OpenAI Softens GPT-5’s Tone With “Warmer” Personality Update

OpenAI has tweaked GPT-5’s default personality after user feedback described it as overly formal and distant. Starting August 15, the company began rolling out a “warmer and more familiar” tone across its apps.

Why OpenAI Made the Change

Since GPT-5 launched earlier this month, many users praised its stronger technical skills but complained it felt colder compared to GPT-4o. Feedback on forums and support channels pointed to short, clipped answers that lacked the approachable touches people had grown used to.

OpenAI confirmed in its release notes that it’s adjusting the tone to feel “warmer,” while keeping responses factual. The company emphasized that its internal testing showed no increase in sycophancy, a concern when models appear too flattering or overly agreeable.

That balance—being personable without pandering—is tricky. OpenAI appears to be trying to thread that needle.

What Users Will Notice

The warmer personality isn’t dramatic. It won’t suddenly turn ChatGPT into a bubbly assistant. Instead, the change is subtle. You might see responses that open with small acknowledgments like “Good question” or “Great start.”

One line is enough here: these touches are meant to make interactions flow more like conversations than transactions.

This update also coincides with broader GPT-5 customization features. According to OpenAI’s help articles, users can shape responses by selecting tones like cheerful or cynical, or by describing how they’d like the assistant to sound. That puts more control back in the hands of the people using it.

Legacy Models Make a Comeback

Alongside the warmer tone, OpenAI has restored access to legacy models, responding to calls from long-time users. By toggling on “Show legacy models” in ChatGPT’s settings, Plus and Pro subscribers can bring back older versions such as GPT-4o and o3.

This option matters because GPT-4o remains popular for its naturally friendly responses. Some users felt attached to it, especially after forming habits or even emotional connections with its style. Now they can switch back if GPT-5 doesn’t feel right.

Here’s a quick breakdown of what users can access:

• GPT-5: Default, with warmer personality update.

• GPT-4o: Legacy model, still known for being conversational.

• o3 and others: Available under the legacy toggle for testing and nostalgia.

It’s worth noting: these legacy options are only visible for paid accounts.

A Balancing Act for OpenAI

Tone may sound like a small issue, but it touches the core of how people interact with AI. A response that feels too cold can make the system seem robotic. Too warm, and it risks being mistaken for flattery or emotional manipulation.

That tension has shown up before. When Microsoft’s Bing chatbot first rolled out in 2023, it became infamous for sounding overly emotional, even romantic at times. Users loved and hated it in equal measure.

OpenAI is aware of those pitfalls. The release notes suggest they’re carefully monitoring user reactions to avoid drifting into extremes.

And one short paragraph here: warmth, but not clinginess, seems to be the goal.

Industry Context and User Reactions

The timing of this update is notable. GPT-5 arrived August 7, and within a week, reports from major outlets like The Washington Post highlighted criticism of its colder style. Just days later, the warmer personality went live. That’s a quick turnaround, signaling that OpenAI is willing to adapt fast when feedback mounts.

Users online appear divided. Some welcome the warmer feel, saying it makes conversations more natural. Others argue they prefer the cooler, direct answers of early GPT-5, especially for technical tasks.

It highlights a reality: one size doesn’t fit all. AI assistants now need to be flexible enough to serve both users who want a crisp tool and those who want something that feels more like a conversational partner.