AI Referral Traffic

AI referral traffic is the traffic that reaches your site from AI assistants and AI search products, such as ChatGPT, Perplexity, Gemini, Claude or Copilot, and that analytics can attribute to them, usually through a referrer or a UTM parameter. It is the visible part of the traffic AI sends you. MLforSEO calls it “the detectable floor, not the ceiling”, because much AI-influenced traffic arrives with no label at all.

How it works

When a user clicks a link in an AI answer, the browser may pass a referrer (for example perplexity.ai), and some products append a parameter such as utm_source=chatgpt.com, which MLforSEO’s analysis notes ChatGPT began adding in June 2025. Analytics tools then record the session as a referral from that source. Many clicks pass nothing, especially from mobile apps, and land in direct; MLforSEO calls this dark AI traffic.

Clicks from Google’s AI Overviews and AI Mode are a special case: Google reports them as part of normal search traffic in the Search Console Performance report (“Web” search type), so they appear as Google organic, not as a separate AI source.

How it’s used in SEO and marketing

• Baseline: in GA4, create a custom channel group (or exploration segment) with a regex on session source that matches known assistants, such as chatgpt.com, openai, perplexity, gemini, claude, copilot.
• Landing pages: see which pages AI systems actually send people to, and compare them with the pages you want cited.
• Quality: compare engagement and conversion rates with other channels; MLforSEO’s posts report that AI-referred visitors often arrive further along in their decision.
• Reporting: combine with prompt tracking, server-log crawl data and the direct-versus-branded-search gap for a fuller picture in Looker Studio.

Example

A publisher’s GA4 shows AI referrals at well under 1% of sessions. After adding an AI channel group, it sees that Perplexity and ChatGPT visits land mostly on three comparison guides and convert better than average. It also notices direct traffic to those guides growing faster than branded search, a hint that more AI-driven visits are arriving unlabelled.

Limitations and things to watch

• It undercounts: no referrer means no attribution, and referrer behaviour changes without notice.
• Regex lists need maintenance as new assistants and domains appear.
• AI Overview and AI Mode clicks cannot be separated from other Google organic clicks in standard reports.
• Low volumes make percentage changes noisy; look at trends over months.

Learn it on MLforSEO

• Dark AI traffic: why 70% of your AI influence hides as direct
• Your rank tracker is lying to you about AI search
• Blog and Site Architecture for SEO and Conversions in the AI Search Era
• 7 Alternative Signals of AI Influence
• AI Search Optimisation & Agentic SEO (course)