ChatGPT is extending from a conversational tool into a web discovery entry point. After receiving an AI answer, users may continue to visit news, product, travel, or traditional search websites. This report compiles public sources including OpenAI, Similarweb, Semrush, Digiday, Microsoft Clarity, and Pew Research Center to observe the growth, destinations, industry differences, and conversion performance of AI web referral traffic from 2025 to 2026. The samples, statistical periods, and traffic definitions used by different organizations are not the same; the relevant data is used to observe trends separately and is not directly combined for calculation.
Research Background
In October 2024, OpenAI launched ChatGPT Search, combining real-time web search with conversational interaction and providing accessible source links within answers. According to OpenAI's official announcement, users can ask questions in natural language and continue to obtain web information through follow-up queries. This establishes a more direct access relationship between AI answers and external web pages.
Traditional search typically guides users to websites through result lists, whereas AI search can first integrate information and then provide some source links. The two access methods may occur consecutively, or they may separately fulfill different information needs.
Therefore, to assess the impact of AI search on websites, three questions need to be distinguished: how much traffic the AI platform itself receives, how much traffic the AI platform sends to external websites, and what behavior these visits generate after entering the website. We focus on the latter two questions, while distinguishing brand mentions, web citations, and actual clicks.
ChatGPT's outbound referral traffic is growing, but platform visits and external referrals are not synchronized
The traffic ChatGPT sends to external websites is growing, but this growth cannot be directly equated with changes in ChatGPT platform visits.
According to Semrush's "ChatGPT Search Insights" analysis of 17 months of clickstream data, ChatGPT's referral traffic to external websites grew by 206% in 2025. This means that in addition to obtaining information within the conversational interface, more and more users recorded in the research sample continue to visit external web pages.
Similarweb's "AI Referral Traffic by Industry" provides another perspective. According to its industry statistics from June 2025 to May 2026, global AI platforms brought an average of approximately 770.7 million referral visits to external websites per month, a year-over-year increase of 117.4%. This metric covers multiple AI platforms and is not the visit volume of ChatGPT alone.
Similar changes have also been observed in the news and publishing sector. According to a 2025 Digiday report, ChatGPT sent approximately 243.8 million visits to the 250 news and media websites covered by the research in April of that year, an increase of about 98% from 123.2 million in January. This data reflects changes within a specific media sample and cannot directly represent all websites.
The three sources together present the phenomenon of increasing AI referral visits, but their statistical subjects differ. Semrush focuses on ChatGPT's outbound clickstream, Similarweb observes industry referrals across multiple AI platforms, and Digiday reports visit changes at specific media websites.
This distinction is important. ChatGPT platform visits measure the scale of users entering the AI service; outbound referral traffic measures whether users continue to visit external websites. A single conversation may have no external clicks, or it may involve multiple sources. Therefore, the actual visits a website receives cannot be inferred from the platform's user scale.
From public research, ChatGPT has already taken on part of the web discovery function. But traffic growth does not mean that all AI answers will bring clicks, nor can it directly indicate that these clicks ultimately produced commercial conversions. The next step is to observe which websites users specifically enter after leaving the AI conversation.
ChatGPT's referral traffic is not evenly dispersed, but concentrated in certain websites and industries
ChatGPT's outbound referral traffic shows a concentrated distribution pattern, with certain traditional search websites and content websites with clear follow-up visit needs receiving more traffic.
Here, "referral destination" refers to the visit destination after users enter external websites through links provided by the AI platform. It is not the same metric as a brand name appearing in an AI answer or a web page being listed as a reference source.
According to Semrush's 17-month clickstream study, the top 10 domains received more than 30% of ChatGPT referral traffic. Among them, Google received approximately 21.6% of outbound referral traffic.
This result shows that AI search and traditional search do not necessarily constitute a mutually exclusive relationship. Some users may first use ChatGPT to organize questions, understand concepts, or get suggestions, and then enter Google to continue searching. In this case, AI conversation assumes the function of preliminary information organization, while traditional search assumes the function of further finding web pages.
At the same time, websites with clear follow-up action needs also appear among the main destinations of AI referral traffic.
According to Similarweb's industry statistics from June 2025 to May 2026, online general marketplaces received an average of approximately 46.8 million AI referral visits per month, while the news and travel industries each received approximately 44.5 million.
