Previsible said Google stays the most important surface for AI-influenced brand discovery, even as ChatGPT dominates measurable referrals from standalone AI assistants.
The AI discovery agency released the third edition of its AI Traffic Study, which tracks generative engine optimisation and modern search trends. The report covers data from November 2024 to May 2026 and analyses 6.77 million LLM-driven sessions across 166 web sites.
Previsible said the study relies on 166 Google Analytics 4 properties and measures referral traffic from standalone LLM platforms. The company excluded Google AI Overviews from the standalone LLM dataset, saying Google’s AI search features operate under a unique measurement model and don’t produce trackable referral sessions in the identical way.
Google leads AI discovery
The study found that AI discovery inside Google, including AI Overviews and AI Mode, represents a bigger volume of AI-influenced traffic than standalone LLM platforms combined.
The web sites included within the study span SaaS, e-commerce, finance, legal, health, insurance, education, publishing, and ticketing. The report identified Google’s AI Overviews and AI Mode as the most important surfaces marketers should deal with within the second half of 2026.
The study individually measured referral traffic from standalone AI assistants. ChatGPT accounted for many measurable LLM referral traffic in that channel.
The report also found that total monthly standalone LLM-referred sessions grew 9.9 times between November 2024 and May 2026. Monthly sessions fell in November 2025 before recovering in December.
Monthly LLM-driven sessions reached 644,478 in May 2026, up from 65,249 in November 2024.
Monthly standalone LLM-referred sessions fell 50% in November 2025, with the decline driven mainly by ChatGPT referrals. ChatGPT referrals dropped from 448,412 in October 2025 to 213,345 in November 2025, before recovering to 442,609 in December 2025.
The report didn’t attribute the November decline to a single confirmed cause.
ChatGPT dominates standalone referrals
Among standalone assistants, ChatGPT remained the biggest source of trackable referral traffic. The study found that ChatGPT accounted for 92.4% of standalone AI referral traffic across the web sites analysed.
ChatGPT’s share increased from about 84% in Previsible’s December 2025 review to 92.4% in the most recent study. At the time, Perplexity held 8.9%, Gemini 4.5%, Copilot 2.1%, and Claude 0.6%.
ChatGPT’s standalone referral traffic rose from 47,606 sessions in November 2024 to 610,910 in May 2026. That represented a 12.8 times increase over the 19-month period.
Other AI assistants also recorded referral traffic growth throughout the study period. Gemini grew 3.2 times and have become the second most visible model behind ChatGPT. Claude grew 64 times over the tracked period and moved ahead of Perplexity in March 2026.
Gemini’s trackable referral traffic rose from 5,598 sessions in November 2024 to 18,119 in May 2026.
Claude’s referral traffic rose from 133 sessions in November 2024 to eight,528 in May 2026. The report said Claude overtook Perplexity in monthly referral sessions in March 2026 and remained ahead through May.
Claude referrals were more common amongst developers, technical buyers, and skilled services audiences, based on the study. The report also found differences in AI referral patterns across industries.
The study also pointed to declines from some standalone AI referral sources. Perplexity peaked at 17,507 monthly sessions in March 2025 before falling to six,788 in May 2026. Copilot dropped from 8,651 sessions in August 2025 to 339 in May 2026.
AI traffic varies by industry
AI referral traffic to e-commerce content rose 37 times, based on the study. Product pages became the most important landing surface for e-commerce LLM traffic.
Insurance also recorded growth in AI traffic penetration, rising 18.9 times to 1.51% of total sessions. Education grew 5.4 times, while finance increased from 0.56% to 1.19%. SMB web sites rose from 0.4% to 1.71%.
Health was the one vertical within the dataset where AI traffic penetration declined, falling from 0.23% to 0.17%. The report showed that LLM traffic remained a small share of total sessions across the dataset, but concentration varied by industry and page type.
The report found that AI-referred traffic doesn’t land on the identical sorts of pages across industries. In SaaS, internal search pages accounted for 34.6% of LLM referrals, while news pages accounted for 54% of publisher referrals. In education, course pages accounted for 52% of LLM referrals.
In publishing, news pages accounted for 54% of LLM referrals within the category. The study found that LLM penetration remained low at 0.08% against greater than 120 million organic sessions.
In e-commerce, product pages were the most important landing surface for LLM traffic. The study found that e-commerce AI traffic was almost entirely driven by ChatGPT.
In health, about pages accounted for 42.1% of LLM referrals, which the report linked to users checking the credibility of a source after being directed by an AI assistant. In legal, referrals were more evenly distributed across blog, about, contact, and placement pages.
In financial services, blog content received the biggest share of LLM referrals, while some location and conversion pages also recorded LLM traffic. Some conversion pages, including enrolment, sign-up, and product-entry pages, recorded LLM-to-total traffic ratios above 2%.
Across industries, AI-referred traffic landed on internal search results pages roughly 25% of the time, based on the study.
Search foundations remain central
David Bell, Previsible’s chief product officer and creator of the report, said website structure and content signals remain essential as AI becomes a part of discovery.
“The foundation brands in-built search matters greater than ever,” Bell said. “Start by becoming a source Google’s AI results wish to cite by constructing the positioning architecture, and content signals AI systems depend on to cite you, then win ChatGPT as the leading standalone surface.”
Bell said teams that deal with AI search now will probably be higher positioned to seem in AI-generated results. Previsible really helpful five areas of focus: constructing citation-worthy evidence, improving authority across trusted third-party sources, making web sites easier for AI systems to read and extract, optimising for answer journeys, and measuring business impact as an alternative of broad visibility alone.
The report also said marketers should measure AI traffic by page type somewhat than relying only on site-wide averages. The study found that AI traffic varied by page type, including pricing, product, course, and conversion pages.
Previsible CEO Jordan Koene said the study draws attention to users who reach and interact with brand web sites through AI-driven discovery.
“David’s research brings focus to probably the most essential, and sometimes ignored, parts of AI search: the worth of engaged users who visit and interact with brand web sites,” Koene said.
Koene said the findings can affect how brands measure AI-driven discovery, construct optimisation strategies, and improve web experiences and messaging for purchasers.
(Photo by Shutter Speed)
See also: AI try-on linked to higher ecommerce conversion
Want to learn more about AI and massive data from industry leaders? Check out AI & Big Data Expo going down in Amsterdam, California, and London. The comprehensive event is an element of TechEx and is co-located with other leading technology events, click here for more information.
Marketing Tech News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.
Read the total article here
