ChatGPT up 48% as Bing fell 50% in US traffic, Semrush data shows

ChatGPT up, Bing down

ChatGPT.com recorded roughly 1.09 billion monthly US visits in July 2026, a 48.38% jump over the same month in 2025, according to Semrush Traffic Analytics data examined in a dev.to analysis. Bing.com moved in the opposite direction over the identical window, shedding about 50.43% of its US traffic. ChatGPT now sits ninth among the most-visited sites in Semrush's US ranking.

The numbers come from Semrush's US Trending Websites dataset, which compares July 2026 against July 2025 for ranked sites. It is not a tally of every search query or AI conversation, but the magnitude of the change turns AI-assisted discovery into an operational measurement problem rather than a distant trend to watch.

Search giants still dominate

The same snapshot shows how far ChatGPT still has to go. Google.com drew about 25.31 billion monthly US visits and YouTube.com about 10.27 billion, dwarfing every other property in the dataset. The dev.to analysis is careful on this point: the figures describe a redistribution of attention across discovery surfaces, not the collapse of conventional search. Google and YouTube remain the largest destinations by a wide margin.

Where AI visits go missing in analytics

The core of the analysis concerns what happens after someone leaves an AI assistant. A person can ask ChatGPT for product options, click a recommended link and land on a site with the originating referrer stripped. Google Analytics 4 then tends to bucket that session as Direct, or leave it otherwise unclassified.

GA4 does offer an AI Assistant channel that consolidates sources such as ChatGPT, Gemini, DeepSeek, Copilot and Grok, but it only captures visits where referral data survives the transition from the AI interface to the website. The gap cuts both ways for decision-making. Teams watching Direct traffic climb may have no signal that AI assistants are upstream, and may underestimate how often their pages surface in answers. At the same time, a traffic bump alone cannot prove that an AI platform caused it, which complicates judgments about which content earns visibility and which channels deliver visitors worth having.

Practical steps the analysis suggests

The dev.to piece recommends a layered approach rather than reliance on default acquisition reports:

• Treat unexplained Direct traffic as an open question, not an assumption of typed URLs or bookmarks.
• Adjust GA4 channel groupings so identifiable AI referral sources land in the right bucket.
• Cross-reference analytics with outside audience and traffic datasets, including Semrush Traffic Analytics, SparkToro and GWI.
• Keep observed referrals and inferences separate: a logged ChatGPT referral is evidence of a source, while rising Direct traffic only signals a possible measurement blind spot.
• Read landing pages and business outcomes alongside acquisition labels so reporting does not rest on a single imperfect field.

None of this recovers lost referrer data. It does reduce the odds that a default channel label gets mistaken for a complete account of how a visitor arrived.

Why it matters

Visibility and attribution are becoming two separate problems. A business can be recommended inside AI answers repeatedly and still watch those sessions surface as unlabelled Direct traffic in its reports, which distorts budget decisions, content priorities and channel bets. Bing's steep decline also raises questions for teams that treated it as a meaningful secondary search channel. The Semrush comparison is a single US snapshot of ranked sites, and Google's lead remains enormous, but the direction of travel is clear: discovery increasingly begins in conversational interfaces, and analytics setups built for a search-click-referrer model will keep undercounting it until they are adapted.