AI discovery is growing, but it is neither evenly distributed nor reliably predictable. Previsible’s 2026 State of AI Discovery Report analyzed 6.77 million AI-driven sessions across 166 Google Analytics 4 properties from November 2024 through May 2026. Its central finding is a practical warning for marketers: ChatGPT dominates trackable standalone LLM referrals, yet Google’s AI surfaces appear to drive more AI-influenced traffic overall, and platform changes can sharply alter referral volume in a single month. According to Previsible’s 2026 State of AI Discovery Report, ChatGPT accounted for 92.4% of trackable referral sessions from standalone LLMs. Monthly AI-referred sessions reached 644,478 in May 2026, a 9.9x increase over the study’s 19-month period. That is a substantial shift in how people find information, products, and websites. It is not, however, a reason to build a marketing plan around one AI platform or one referral channel. The more important conclusion is that AI discovery has become a distributed acquisition environment. A business can be surfaced by ChatGPT, encountered through Google AI Overviews or AI Mode, and reached via an internal search page on its own site. Each route may be measured differently, and each can change as AI products evolve. AI referral growth does not equal a stable channel ChatGPT’s share of trackable standalone LLM referral traffic makes it an important source to monitor. But the report also documents how exposed businesses can be when a platform changes its product or model behavior. Between October 2025 and November 2025, ChatGPT referrals in the study fell from 448,412 to 213,345, or roughly half in one month. That decline is a reminder that referral traffic from AI assistants is not equivalent to traffic from a mature, fully controllable marketing channel. A change in how an AI product retrieves, cites, links, or presents answers can affect outbound clicks without any corresponding change to a company’s website, content quality, or technical SEO. The report also points to a changing competitive landscape among standalone AI platforms. Claude and Gemini rose, while Copilot referrals fell about 96% from their peak and Perplexity declined. These movements reinforce two points: traffic is concentrating around a small number of platforms, and the relative importance of those platforms can change quickly. AI discovery signal in the study What Previsible found Practical implication Standalone LLM referrals ChatGPT represented 92.4% of trackable referral sessions from standalone LLMs. ChatGPT deserves attention, but concentration creates dependency risk. Google AI surfaces AI Overviews and AI Mode collectively drove more AI-influenced traffic than all standalone LLMs combined. Referral reports alone may understate the role of AI in discovery. Internal search pages 34.2% of AI traffic landed on search pages. On-site search experiences can become meaningful acquisition surfaces. ChatGPT referral volatility Referrals fell from 448,412 in October 2025 to 213,345 in November 2025. Do not treat a short period of AI referral growth as a dependable baseline. The study does not suggest that businesses should disregard ChatGPT referrals. Instead, it shows why a single referrer is a weak proxy for total AI visibility. Google’s AI experiences are especially significant because they can influence discovery at a scale larger than the standalone LLM referral data, even when traditional referral attribution does not fully capture that influence. What businesses should change in measurement and content planning The first task is to separate AI referral traffic from the broader question of AI-influenced discovery. Referral data can identify some visits sent by standalone assistants. It cannot, by itself, explain every user journey that began with an AI-generated result inside Google’s search environment. That distinction should affect reporting. Rather than treating AI traffic as one stable acquisition bucket, teams can review it alongside organic search performance, landing-page behavior, branded demand, conversions, and on-site search activity. A sudden change in LLM referrals may matter, but it should be interpreted in context before budgets or content priorities are changed. The report’s finding on internal search pages is also unusually useful. If more than a third of observed AI traffic lands on search pages, website owners need to understand what visitors encounter there. A search results page that is slow, poorly filtered, empty, or disconnected from relevant conversion paths can waste valuable demand. Conversely, a clear internal search experience can help users move from an AI-driven query to the information, product, or service they need. A practical response includes: Track AI referral sources separately where analytics platforms identify them, while avoiding decisions based on one platform’s monthly swings. Review search-page landing behavior, including the queries, result quality, engagement, and next steps available to visitors. Keep core content useful and specific so it can serve users arriving from Google search, AI assistants, and direct visits alike. Use diversified measurement that considers conversions and user behavior, not only visits attributed to an AI referrer. Avoid overcommitting spend to one AI source until performance is sustained and connected to meaningful business outcomes. For marketers, the underlying strategic shift is straightforward. Visibility in AI systems is increasingly important, but it is not a replacement for sound search, website, and analytics fundamentals. The businesses best positioned to learn from AI discovery will be those that can identify where users land, what they seek, and whether those visits lead to useful outcomes. AI-driven discovery can create new demand, but volatile referrals and opaque attribution make reactive decisions costly. Scalevise helps businesses assess where their brand appears across AI search experiences, identify gaps in content and measurement, and turn findings into a focused visibility plan. Use the Scalevise AI Visibility and GEO Checker to understand your current AI search presence before allocating more budget or rebuilding your content strategy. Start an AI visibility scan today. Frequently Asked Questions What share of standalone LLM referral traffic came from ChatGPT? Previsible found that ChatGPT accounted for 92.4% of trackable referral sessions from standalone LLMs in its analysis of 6.77 million AI-driven sessions. Why did ChatGPT referrals fall sharply in late 2025? The study recorded a drop from 448,412 referrals in October 2025 to 213,345 in November 2025. It uses the decline to show how a product or model change can quickly affect AI referral traffic. Do Google AI Overviews and AI Mode matter more than standalone AI assistants? In Previsible’s study, Google AI Overviews and AI Mode collectively drove more AI-influenced traffic than all standalone LLMs combined. This means AI discovery extends beyond trackable referrals from tools such as ChatGPT. Why are internal search pages important for AI traffic? Previsible found that 34.2% of AI traffic landed on search pages. This makes internal search results an important part of the visitor experience for people arriving through AI-driven discovery. Conclusion Previsible’s findings show that AI discovery is becoming a meaningful but uneven source of website traffic. ChatGPT currently dominates trackable standalone LLM referrals, while Google’s AI surfaces have a broader role and internal search pages absorb a substantial share of AI visits. The sharp month-to-month ChatGPT decline is the clearest reason to measure AI traffic carefully, improve landing experiences, and avoid treating any single AI platform as a dependable acquisition channel.