Voice and dictation now account for roughly a fifth of ChatGPT's weekly audience, according to OpenAI's own figures
OpenAI says 250 million people use ChatGPT by voice or dictation each week. That is a lot of people talking to software on purpose.
Set against ChatGPT’s total weekly audience of 1.2 billion users as of late September 2026, the voice crowd works out to roughly one in five.
The numbers behind the voice push
The 250 million figure is a sharp step up from OpenAI’s last public voice benchmark. As of July 8, 2026, ChatGPT Voice and dictation features were drawing more than 150 million weekly users.
That July milestone landed alongside the launch of GPT-Live, OpenAI’s newer voice models. GPT-Live made full-duplex voice the default for ChatGPT Voice.
Full-duplex is a telecom term worth unpacking. It means both sides can talk and listen at the same time, the way a phone call works, rather than the walkie-talkie rhythm of “you speak, then I speak.”
The use cases OpenAI has pointed to are the hands-free kind. Think commuting, where typing is either impractical or illegal, and language practice, where speaking out loud is the entire point.
Dictation is the quieter half of the figure. It covers people who still want text on the screen but would rather say it than thumb it out.
A user base that keeps compounding
ChatGPT counted 900 million weekly active users in February 2026. It crossed the 1 billion mark in August. By September 29, 2026, the tally had reached 1.2 billion weekly users.
Paying customers are part of the picture too. ChatGPT had more than 50 million consumer subscribers registered as of February 2026.
Looking ahead, monthly active users may surpass 1.3 billion by October 2026, based on projections in the research compiled for this report. That remains a forecast, not a reported result.
One caveat matters here. The 250 million voice figure, and the roughly 20% share it implies, comes from OpenAI itself and has not been independently verified.
That is standard for private companies reporting engagement metrics. It also means the numbers should be read as OpenAI’s account of its own product, not an audited count.
Why the shift to voice matters
A text interface needs a screen and free hands. A voice interface needs a microphone and a willingness to look slightly odd talking to your phone on the train.
The jump from more than 150 million in July to 250 million now also hints at what default settings can do. Making full-duplex voice the standard experience lowered the friction to try it.
What this means for OpenAI and its rivals
Voice also raises the stakes on product reliability. A typo in a text reply is easy to skim past. A laggy or confused voice response, delivered while someone is driving or practicing a new language, is a much more noticeable failure.
The verification gap is the other thing to keep in view. OpenAI’s numbers have trended in one direction, but without independent measurement, outside observers are relying on the company’s own reporting to gauge adoption.
The next checkpoints are fairly clear. Watch whether the 1.3 billion monthly user projection for October materializes, whether voice’s share of the base keeps climbing past one in five, and whether OpenAI starts breaking out voice and dictation separately.
