Why ChatGPT referral dominance does not mean AI discovery dominance

A referral dashboard can make ChatGPT look like the clear center of AI discovery. Previsible’s AI Traffic Study found that ChatGPT generated 92.4% of measurable standalone AI referral traffic across the websites it analysed. Previsible also says Google’s AI features generate more AI-influenced traffic than all standalone LLM platforms combined. These findings answer different questions because they measure different parts of discovery.

For CEOs and CTOs, that distinction affects resource allocation. Identifiable referrals show which standalone assistants send measurable visits to a website. Google’s AI Overviews and AI Mode sit outside the standalone referral dataset used for the 92.4% figure. A broader view of AI discovery requires executives to keep these measurement categories separate.

AI referral share has a defined boundary

Previsible’s third AI Traffic Study separates standalone large language model, or LLM, referrals from AI discovery embedded in Google search. The standalone analysis covers website visits attributed to assistants such as ChatGPT, Claude, Gemini, Perplexity, and Copilot in Google Analytics 4 (GA4). ChatGPT dominates this category. Previsible separately identifies AI Overviews and AI Mode as part of a broader set of AI discovery experiences within Google.

Previsible says Google remains the main surface for AI-influenced brand discovery and that its AI experiences represent a larger volume of AI-influenced traffic than standalone LLM platforms combined. That is Previsible’s conclusion and should be treated as such. Its study uses different measurement approaches for Google AI experiences and standalone LLM referrals. A standalone-referral ranking therefore cannot rank all AI discovery.

This boundary changes how leaders should read a dashboard. Standalone referral data can identify the assistants producing measurable website sessions within the dataset. It cannot establish ChatGPT’s share of total AI-driven discovery because Google’s AI experiences fall outside that denominator. The metric is useful when the management question matches what it measures.

Referral dashboards capture a specific part of discovery

Previsible based its standalone LLM analysis on 166 Google Analytics 4 properties. The study covers sessions attributed to individual standalone assistants, allowing direct comparison of their referral volumes. Previsible excluded Google AI Overviews because, the company says, Google’s AI search features do not produce trackable referral sessions in the same way. The resulting ranking describes the standalone referral category.

This distinction also affects performance measurement. A page cited through a Google AI experience and a referral attributed to ChatGPT enter Previsible’s analysis through different measurement models. Executives should report standalone referrals separately from Previsible’s broader assessment of Google’s AI influence. Combining them into one platform ranking would treat different measures as equivalent.

David Bell, Previsible’s chief product officer and author of the report, links AI discovery to established search practices. “The foundation brands built in search matters more than ever,” Bell said. “Start by becoming a source Google’s AI results want to cite by building the site architecture, and content signals AI systems rely on to cite you, then win ChatGPT as the leading standalone surface.” Previsible sells search and AI-search services, so it benefits commercially when companies invest in the practices it recommends.

Standalone AI referrals are growing quickly

Previsible analysed 6.77 million LLM-driven sessions across 166 websites from November 2024 through May 2026. The websites span SaaS, e-commerce, finance, legal, health, insurance, education, publishing, and ticketing. Monthly standalone LLM-referred sessions increased 9.9 times over the period, from 65,249 to 644,478. The aggregate growth makes the measurable channel relevant even though individual platforms followed different paths.

ChatGPT’s share had been about 84% in Previsible’s December 2025 review. At that point, Perplexity held 8.9%, Gemini 4.5%, Copilot 2.1%, and Claude 0.6%. Claude later overtook Perplexity in monthly referral sessions in March 2026 and remained ahead through May. Previsible found Claude referrals were more common among developers, technical buyers, and professional services audiences.

The figures show why leaders need both growth rates and absolute volumes. Claude’s growth rate was much faster than Gemini’s, while Gemini’s later session count was more than twice Claude’s. Perplexity and Copilot moved in the other direction. An older platform ranking can quickly lose value as the referral mix changes.

ChatGPT itself showed substantial month-to-month movement. Its referrals fell from 448,412 in October 2025 to 213,345 in November 2025, then recovered to 442,609 in December 2025. Total monthly standalone LLM-referred sessions fell 50% that November, driven mainly by ChatGPT. One abrupt monthly movement gives executives limited evidence for a lasting resource shift.

Results also vary by industry. Previsible measured rapid growth in several sectors and much lower penetration in others. Health was the only vertical in the dataset where AI traffic penetration fell. Publishing remained at 0.08% LLM penetration against more than 120 million organic sessions.

