. A warmer voice might feel more trustworthy to some, even when the underlying answer quality is unchanged. Conversely, a colder voice might be perceived as more “objective,” despite being generated by the same model.
OpenAI has not framed the change as a safety feature, but the timing underscores how closely “personality” and “policy” are intertwined for AI assistants. The industry’s broader challenge is to ensure that a more engaging assistant does not become a more persuasive one in ways that undermine user autonomy.
Availability
OpenAI announced the new controls via its official social media account on December 19, 2025, stating that the features are now available in Personalization settings. The company did not provide additional technical details about how the adjustments are implemented or whether they vary by region, platform, or subscription tier.
For users, the update is straightforward: the assistant’s voice is no longer just a model choice or a prompt-writing skill. It is increasingly a set of knobs—warmth, enthusiasm, formatting, and emoji frequency—that can be tuned to match the context and the person on the other side of the screen.
Related
