OpenAI began rolling out GPT-6 in ChatGPT on October 7, paired with an Intelligent UI that lets the model choose whether to answer in text, charts, buttons, forms, or a tool it builds on request. Search Engine Journal reported the launch, in a piece by Matt G. Southern. For search teams, the stakes are direct: more tasks can be finished inside the chat window, with no visit to a website.
The rollout started October 7 for four paid plans (Plus, Pro, Business, and Enterprise), inside the Chat tab. Free and Go users get access from October 8. According to OpenAI’s help page, as relayed by Search Engine Journal, the paid plans in this rollout run on GPT-6 Sol, while Free and Go run on GPT-6 Luna. Enterprise access varies with admin configuration.
Availability is uneven by design. GPT-6 replaces the “Latest” option in the model menu, and GPT-5.6 Sol stays selectable as a separate choice. The Intelligent UI is arriving gradually on the web and in updated apps. The Pro thinking level runs on GPT-6 Astra and does not support it, and neither do older macOS and Windows desktop apps. Two users on the same day can see different formats for the same question.
According to OpenAI, GPT-6 was trained to pick a layout based on the question, and plain text remains available when that serves best. Users can request a purpose-built tool from ChatGPT, and OpenAI’s examples are a savings calculator, a bill splitter, and a game played in the conversation. On the web, a Layout and visuals setting reduces the visuals, though the help page warns that some may still appear.
Speed is the other headline claim. On queries that trigger a web search, OpenAI claims GPT-6 Instant starts its answer 44 percent faster, on average, than GPT-5.6 Instant does. That figure is the company’s own, measured against its own prior model. OpenAI also says GPT-6 decides better when to look something up, but its internal test on difficult problems came without a number.
Southern offered an analysis of what this means for tool pages. He noted that the savings calculator and bill splitter rely on ordinary inputs and publicly known formulas. In August, he wrote that Google’s generative UI could most easily copy tools of that kind once it began appearing in AI Overviews. In his reading, calculator and converter pages now face the same pressure from ChatGPT-built tools that users can request and run without leaving the conversation.
The open question is attribution. The announcement does not say how sources appear when answers take the form of charts, comparisons, or tools. ChatGPT’s search help page says answers that use web search can carry citations, reachable through a Sources button when one is available. Whether an interactive answer exposes that button is not confirmed.
That gap shapes how search teams should read their data. Three considerations follow from the reporting:
- Click-out exposure: Queries that ask to calculate, convert, or split are the most likely to be completed in the chat. Compare ChatGPT referral traffic to your tool pages before and after the rollout reaches your audience.
- Page differentiation: A calculator built on public math gives a model little reason to send anyone onward. Pages with proprietary data, live inputs, or context the model cannot generate are harder to replace.
- Visibility tracking: Monitoring built around text answers and cited links can miss interactive responses. Record the plan tier, app version, and whether a Sources button appears for each tracked prompt.
Teams that own calculator and converter pages should log how ChatGPT answers those queries on each plan tier now, because until citation behavior in interactive answers is documented, lost click-out cannot be measured, only inferred.
Search Engine Journal, Matt G. Southern, October 7, 2026, reporting on OpenAI’s GPT-6 announcement and help pages.
