Semrush Traffic Analytics data for July 2026, published by Search Engine Land on Thursday, puts concrete numbers on the AI search transition that GEO practitioners have been modelling for eighteen months. ChatGPT reached 1.09 billion US monthly visits — ranking ninth nationally — up 48.38 per cent year on year. Bing fell 50.43 per cent, the steepest single decline in the US top-20 website rankings. Google grew 10.72 per cent. DuckDuckGo fell 15.83 per cent. The pattern is not ambiguous: AI chat is absorbing the query volume that was flowing to secondary search engines, while Google continues to grow from a dominant base.

The more consequential finding for practitioners is buried in the measurement discussion. A case study cited in the Search Engine Land analysis documents a client whose Google Analytics 4 showed direct traffic surging 157 per cent year on year, now representing 78 per cent of all sessions. The author's assessment: virtually all of that "direct" traffic is AI-chat referral traffic with referrer headers stripped — users arriving from ChatGPT, Perplexity, or Claude responses that do not pass source attribution to the destination site. GA4's attribution model was built for a web in which traffic arrives with referrer data intact. It was not designed for a web in which a significant and growing share of traffic originates from AI systems that do not pass referrer headers. The practical implication is that brands relying solely on GA4 to measure AI search impact are operating with a structural blind spot that, in this case, reached 78 per cent of sessions.

The recommended measurement supplement is a stack of audience intelligence tools — Semrush Traffic Analytics, SparkToro, GWI, YouGov, and direct customer surveys — that can triangulate AI chat influence from the demand side rather than the referral side.

Separate Semrush research presented ahead of the Spotlight London conference in October adds a market-structure dimension to the traffic data. Analysis of 1,094 topic categories found that only 15.2 per cent had an established "category owner" in AI search results — a brand or source that AI systems consistently cite when answering queries in that category. Eighty-five per cent of categories remain unclaimed. AI search visitors, when they do arrive at a destination site, convert at 4.4 times the rate of organic search visitors — a finding consistent with the Brainlabs 335 per cent conversion uplift data reported in Issue 195.

The combination of these data sets defines the GEO market moment precisely: the traffic volumes are real and shifting rapidly, the conversion quality of AI-referred traffic is substantially higher than organic, 85 per cent of topic categories have no incumbent citation owner to displace, and the standard analytics infrastructure is unable to measure most of the action. For brands that can accept measurement uncertainty while building toward the conversion data that will eventually come through, the opportunity is the category-capture window. For brands that require complete measurement coverage before investing, the window is closing before the tools exist to see it.