OpenAI just gave users something they've been quietly asking for: the ability to tell ChatGPT exactly how to talk to them. On December 19, 2025, the company rolled out granular tone controls that let users adjust the AI's enthusiasm, warmth, and personality characteristics through intuitive slider interfaces. It's a small feature with potentially significant implications for how we interact with conversational AI.

What's Actually New Here

Previously, ChatGPT offered base style presets—options like "candid and friendly" or "direct and concise" that provided broad-stroke personality adjustments. The new system goes considerably deeper. Users can now access Settings → Personalization and fine-tune specific personality dimensions independently, according to OpenAI's release notes.

The controls allow adjustment of:

  • Enthusiasm level – from measured and neutral to energetic and expressive
  • Warmth – controlling how personable versus professional responses feel
  • Additional personality characteristics – though OpenAI hasn't disclosed the full range of available sliders

Changes apply immediately across all conversations, including existing chat threads—a technical detail that suggests these controls operate at the inference layer rather than through system prompt modifications. The feature integrates with ChatGPT's real-time router system, the same infrastructure used for sensitive topic detection, as reported by TechCrunch.

The Technical Context

This isn't just a UI update—it represents a meaningful evolution in how large language models handle personality consistency. Traditional approaches to LLM personality customization rely heavily on system prompts, which can drift over long conversations and compete with user instructions for attention in the context window.

OpenAI's implementation appears to work differently. By tying into the router system, the tone controls likely influence response generation at a more fundamental level, potentially through steering vectors or conditional generation parameters. This would explain why changes propagate instantly to existing conversations without requiring prompt regeneration.

Currently, the feature is available for Plus, Business, and Pro subscribers on chatgpt.com. Mobile app synchronization is planned for the coming weeks, suggesting the backend infrastructure is ready but client-side implementation requires additional work.

Why This Matters: The Market Context

OpenAI isn't building this in a vacuum. The global AI chatbot market is valued at $10-15 billion in 2025, growing at a 24-30% CAGR. In that competitive landscape, personalization has become a key differentiator.

The data supports this focus. According to Zendesk's 2025 research, 70% of CX leaders believe chatbots are becoming skilled architects of personalized customer journeys. Meanwhile, industry analysis indicates that personalized AI chatbots increase customer engagement by 30%, and AI-powered personalization drives a 27% improvement in customer satisfaction scores.

Perhaps most tellingly, 73% of businesses now agree that AI improves personalization strategies—a consensus that's pushing vendors to offer increasingly granular customization options.

Reality Check: What This Is and Isn't

Let's be clear about what granular tone controls actually deliver. This is not fine-tuning in the traditional ML sense. Users aren't modifying model weights or creating custom model variants. They're adjusting parameters that influence generation behavior at runtime.

The practical implications:

  • Consistency – Slider-based controls should provide more reliable personality consistency than prompt-based approaches
  • Limitations – There's only so much personality variation you can achieve without fundamentally changing how the model was trained
  • Edge cases – How these controls interact with complex prompts, multi-turn conversations, and tool use remains to be seen

It's also worth noting that "enthusiasm" and "warmth" are subjective qualities. What feels appropriately warm to one user might feel saccharine to another. OpenAI will need to calibrate these sliders carefully to ensure they produce meaningful, predictable changes across the spectrum.

Implications for Developers and Enterprises

For developers building on OpenAI's APIs, this consumer feature signals where the platform is heading. Expect API-level access to similar controls in future releases—a development that could significantly impact enterprise deployments.

Consider the use cases:

  • Brand voice consistency – Companies could deploy ChatGPT-powered interfaces with precisely calibrated personality settings that match their brand guidelines
  • Context-appropriate responses – A legal services chatbot might dial enthusiasm to minimum while a gaming community bot cranks it up
  • User preference learning – Systems could potentially adjust tone based on user feedback or interaction patterns

The integration with the sensitive topic router is particularly interesting for enterprise applications. It suggests OpenAI is building infrastructure that can modulate response characteristics based on multiple contextual factors simultaneously—tone settings, content sensitivity, and potentially other signals.

The Bottom Line

OpenAI's granular tone controls represent an incremental but meaningful step in AI personalization. It's not revolutionary—we're still working with the same underlying model—but it demonstrates a maturing understanding of what users actually want from conversational AI: not just accurate responses, but responses that feel right.

For the average user, this means ChatGPT can finally stop being quite so relentlessly cheerful if that's not your style. For enterprises and developers, it's a preview of more sophisticated personality customization tools to come.

Resources