ChatGPT is not merely leading the emerging market for AI referral traffic in BrightEdge’s latest dataset. It is absorbing almost the entire category.
New data released by BrightEdge on September 24 says referral traffic from ChatGPT increased 101% between January and August 2026, while ChatGPT’s share of AI-generated referrals reached a record 95.1%. In BrightEdge’s measurements, that makes ChatGPT responsible for roughly 22 times the combined referral share of Gemini, Claude and Perplexity.
The figures come from BrightEdge’s proprietary data rather than a census of the entire web, so the 95.1% figure should not be generalized into a universal market-share number. But within the websites and AI referrals BrightEdge observes, the gap has become extraordinary: one assistant is currently responsible for nearly all measurable AI-driven visits.
That creates an important distinction for marketers. The AI platform that receives attention in prompts, citations or product announcements is not necessarily the platform sending users back to websites. In BrightEdge’s current dataset, ChatGPT has separated itself dramatically on the latter metric.
ChatGPT referral traffic doubled in eight months
BrightEdge says ChatGPT referrals increased 101% from January through August 2026. That means the channel more than doubled within eight months even as the broader AI-search market continued changing rapidly.
The more striking number, however, is share. ChatGPT reached 95.1% of the AI referral traffic measured by BrightEdge, leaving only a small fraction for competing assistants.
BrightEdge describes the result as roughly 22 times the combined share of Gemini, Claude and Perplexity. The comparison is especially notable because those three products represent some of the largest alternative AI discovery environments available to users.
This is not a statement about usage, total queries, model quality or overall market share. It is specifically a measurement of referral traffic: visits that move from an AI platform to an external website and can be observed as such.
95.1% does not mean ChatGPT drives 95% of search traffic
The headline number needs an important denominator.
BrightEdge’s own broader research continues to say that AI search platforms collectively account for less than 1% of total referral traffic in its market measurements, while Google remains responsible for more than 90% of search traffic. The new 95.1% figure is therefore a share inside the AI-referral category, not a claim that ChatGPT now drives 95% of website discovery overall.
That difference is essential. A channel can dominate a small but rapidly growing category without yet approaching the absolute traffic scale of traditional organic search.
The correct interpretation is that when BrightEdge sees a measurable referral arriving from an AI assistant, the overwhelming probability in its current dataset is that the referral came from ChatGPT.
The AI market looks very different when measured by outbound clicks
AI platforms can be compared in many ways: monthly users, prompts, citations, brand mentions, answer quality, shopping recommendations or website referrals. Those metrics describe different behaviors and can produce very different competitive pictures.
A system can influence millions of decisions without sending many direct visits to publishers. Another can expose external links prominently and become a larger source of trackable traffic even with a smaller overall audience.
BrightEdge’s dataset focuses attention on that distinction. For publishers and brands, referrals are particularly concrete because they cross the boundary from an AI interface into an owned website where the business can measure engagement, leads, subscriptions and purchases.
ChatGPT’s 95.1% share suggests that, within BrightEdge’s observed universe, OpenAI’s product is currently making that handoff far more often at scale than its named AI-first competitors.
AI visibility and AI traffic are not the same KPI
The result should make GEO teams cautious about treating citation share as a proxy for traffic.
A brand can appear frequently inside an AI answer while receiving few clicks. The model may provide enough information that the user never leaves the interface, or a citation may be visually secondary to the generated response. Conversely, a smaller number of highly actionable references can produce meaningful visits.
That means organizations increasingly need at least two separate measurement layers: whether their content or brand appears in AI responses, and whether those appearances generate measurable visits to their own properties.
BrightEdge’s new numbers describe the second layer. They do not tell us that 95.1% of AI citations come from ChatGPT, nor that ChatGPT controls 95.1% of AI usage.
ChatGPT is becoming the AI referral channel marketers can actually see
For analytics teams, a referral is valuable partly because it is observable. A visitor reaches the site, creates a session and can potentially be connected to downstream behavior.
As ChatGPT’s share of AI referrals rises, it becomes increasingly important to separate that traffic cleanly inside analytics rather than burying it inside a generic referral bucket. Landing pages, engagement patterns and conversion paths can then be compared with traditional organic search, direct traffic and other AI assistants.
The 101% growth rate also means historical baselines can become obsolete quickly. A company that concluded earlier in 2026 that AI referrals were negligible may see a materially different picture by the end of the year.
That does not justify shifting budgets blindly from SEO into ChatGPT optimization. It does justify measuring the channel continuously rather than relying on assumptions formed when AI referral volumes were much smaller.
Google can dominate search while ChatGPT dominates AI referrals
There is no contradiction between Google remaining overwhelmingly dominant in traditional search traffic and ChatGPT accounting for almost all referrals inside BrightEdge’s AI-assistant category.
The two numbers describe different markets. Google’s search ecosystem operates at vastly greater absolute scale. ChatGPT is dominating a newer slice of traffic generated when users interact with conversational AI products and then follow links to the open web.
