Testing ChatGPTs New Conversational Voice Mode

The Awkward Reality of Talking to AI in Public

I found myself in the lobby of an auto body shop, waiting for an estimate and realizing I'd forgotten my earbuds. This wouldn't normally be a problem, but I was in the middle of a conversation with my phone—specifically, with ChatGPT. The feeling was as cringey as shouting a question at Siri from across the room or joining a work meeting on speakerphone in a quiet office.

I was testing the advanced voice mode included with GPT-5, the latest AI model from OpenAI. Released this summer after months of anticipation, GPT-5 promised a faster and more intelligent chatbot. Whether it has fully delivered on that promise is still being debated. (Disclosure: CNET's parent company, Ziff Davis, has a pending lawsuit against OpenAI regarding copyright infringement in its AI training.)

What's New with GPT-5's Voice Mode

The updated GPT-5 includes significant improvements to its advanced voice mode, which allows you to have a spoken conversation with ChatGPT. This feature is now available to free users, who previously only had access to a basic version, while paid subscribers get higher usage limits. Another new addition is the ability to assign personalities to your AI, with options like sassy, nerdy, or robotic.

Using the voice mode is straightforward. You just open the ChatGPT app, tap the audio icon next to the text prompt, and start speaking. You can also customize the voice ChatGPT uses in the settings menu.

A Conversation with a Human-Like AI

I decided to treat ChatGPT like a friend, starting our chat with an enthusiastic "Heyyyy girlfriend!" The AI's resulting laugh felt both amusing and slightly patronizing. The voice I chose, named "ember," spoke with a natural, friendly cadence, much like a helpful customer service representative. The chatbot even explained that the upgraded mode was designed to make it sound more human.

One peculiar detail was that the voice would often pause to take breaths, just as a person would. It was a strange affectation, considering we both knew it didn't actually need to breathe. The AI also showed more empathy than I anticipated. When I mentioned I was having a bad week after a car accident, it consistently offered sympathetic phrases, saying it was sorry and agreeing that dealing with insurance is a headache.

While a sympathetic robot might sound harmless, it can be a symptom of a larger issue known as "sycophantic AI," where the model is overly agreeable or emotional. This can be annoying for users who just want facts and potentially harmful for people who use AI as a substitute for therapy, something OpenAI's own CEO has warned against. In the past, versions of ChatGPT have been updated to reduce these sycophantic tendencies.

Voice vs Text Which Is Better for Facts

When it came to factual questions, the voice mode's friendly demeanor was less helpful. I asked for the average cost of car repair labor in North Carolina and for suggestions on where to get a second estimate. Its response was more like a friend's recommendation than a data-driven answer.

In contrast, when I typed the same query into ChatGPT on my laptop, it provided a map and a detailed list of shops with pricing information and hours. The voice chat offered fewer options and described them with vague marketing phrases like "They've been around for quite a while" and "known for quality service." It also failed to provide any links or sources, which is a major drawback for research.

The difference in detail between text prompts (left) and voice chats (right) is clear in the transcribed conversation.

Where Voice Mode Truly Shines Brainstorming

One area where the voice mode excelled was as a brainstorming partner. I used it to help plan a sky-diving-themed birthday party, and it was great for bouncing ideas around, offering new suggestions, and helping me refine my existing plans.

The technology was impressive in its responsiveness. I interrupted the AI multiple times, and it pivoted without issue. I tend to speak quickly, but the chatbot had no trouble keeping up. I could ramble and go on tangents, and thanks to improvements in ChatGPT's memory, it could easily pick up where we left off when I returned to an earlier topic.

The Final Verdict Is ChatGPT Voice Worth Using

Overall, ChatGPT's voice mode is a nice addition, but its usefulness is situational. If you need to conduct in-depth research or gather detailed information with sources, you should stick to the text-based interface. However, if you want to work through a problem out loud or simply have a sounding board for your ideas, the voice mode offers a convenient alternative to typing everything out.

I'm still convinced that we haven't quite normalized talking to our AI assistants in public, at least not without headphones. But for those who think better aloud, it can be a genuinely useful tool. For more on this topic, you can explore how AI is changing search engines and find the best AI image generators.