OpenAI is retiring two of ChatGPT's three voice modes. Advanced Voice and Standard Voice will disappear from the app's voice settings on December 11, 2026, and anyone still using either one will be switched automatically to GPT-Live, the model that can listen and talk at the same time.
OpenAI confirmed the retirement on November 10, 2026, in a help centre article and in its release notes. From December 11, users will no longer be able to select Advanced or Standard voice from the model picker in Settings and Voice. Accounts that had either mode selected will default to Live voice, which is powered by GPT-Live, the voice model OpenAI introduced in July 2026. The change applies wherever ChatGPT offers voice, and OpenAI has named no regional exceptions.
The two retiring modes work in different ways. Standard Voice, the original version from 2023, turns speech into text, generates a written reply, and then reads it out, one turn at a time. Advanced Voice processes audio directly but still follows a turn-based format, waiting for the user to finish speaking before it answers. GPT-Live keeps the microphone open while it speaks, so a user can interrupt mid-sentence, add a detail in the middle of an answer, or pause to think without the conversation stalling.
OpenAI's justification is adoption. The company says voice usage in ChatGPT has nearly doubled since GPT-Live launched in July, and that virtually all voice users are already on Live. With the migration effectively complete, keeping two older modes alive means maintaining three separate voice systems for a shrinking group of holdouts. Dropping them lets OpenAI focus its work on improving Live, including giving fuller answers, carrying context across conversations, and adapting to user preferences.
There is also history here. In September 2025, users pushed back when OpenAI tried to remove Standard Voice, and the head of ChatGPT, Nick Turley, posted that it would stay. This time the company has put a fixed cutoff in its release notes and attached a number to justify it, which suggests it does not expect to reverse course again. OpenAI says feedback from people who preferred the older modes will shape how GPT-Live develops.
Not everything built on the older modes disappears on December 11. Live video and screen sharing, the features that let ChatGPT look through a phone camera or at a screen during a voice chat, will keep running on Advanced Voice because GPT-Live cannot do either yet. OpenAI has given no timeline for adding those capabilities to GPT-Live, so for now the December date marks the end of the two selectable voice modes, not the end of Advanced Voice entirely.
Past conversations are also safe. OpenAI says chats created with the older modes will stay in ChatGPT history, and users can reopen one and carry on by typing or speaking. If they speak, the conversation will use Live. The same December 11 date also applies to custom GPTs, the tailored chatbots people build on top of ChatGPT, which use Advanced Voice and are scheduled to retire that day under a separate policy, with later deadlines for some workspaces.
The retirement matters beyond ChatGPT's settings menu because voice is becoming a real search surface. When people talk to an assistant instead of typing, they ask longer, more conversational questions, and they expect a spoken answer rather than a list of links. GPT-Live's ability to listen and speak at the same time makes that kind of exchange feel natural, which is why usage nearly doubled in four months. Every other assistant maker is chasing the same experience, and the winner will shape how people discover products, brands and information by voice.
For publishers and marketers, the shift reinforces a trend that has been building all year: the answer, not the link, is becoming the product. A voice assistant that answers in full sentences has little room for ten blue links, and the sources it cites are the ones whose content is clear, factual and easy to quote aloud. Optimising for that means writing content that answers a question directly in its opening lines, with the detail behind it for anyone who wants more.
The practical takeaway is to treat voice as a citation game. GPT-Live and its rivals draw on the same web that traditional search indexes, but they surface a much smaller number of sources per query, and they favour pages that state facts plainly and back them up. Content that is structured as a direct answer, with supporting evidence close behind, is the content most likely to be quoted, and a quote in a voice answer is the new equivalent of a top ranking.
There is also a measurement gap to plan around. Voice answers rarely produce a click, so the traffic that used to show up in analytics simply does not arrive, even when a brand is the source of the answer. SEOs should start tracking brand mentions and citations inside AI answers alongside clicks, because the December 11 change will push more ChatGPT users onto the model that answers most fluently, and the brands it names will be the ones that win the next round of discovery.
Sources: Keywords Everywhere, November 10, 2026.
Related reading on this site: ChatGPT Business Hub Goes Live — With Click Data for Websites, OpenAI Launches Businesses in ChatGPT With Free Profiles and AI Agents, ChatGPT Sends Just 0.62% of Google's Organic Traffic, ChatGPT Ads: Advertisers Surge, Average Spending Falls, ChatGPT Visual Ads Appear While Your Image Generates — Then Vanish, and AI Assistants Hit 76 Billion Monthly Sessions, Graphite Estimates.
Linkbaits helps you create content that earns citations and backlinks. Start with the toolkit and build your first linkable asset this week.
Start free trial