AI Search Fragmentation Is Replacing the ChatGPT-Only Era

AI search fragmentation is turning a once ChatGPT-dominated market into a multi-player discovery layer. Similarweb estimates that generative AI websites reached 9.5 billion monthly visits between June 2025 and May 2026, up 70% year over year, while app downloads reached 4.4 billion. The practical takeaway for AI builders is simple: users, referrals and monetization are no longer concentrated in one assistant.

AI search is growing, but ChatGPT no longer owns the journey

ChatGPT remains the largest standalone AI website, yet its share of web visits is falling as Gemini, Claude, Perplexity and Grok build their own audiences. In the US, reported monthly active user growth was strongest for Meta AI at 435% and Claude at 349%, followed by Grok at 117%, Perplexity at 94% and ChatGPT at 87%. Gemini added 31%, while Microsoft 365 Copilot and DeepSeek declined 31% and 23%, respectively. These are Similarweb traffic estimates, not a complete measure of model usage.

This is the core of the Generative AI search landscape: the market is expanding while attention fragments across products with different distribution, model access, interfaces and integrations. A team building an AI workflow should assume that its users may arrive through several assistants, search surfaces or embedded experiences.

Google Overviews makes discovery even more hybrid

Google is pushing AI-native search into its existing product. Similarweb estimates that Google Overviews appear on nearly four in ten US searches. That can pull more research into an answer layer before a user visits a website, putting additional pressure on publishers and product pages that depend on conventional search clicks.

But this is not a clean replacement of Google by chatbots. Some 95% of ChatGPT users also use Google, pointing to a hybrid discovery habit. Users may ask an assistant for synthesis, return to Google for verification, then visit a product site directly. Builders should measure those paths separately instead of treating AI traffic as a single acquisition channel.

AI ads in chats change the product surface

Advertising is moving into conversational interfaces. The share of US ChatGPT chats carrying an ad rose from 14% to 26% in one month, while the reported click-through rate was 0.50%. Early advertisers included Resume.io, Monday.com and Framer. The figures suggest that AI ads in chats are entering an early monetization phase, but they do not yet show whether users will consider the format useful, intrusive or trustworthy at scale.

For product teams, ads are more than a marketing issue. They affect answer ranking, interface clarity, disclosure requirements and user trust. Teams integrating assistants into workflows should define how sponsored results are separated from model recommendations, especially when the user is making a purchase or selecting a business tool.

What builders should change now

AI referral traffic is rising sharply in several categories: 312% for marketplaces, 278% for news and 237% for travel year over year. Yet 58.8% of AI-referred visits land on homepages, even when the assistant cited a page several folders deeper. That mismatch creates an onboarding problem: a visitor can arrive with a specific expectation and still be dropped into a generic entry point.

Product and marketing teams should make the homepage resilient to that context. Explain the core product quickly, link to high-intent workflows, and preserve clear routes to documentation, pricing and the exact task an AI assistant may have recommended. Track referrals from ChatGPT, Gemini, Perplexity and other sources where