This distinction matters because marketers can otherwise overreact to a percentage that sounds much larger than its denominator. A 95.1% share of AI referrals can coexist with AI referrals representing less than 1% of overall referral traffic in BrightEdge’s broader measurements.
The strategic implication is not that Google has been displaced. It is that ChatGPT has emerged as the clearest leader inside a new referral channel that is growing from a much smaller base.
The 22× gap raises questions for Gemini, Claude and Perplexity
Gemini, Claude and Perplexity are all capable of web-connected research and external sourcing, yet BrightEdge says their combined AI referral share is dwarfed by ChatGPT’s.
The dataset by itself cannot establish why. Differences in audience size, interface design, citation presentation, user intent, link behavior and measurement can all affect referral volume.
It would therefore be speculative to conclude that users trust one engine 22 times more, or that one engine is 22 times more influential. Referral share measures a specific observable action, not the entire discovery process.
What marketers can say is narrower and still significant: BrightEdge currently observes far more website visits originating from ChatGPT than from those three competing assistants combined.
A click can understate the influence that happened before it
Referral analytics also have a structural blind spot. AI systems can influence a purchase or brand decision without receiving credit for the final visit.
A user might research products in ChatGPT, remember a brand and later type its URL directly. Another might move from an AI conversation to Google and then click an organic result. In both cases the AI assistant affected the journey but may disappear from conventional last-click attribution.
BrightEdge has previously characterized AI search as heavily oriented toward research and discovery, with users often converting later through organic search or direct channels. That makes referral traffic a useful but incomplete measure of AI influence.
The 95.1% result should therefore be read as dominance in observable AI referrals, not a complete attribution model for AI-assisted customer journeys.
For publishers, the open-web question is becoming measurable
One of the central tensions surrounding generative AI is whether assistants will replace website visits or create a new discovery layer that sends valuable traffic outward.
BrightEdge’s data shows that at least one AI assistant is producing a rapidly growing stream of measurable referrals. ChatGPT traffic more than doubling between January and August is evidence that outbound discovery has not disappeared as conversational interfaces become more capable.
But the scale question remains. AI referrals are still small compared with organic search in BrightEdge’s broader datasets, and publishers cannot assume that a citation will produce a click.
The opportunity is therefore real but uneven: AI can become a referral source while simultaneously answering many questions without a website visit.
GEO strategy may need to become platform-weighted
The early phase of generative engine optimization often treated ChatGPT, Gemini, Claude and Perplexity as a relatively flat list of engines that brands should monitor equally.
BrightEdge’s referral data argues for a more nuanced approach. If the business objective is measurable website traffic rather than visibility alone, platforms can have radically different practical value at a given moment.
That does not mean brands should ignore Gemini, Claude or Perplexity. Each can influence awareness and customer decisions, and referral behavior can change rapidly as interfaces and user bases evolve. A platform with little outbound traffic today can alter link presentation or grow substantially tomorrow.
But resource allocation should reflect the KPI being optimized. A team seeking AI mentions may use one platform mix; a team seeking trackable referral sessions may currently weight ChatGPT much more heavily.
The proprietary dataset is a signal, not a census
BrightEdge’s figures are valuable precisely because large-scale AI referral data is difficult to obtain, but they need the same methodological caution applied to any proprietary measurement system.
The dataset reflects the sites, customers, tracking methods and classifications available to BrightEdge. It does not observe every website on the internet, every ChatGPT user or every interaction with Gemini, Claude and Perplexity.
Different analytics providers can therefore produce different shares depending on their customer mix and methodology. Industries may also vary sharply: an enterprise software site, ecommerce retailer and news publisher can have very different AI referral profiles.
The strongest use of the 95.1% figure is as evidence of an extreme pattern inside BrightEdge’s measured environment, not as a universal law of web traffic.
The bigger story is concentration
The 101% growth rate demonstrates momentum, but the 95.1% share reveals something more strategically important: concentration.
A new discovery channel is emerging, yet its measurable outbound traffic is not currently distributed evenly across the leading AI assistants in BrightEdge’s data. It is concentrated overwhelmingly in ChatGPT.
That concentration creates opportunity and risk. Brands that gain visibility in ChatGPT can potentially access the dominant source of AI referrals observed by BrightEdge. At the same time, an ecosystem dependent on one assistant can change abruptly if product design, citation behavior or user habits shift.
Marketers should therefore exploit the current opportunity without confusing today’s distribution with a permanent market structure.
ChatGPT has become the benchmark for AI traffic, not a replacement for Google
The cleanest conclusion from BrightEdge’s new dataset is not that AI search has overtaken traditional search. It has not, according to BrightEdge’s own broader traffic measurements.
The change is inside the AI category itself. ChatGPT referral traffic more than doubled from January through August, and its share reached 95.1% of the AI referrals BrightEdge observed. Gemini, Claude and Perplexity combined were roughly 22 times smaller by this measure.
For SEO and GEO teams, that turns ChatGPT into the immediate benchmark for a specific question: can generative AI send measurable traffic back to the open web?
BrightEdge’s answer, in its own dataset, is increasingly yes—and almost all of that measurable AI referral growth is currently arriving through one door.