AI Referral Traffic Share: ChatGPT, Gemini, Claude & Perplexity in 2026

Two headlines circulated simultaneously in 2026: "ChatGPT controls 90% of AI search" and "ChatGPT's share dropped to 63%." Both are accurate. The real issue with AI referral traffic data this year is not the numbers themselves, it is measurement methodology. Depending on which panel you use and which industry vertical you examine, the same platform's share can swing by more than 30 percentage points. This article cross-analyzes four major studies published in the first half of 2026, then draws out what Gemini, Claude, and Perplexity's shifting positions mean for GEO strategy.

What AI Referral Traffic Is

AI referral traffic refers to website sessions that originate when a generative AI platform, ChatGPT, Gemini, Claude, Perplexity, or similar, presents a link to a user and the user clicks it. The mechanics mirror traditional search referrals (google.com → site), except the HTTP referrer value stamps chatgpt.com, gemini.google.com, claude.ai, or perplexity.ai instead.

AI referral share is the fraction of all AI-originated referral sessions attributable to a specific platform. "ChatGPT share 78%" means 78 of every 100 AI-sourced referral visits came from ChatGPT, not 78% of total web traffic.

Three distinct measurement approaches produce the divergent figures in circulation. The diagram below summarizes each.

Four Studies, First Half 2026: Cross-Comparison

To make sense of the variance, each study's underlying data must be examined alongside the headline figure.

StatCounter's HTTP referrer census carries no industry bias, making it the broadest proxy for the web overall. SE Ranking's website panel skews toward content-heavy sites. Previsible's GA4 method excludes sessions where AI strips the referrer value and counts only confirmed click-throughs, which pushes ChatGPT's share higher. Goodie's B2B panel, drawn from audiences with active purchasing decisions, reflects heavier Claude and Gemini usage relative to the general web.^[1][2][3][4]

Platform-by-Platform Analysis

ChatGPT: Dominant but no longer above 80%

ChatGPT remains the single largest source of AI referral traffic by a wide margin. For the first time since StatCounter began tracking, though, its share fell below 80%: from 84.21% in April 2025 to 78.16% in March 2026, a drop of 6 percentage points.^[1] SE Ranking confirms the direction, annual average of 79.74% in 2025 declining to 74.78% in 2026.^[3]

In absolute terms, volumes are growing. Previsible's analysis of 166 GA4 properties found that LLM-attributed sessions increased 9.9x over 19 months (November 2024 to May 2026), with 92.4% of those sessions originating from ChatGPT.^[2] The platform is ceding share percentage points while its raw traffic grows, the pie is expanding faster than ChatGPT's slice.

Gemini: 3.7x in 12 months, now in second place

Gemini's climb is unambiguous in the data. StatCounter records a rise from 2.31% in April 2025 to 8.65% in March 2026, 3.7x, enough to displace Perplexity from second place.^[1] SE Ranking's panel puts Gemini at 11.56%, indicating some variation by methodology, though the direction is consistent.^[3]

The growth tracks Google's decision to embed Gemini natively into Android, Chrome, and Google Search. Users encounter Gemini through existing touchpoints without downloading anything separately, which raises the probability that a Gemini response turns into a referral click.

Claude: Fastest growth rate, with a strong B2B signal

Claude's absolute share remains small, but its trajectory stands out. Previsible's study recorded 64x growth over 19 months, with Claude passing Perplexity in March 2026.^[2] SE Ranking separately documents a 320% year-over-year increase.^[3]

The B2B dimension is where the Claude picture shifts most sharply. In Goodie's B2B brand panel, Claude holds 18.5%, second place, ahead of Gemini at 10.6%.^[4] Among developers, technical buyers, and professional services audiences, Claude's referral share runs well above the headline averages. For companies selling into technology or B2B markets, treating Claude as a minor platform understates its actual reach in those segments.

Perplexity: Share declining, absolute traffic still growing

Perplexity held second place in April 2025 at 12.07%, then fell to 7.07% by March 2026, surrendering that position to Gemini.^[1] This is a relative decline: Perplexity's absolute visitor counts continue to rise. Faster growth from Gemini and Claude compressed Perplexity's share of an expanding market, which is not the same as losing users.

Source: StatCounter (Mar 2026)^[1]

South Korea: What the Global Figures Mean Locally

The global share numbers carry direct implications for South Korea. OpenSurvey's 2026 study found that 54.5% of Korean internet users had used ChatGPT in the past three months, a 15-percentage-point jump year-over-year, and ChatGPT's app user base in Korea exceeded 20 million.^[5]

As South Korean users spend more time on AI platforms, AI-originated referral traffic to Korean sites will grow in parallel. Absolute volumes still trail traditional organic search, but AI has become a primary starting point for information-seeking, which makes AI referral a channel that warrants dedicated tracking rather than folding into "other."

B2B, professional services, and technology companies face an additional consideration: Gemini and Claude's referral weight in those segments tends to run higher than global averages suggest. Goodie's B2B panel data illustrates the point, platform mix follows audience profile, so the right optimization priority list starts with understanding which platforms your specific buyers use.^[4]

Measurement and Response: Four Practical Steps

Step 1, Check your own GA4 first. In GA4's traffic source report, filter sessions by source: chatgpt.com, gemini.google.com, claude.ai, perplexity.ai. Global averages describe the population, not your site. Your own data is the baseline.

Step 2, Separate session quality by AI channel. AI referral visitors tend to arrive deeper in the information-seeking process. Compare bounce rate, dwell time, and conversion-stage completion across channels to determine which AI platform sends the most valuable traffic to your specific site.

Step 3, Build content visibility for each platform's profile. ChatGPT referrals flow primarily from citation and mention rather than direct link clicks; Gemini responds well to structured content integrated with Google Search; Claude's referrals skew toward technical documentation and in-depth content. The GEO KPI measurement guide covers platform-specific measurement approaches.

Step 4, Establish a recurring measurement cadence. The AI referral ecosystem shifts fast. Perplexity held second place in April 2025 and dropped to third by March 2026. Revisiting platform-level share every six months is the minimum. BOIDA and similar multi-dimensional AI visibility tracking tools, or GA4 automated segments, reduce the manual overhead.

Key Takeaways

AI referral traffic share figures can differ by more than 30 percentage points depending on measurement methodology. ChatGPT's dominance is real, but Gemini's 3.7x climb and Claude's 64x growth over 19 months show the risk of optimizing for a single platform. The starting point for any practical GEO strategy is your own GA4 data and your industry's specific platform distribution, not a headline average. As AI search grows, GEO shifts from an optional experiment to a standard component of channel management.

---