These results put growth rates in context. A channel can grow rapidly while remaining a small share of total acquisition in a given sector. Executives need a denominator to judge whether a large growth multiple warrants more resources. The relevant scale differs by industry and website.

Page-level patterns show where referrals arrive

Platform shares show where measurable referrals originate, while page-level analysis shows where they land. Previsible recommends measuring AI traffic by page type because site-wide averages can hide strong concentrations within a website. Its findings differ sharply by business model.

For SaaS companies, the concentration on internal search pages makes those pages relevant to the AI referral journey. Education referrals concentrate on course pages, while publishing referrals concentrate on news pages. These distributions give product and marketing leaders a more precise basis for choosing which landing experiences to inspect. Site-wide averages can conceal those differences.

Health followed another pattern. Previsible linked the concentration of LLM referrals on about pages to users checking a source’s credibility after an AI assistant directed them there. In legal, referrals were spread more evenly across blog, about, contact, and location pages. The landing-page distribution is observed in the dataset; the explanation of user behavior is Previsible’s interpretation.

Financial services adds a commercial dimension. Blog content received the largest share of LLM referrals, while location and conversion pages also attracted AI traffic. Across the study, some conversion pages, including enrolment, sign-up, and product-entry pages, recorded LLM-to-total traffic ratios above 2%. Page-level measurement can identify where AI-referred visits intersect with conversion journeys.

Jordan Koene, Previsible CEO, frames the value around visitors who reach company websites. “David’s research brings focus to one of the most important, and often overlooked, parts of AI search: the value of engaged users who visit and interact with brand websites,” Koene said. He said the findings can affect how brands measure AI-driven discovery, build optimisation strategies, and improve web experiences and messaging for customers. Like Bell, Koene speaks for Previsible, which has a commercial stake in companies increasing investment in search and AI-search work.

Build priorities around the measurement boundary

Executives can manage standalone referrals and broader AI-influenced discovery as separate categories. ChatGPT’s referral dominance matters when the goal is measurable visits from standalone assistants. Previsible identifies Google’s AI Overviews and AI Mode as the main surfaces marketers should focus on in the second half of 2026. That recommendation comes from a company that sells services in this market and should remain attributed to Previsible.

Platform comparisons also benefit from several measures. Absolute sessions show current scale, growth multiples show how sources are changing, and page-level penetration shows where AI traffic matters relative to other acquisition. The November 2025 swing illustrates the risk of making a major allocation decision from one month’s movement. A longer observation period can separate a lasting shift from a short-lived change.

Previsible recommends building citation-worthy evidence, strengthening authority across trusted third-party sources, making websites easier for AI systems to read and extract, optimising for answer journeys, and measuring business impact. Bell’s recommendation connects this work to site architecture and content signals developed through established search practices. For executives, the measurable operating question is where AI visibility leads to relevant page visits and business outcomes.

Page type can then guide execution. Pricing, product, course, internal-search, and conversion pages can carry very different concentrations of AI traffic. Teams can use observed landing patterns to decide where content, navigation, credibility signals, messaging, and conversion paths deserve attention. That ties resource allocation to measured visitor behavior and the specific discovery channel being evaluated.

Key executive takeaways

• Separate referral share from total AI discovery: ChatGPT generated 92.4% of measurable standalone AI referrals in Previsible’s dataset, but that figure excludes Google’s embedded AI experiences. Leaders should keep these categories separate when allocating resources.
• Measure the discovery channel that matches the business question: Standalone referral dashboards capture attributable visits from assistants such as ChatGPT, Claude, and Gemini, not the full influence of Google AI Overviews and AI Mode. Avoid combining different measurement models into a single platform ranking.
• Track scale, growth, and industry context together: Standalone LLM referrals grew 9.9 times from November 2024 to May 2026, but platform shares shifted sharply and AI penetration varied by sector. Use longer-term trends and total traffic share before changing investment priorities.
• Use landing pages to identify where AI traffic creates value: AI referrals concentrate on different page types by industry, from SaaS internal search pages to education course pages and publishing news pages. Measure AI traffic and conversion performance at the page level rather than relying on site-wide averages.
• Build priorities around measurable business outcomes: Treat Google AI discovery and standalone LLM referrals as distinct channels, then connect each to relevant visits, conversion paths, and business results. Prioritize site architecture, credible content, and high-value landing experiences where the data supports investment.