ChatGPT Traffic Surges Back, Leaving Rivals Stuck Below 10%

TL;DR Despite billions spent on competing models throughout 2025 and 2026, OpenAI’s ChatGPT has rebounded from its mid-year plateau, holding over three-quarters of global consumer web and app queries while rivals struggle to crack double digits.

Late 2025 fueled speculation that consumer AI had reached a commoditized stalemate. Open-weights models were rapidly closing performance gaps on benchmarks, Google baked Gemini into every Android touchpoint, and Anthropic’s Claude earned high praise across software engineering circles. Pundits forecast a fragmented ecosystem where user attention would scatter across half a dozen distinct specialized interfaces.

Autumn 2026 data shows that thesis falling apart.

According to consolidated digital telemetry tracked across global desktop and mobile web visits via Similarweb web traffic datasets and app analytics, ChatGPT has sustained an aggressive autumn traffic rebound. OpenAI’s direct consumer portal now captures between 74% and 78% of global chatbot web visits. More striking than OpenAI’s sustained dominance is the ceiling suffocating everyone else: not a single competing standalone assistant—including Google Gemini, Microsoft Copilot, Anthropic’s Claude, or Perplexity—commands more than 9% of direct traffic.

analytics dashboard screen displaying consumer web traffic statistics and market share charts — Photo by Luke Chesser on Unsplash

The Illusion of the Multi-Assistant Ecosystem

The gap between benchmark parity and actual user habituation has widened into an insurmountable moat. When users want an answer, a drafted email, or a synthesized document, they open ChatGPT out of muscle memory.

The dynamics resemble the search engine dynamics of the early 2000s. While alternatives offer parity—and in specific programming workflows, Claude 3.5 and its successors frequently claim superiority—the consumer internet does not run on marginal technical wins. As developers evaluate ai models for backend infrastructure, end consumers simply stick with the name brand they registered for two years ago.

The numbers reflect an entrenched consumer habit that marketing spend has failed to unseat:

Assistant Platform	Estimated Global Traffic Share	Primary Growth Driver	Key User Churn Vulnerability
OpenAI ChatGPT	76.2%	Default brand habit, SearchGPT, Voice Mode	Subscription price fatigue
Google Gemini	8.8%	Android integrations, Workspace bundling	Brand confusion, Search cannibalization
Anthropic Claude	5.4%	Developer adoption, Artifacts workspace	Weak consumer brand recognition
Perplexity AI	4.9%	Academic research, factual citation workflows	Dependency on upstream model APIs
Microsoft Copilot	4.7%	Windows 11 tasks, Enterprise Office tie-ins	Low spontaneous consumer intent

Google has thrown immense distribution weight behind Gemini, integrating it directly into Android system notifications and Chrome search bars. Yet on standalone web destinations—where users intentionally visit a destination to solve a complex intellectual problem—Gemini hovers beneath 9%. Microsoft Copilot faces an even sharper version of this dilemma; despite native OS integration on hundreds of millions of enterprise PCs, independent visits to the Copilot portal sit below 5%.

Why Single-Digit Ceilings Persist

Three structural factors explain why competing standalone chatbots are running in place.

First is the identity layer. Hundreds of millions of consumers have established deep contextual history in ChatGPT: custom instructions, past projects, memory banks, and uploaded personal knowledge. Moving to Anthropic or Perplexity requires abandoning that accumulated personal context. For enterprise buyers covered in biz it, switching costs center on enterprise governance and SOC 2 compliance; for consumers, switching costs are pure friction and memory loss.

modern smartphone displaying a clean conversational artificial intelligence user interface — Photo by Jakub Żerdzicki on Unsplash

Second, OpenAI evolved ChatGPT from a pure prompt-and-response box into an operating layer for everyday discovery. The full deployment of real-time search capabilities and high-fidelity native voice during the past year transformed the platform from a specialized writing assistant into a general-purpose replacement for informational search. Rather than opening Google to find a recipe, a travel itinerary, or a comparison sheet, consumers increasingly query ChatGPT directly.

Third, regulatory scrutiny around default ecosystem placement has made aggressive bundling legally precarious. With global regulators and bodies like the Federal Trade Commission aggressively policing platform self-preferencing and anti-competitive default placements, platform holders cannot simply overwrite consumer choice to force their proprietary assistants onto hardware without friction.

Distribution Alone Is No Longer Enough

The traffic distribution carries sobering implications for venture-backed startups attempting to build general-purpose consumer front-ends. Thin wrapper interfaces have already suffered mass casualties, but even well-capitalized foundation model companies now recognize that acquiring consumer eyeballs directly is prohibitively expensive.

When a user acquires an assistant habit, their customer acquisition cost amortizes over thousands of queries. If an upstart has to pay twenty dollars in advertising just to coax an active user away from their established ChatGPT chat logs, the economics do not clear. Anthropic has sensibly leaned into developer APIs and enterprise tooling, leaving consumer distribution largely to platform aggregators. Perplexity continues to carve out a loyal cohort of professional power users, yet its growth curve faces friction as OpenAI enhances its own integrated browsing systems.

OpenAI still faces vulnerabilities. The computing costs required to serve hundreds of millions of non-paying active users remain astronomical, putting perpetual pressure on conversion rates to Plus and Team subscriptions. But suggestions that ChatGPT would fade into the background as an underlying utility have proven premature. Brand identity and consumer habit remain the most defensible assets in software, and for now, OpenAI owns both.