What is new
On 7 October 2026 OpenAI announced that GPT-6 is coming to ChatGPT on every plan, worldwide. Plus, Pro, Business and Enterprise got the new model the same day, Free and Go from 8 October. The paid plans run on GPT-6 Sol, Free and Go on GPT-6 Luna. According to the accompanying system card, both replace their predecessors GPT-5.6 Sol and GPT-5.6 Luna.
The change applies to the “Chat” mode, which is what most people mean by ChatGPT. According to OpenAI, the models behind Work and Codex stay as they are for now.
What changes in web search
Whether a business turns up in an answer is mostly decided at the web search. When someone asks for a roofer in Lüneburg, ChatGPT can answer from what it learned in training or look it up. Only when it looks things up do current details from websites, review platforms and directories make it into the answer.
The announcement states that GPT-6 is better at deciding when it needs to search and more reliably finds information that actually supports its answer. For questions that need a web search, the answer reportedly starts 44 percent sooner on average than with the predecessor. How often the model searches now, and which sources it prefers, remains open.
What answers can look like now
GPT-6 brings a new kind of presentation that OpenAI calls Intelligent UI. Answers can now contain graphics, tappable buttons, forms, charts and small interactive tools. OpenAI's own example shows the stops of a road trip on a map.
For local questions ChatGPT already showed map cards with businesses, star ratings and photos before the switch. Whether Intelligent UI changes that presentation, OpenAI leaves open. We will add it here as soon as we see it in our own queries.
What we have seen ourselves
We query ChatGPT for our clients while logged out, so that no account and no stored memory colours the result. For one question we have tracked since 5 September, the answer on 12 October carried different recommendations than in September. Two small calculator sites named back then had disappeared, and a different calculator, absent in September, now came first.
Whether that is down to the new model or to five weeks of change on the web cannot be separated from a single before-and-after. Answers also vary without any model change: in our own repeatability study, the same question asked three times on the same day gave differing results in almost three out of ten cases.
What this means for a business here
If you track your visibility in ChatGPT, your time series has a break since 8 October. A figure from before the switch and one from after compare two things at once: the work on your own presence and the model that answers. A fresh measurement as a new baseline makes sense, with every question asked three times, so that a single outlier stays recognisable as one.
The foundation stays the same. GPT-6, too, names what it finds when it looks things up, and according to OpenAI it now searches more deliberately for details that support an answer. For a business that mostly means a website that clearly states service, location and contact. It also means a well-kept Google profile and reviews written by other people. How to check for yourself whether ChatGPT names you is covered in How to check whether AI already recommends your business.