If your customers ask ChatGPT which business to use, a different model answered that question this week than did last week.
On August 6, 2026, OpenAI made GPT-5.6 the default engine for everyday ChatGPT conversations. Free users now default to GPT-5.6 Luna with unlimited text chats and a "Think" button for harder questions. Plus and Pro users get a version of GPT-5.6 Sol tuned specifically for everyday conversation, plus a slider to dial reasoning effort up or down. The older GPT-5.5 Instant is no longer the model composing most consumer answers.
This is not a cosmetic version bump. For anyone who cares about being recommended by AI, the engine making the recommendation is the whole game, and it just got replaced with a more capable one. OpenAI reports that in an internal evaluation across financial, medical, and legal prompts, answers containing at least one factual error were 68% less common with the new GPT-5.6 Sol than with GPT-5.5 Instant. Take the exact number with the usual caution about vendor-run evals, but the direction is clear: the everyday answer engine is now more accurate and more willing to commit to a clear recommendation. So, if you operate in a high-stakes category e.g. finance, health, legal, professional services; it's worth re-checking how you show up now, because the model deciding whether to name you has changed its behaviour.
The measurement question also got sharper
The more useful development this week is a shift in how AI visibility should be measured at all. Semrush published a study of more than 50,000 brands across 1,094 categories in ChatGPT, covering roughly 600,000 citations. The headline finding: only 15.2% of topic categories had a clear brand owner: a single brand that consistently shows up across the related questions buyers actually ask. Another 31.2% had an "emerging leader" appearing in some prompts but without a decisive lead. The rest were fragmented, with no brand reliably owning the conversation.
The definition of "ownership" is the part worth internalizing. A brand only counted as owning a topic if it appeared in at least four of five related prompts with a meaningful lead over the runner-up. In other words, showing up once, for one phrasing of one question, is not visibility. AI visibility is showing up *every time* the AI answers around your topic. As Search Engine Land summarized it, topic ownership in ChatGPT is rare, and traditional SEO strength alone doesn't explain who has it.
This lines up with the strongest academic evidence we have on what drives AI citations. A SIGIR 2026 study of 252,000 trials across six models found that topical relevance and placement, not brand familiarity, are the biggest drivers of being cited first. The practical version of both findings is the same: be the consistently relevant answer across a topic, not the loudest brand name for a single query.
The market context
Underneath these changes, the AI-search market keeps re-baselining. Similarweb's 2026 data shows ChatGPT's share of worldwide generative-AI web traffic sliding from around 76% a year ago to roughly 53% today, with Gemini climbing past a quarter of the category and Claude the fastest grower. Citation presence in US ChatGPT prompts rose from about 1.6% in June 2025 to roughly 6.8% by May 2026, but that average hides enormous variation. Travel and hospitality queries cite sources about 23% of the time and automotive around 20%, while professional services sit under 4%. That spread is critical for expectation-setting. If you run a professional-services firm, you should not benchmark your AI citation frequency against a hotel chain's, because the engines simply cite far less often in your category. The right goal isn't a citation-rate number borrowed from another industry; it's owning your topic within the citation behaviour your category actually generates.
What to do about it
Three moves follow from this week. First, if you're in a high-stakes vertical, re-test your ChatGPT visibility now that GPT-5.6 is the default. Last quarter's read was taken on a different engine. Second, stop measuring visibility one prompt at a time. Pick the five or six questions a real buyer asks around each of your core topics and check whether you show up consistently, with a lead. That's what ownership means. Third, set realistic targets using your own category's citation behaviour, not a cross-industry average.
The through-line across every credible study this year is consistent: AI doesn't reward the brand with the biggest name. It rewards the source that is reliably the most relevant answer, everywhere the question gets asked. That's a target you can actually work toward.
---
Clarity Digital Advisory helps B2B companies get recommended, not just mentioned, by the AI platforms buyers now use to shortlist vendors.