The common characteristic of these industries is that after obtaining preliminary answers, users may still need to enter websites to complete other tasks. For example, product information may require further comparison of prices, specifications, and purchase conditions; news content may require reading the full report; travel plans may involve verifying flight, accommodation, date, and booking information.
These behavioral needs provide an explanation for web visits after AI answers, but it cannot be concluded that all users will act in the same order.
It should also be noted that Semrush's Google domain share and Similarweb's industry visit volume come from different studies. The former describes the distribution of ChatGPT traffic among domains, while the latter describes the industry visit scale brought by multiple AI platforms. The two cannot be directly added together or used for a unified ranking.
Overall, AI conversation has not eliminated users' need to continue visiting web pages, but in some scenarios has changed the path by which users reach web pages.
The scale and growth rate of industry referral traffic vary, with beauty, fashion, and other industries growing rapidly
The industries with larger AI referral traffic scale and those with faster year-over-year growth do not completely overlap. When observing industry changes, it is necessary to distinguish both total visits and growth rate.
According to Similarweb's analysis from June 2025 to May 2026, online general marketplaces received an average of approximately 46.8 million AI referral visits per month, a year-over-year increase of 237.3%. The news and travel industries each received an average of approximately 44.5 million visits per month.
By contrast, the beauty and fashion industries stood out in terms of year-over-year growth rate. The Similarweb report shows that AI referral traffic for the beauty industry grew by 312.5% year over year, while the fashion industry grew by 278.2%.
| Industry | Average Monthly AI Referral Visits | Year-over-Year Growth Rate |
|---|---|---|
| Online General Marketplaces | 46.8 million | 237.3% |
| News | 44.5 million | — |
| Travel | 44.5 million | — |
| Beauty | — | 312.5% |
| Fashion | — | 278.2% |
Data source: Similarweb "AI Referral Traffic by Industry," statistical period from June 2025 to May 2026. "—" indicates that the original report framework and outline did not provide a verifiable corresponding value, and does not mean that the industry had no traffic or growth.
This set of data reflects two different types of industry changes. Online general marketplaces have both a relatively large referral visit scale and a relatively high year-over-year growth rate; beauty and fashion stand out in the disclosed growth indicators, but lack average monthly visit volumes that can be used for direct comparison.
From the perspective of user needs, beauty and fashion involve product selection, usage scenarios, style preferences, and product comparison. AI conversation can help users narrow down the information scope, but product details, inventory, prices, and transaction conditions may still need to be confirmed on external websites.
Similarweb also observed that among the citation links displayed in ChatGPT beauty-related conversations, 54.7% point to retail and e-commerce websites. This shows that retail websites have a certain presence among the information sources for related AI answers.
However, the industry distribution of citation links does not equal the industry distribution of actual clicks. The fact that a certain type of website receives more citations does not mean users will definitely click these links, nor can it directly infer order growth in that industry.
News and travel scenarios also have follow-up visit needs. News users may need to verify the original report, publication time, and background information; travel users may need to further confirm booking conditions, service details, and real-time prices.
Therefore, differences in AI referrals across industries are not only related to the scale of platform usage, but may also be related to the depth of information and operational steps required for users to complete follow-up tasks. Current public data can help identify these differences, but is still insufficient to directly predict the traffic performance of an individual website.
AI referral traffic still accounts for a limited overall share, but some studies have observed different conversion performance
The rapid growth of AI referral traffic does not mean it already occupies a large proportion of overall website traffic. At the same time, some website samples have observed different conversion performance between visitors from AI sources and visitors from traditional search sources.
According to Conductor data cited in Digiday's 2025 "In Graphic Detail: The State of AI Referral Traffic in 2025," among the 10 major industries covered by the research, AI platforms contributed an average of about 1% of overall web traffic.
This proportion shows that within the scope of that study, AI referrals remain a relatively limited component of website traffic structure. Even with a high year-over-year growth rate, it cannot be directly concluded that AI traffic has replaced traditional search traffic.
But visit scale is only one dimension. Whether users register, subscribe, or continue browsing after entering the website also affects the assessment of traffic value.
Microsoft Clarity's 2025 AI traffic study analyzed visit data from more than 1,200 news and publishing websites. The study observed that LLM-source visitors and traditional search-source