ChatGPT US traffic up 48% year over year as Bing referrals halve, Semrush data shows

What the Semrush data shows

ChatGPT has become a substantially larger entry point for web activity in the United States. Semrush Traffic Analytics figures for July 2026, reported by dev.to, estimate that chatgpt.com drew about 1.09 billion US visits that month, up 48.38% from a year earlier.

The same dataset points to movement elsewhere in search. Referral activity attributed to bing.com dropped by roughly half, about 50.43%, while Google's direct numbers climbed about 10.72%. Google still held its position as the largest referrer overall, so the absolute gap remains wide even as the growth rates diverge sharply.

There are caveats worth stating plainly. Semrush builds these numbers from clickstream data across desktop and mobile, using a panel it says covers around 200 million anonymized users, so these are third-party estimates rather than server logs from OpenAI, Google or Microsoft. The comparison also mixes different metrics — visits to chatgpt.com on one side, referral signals for bing.com and Google on the other — so it is best read as a directional contrast, not a like-for-like market-share table.

Growth, but concentrated

Broader Semrush research adds context. An April 2026 study found that outbound referral traffic from ChatGPT grew 206% during 2025. In its July 2026 analysis, roughly 21.6% of ChatGPT referrals went to Google, and more than 30% of referrals went to just the top 10 domains. That skew suggests AI-driven referral traffic may flow disproportionately to a small set of large destinations rather than spreading evenly across the web — a detail that matters for smaller publishers hoping to ride the trend.

Chat journeys often loop back to search

The numbers do not add up to search being displaced. A common pattern is a user asking a conversational question, receiving an answer with links or suggested searches, and then continuing the journey on Google or another site. Semrush's own referral findings support this, showing that ChatGPT sends meaningful traffic toward Google. For site owners, the practical question is less about choosing between AI platforms and search engines, and more about whether their content can be found and cited in both environments.

Attribution is the harder problem

The dev.to report frames measurement as the more immediate headache for marketers. Visits that begin inside an AI chat interface do not always arrive at a website as a clean, consistent referral source. Depending on how the originating surface hands the session to a browser, GA4 may record it as direct traffic, an ambiguous referral or something else entirely. A low or missing AI referral count is therefore not evidence that AI discovery is irrelevant; the visitor may have followed a link, pasted a URL or run a follow-up search that obscures the original interaction.

The report suggests layering several signals instead of trusting one view: review GA4 acquisition reports for identifiable referral sources and shifts in direct traffic, benchmark against third-party panel data while treating it as directional, track branded search interest and landing-page performance alongside referral labels, and simply ask customers how they found the business.

Why it matters

The growth differential is the headline: ChatGPT up 48% year over year while Bing referrals halve and Google grows in the low double digits. AI chat has moved from novelty to a genuine discovery layer, even though Google remains dominant in absolute terms. The quieter consequence is measurement — analytics stacks designed for a search-and-social web will undercount journeys that pass through chat interfaces. Teams that optimize and measure only for conventional SEO risk misreading where their audience actually comes from, and the sensible response is content that answers specific questions directly, plus reporting that accepts attribution will be incomplete.