{"id":22647,"date":"2026-10-12T08:37:00","date_gmt":"2026-10-12T06:37:00","guid":{"rendered":"https:\/\/www.lukaswojcik.com\/blog\/?p=22647"},"modified":"2026-10-10T22:59:08","modified_gmt":"2026-10-10T20:59:08","slug":"ai-agents-web-analytics-agentic-traffic-ga4-server-side","status":"publish","type":"post","link":"https:\/\/www.lukaswojcik.com\/blog\/en\/digital-analytics\/ai-agents-web-analytics-agentic-traffic-ga4-server-side\/","title":{"rendered":"AI Agents in Web Analytics: How Agentic Traffic Skews GA4 and How It Can Be Separated on the Server"},"content":{"rendered":"

In 2026, AI agents have moved from experiment to measurable traffic factor \u2013 and most analytics setups simply count them as humans.<\/strong> Agents open pages, scroll, fill in forms and add products to the basket. In the GA4 report this shows up as an ordinary session with a Chrome browser, often as \u201cDirect\u201d, sometimes even as organic traffic.<\/p>\n

The scale is no longer negligible. According to HUMAN Security<\/a>, traffic from AI agents and agentic browsers grew by 7,851% year on year in 2025. In the August 2026 benchmark<\/a>, agentic traffic rose by another 27% and reached a new high.<\/p>\n

At the same time, Google introduced a dedicated \u201cAI Assistant\u201d channel in GA4 in May 2026. That sounds like a solution, but it measures only part of the picture. This article shows which kinds of AI visits exist, which of them GA4 actually recognises, and how agentic traffic can be classified on the server and carried cleanly into reporting \u2013 with concrete code examples.<\/p>\n

Three classes of AI visits<\/h2>\n

\u201cAI traffic\u201d is not a single phenomenon; technically it splits into three classes with completely different measurement behaviour.<\/strong> Lumping them together yields neither clean user figures nor a reliable statement about the value of AI as a channel.<\/p>\n\n\n\n\n\n
Class<\/th>\nExamples<\/th>\nExecutes JavaScript<\/th>\nAppears in GA4<\/th>\nDetectable via<\/th>\n<\/tr>\n
AI-referred humans<\/td>\nClick on a source in ChatGPT, Gemini, Claude, Perplexity<\/td>\nyes<\/td>\nyes, often in the \u201cAI Assistant\u201d channel<\/td>\nReferrer, UTM parameters<\/td>\n<\/tr>\n
Server-side fetches on behalf of a user<\/td>\nChatGPT-User, Claude-User, Perplexity-User<\/td>\nmostly no<\/td>\nno<\/td>\nUser-Agent, IP lists, server logs<\/td>\n<\/tr>\n
Agentic browsers<\/td>\nPerplexity Comet, ChatGPT Agent, Chrome Auto Browse, Google-Agent<\/td>\nyes<\/td>\nyes, as a normal session<\/td>\ndifficult: partly User-Agent tokens, partly signatures, partly not at all<\/td>\n<\/tr>\n<\/table>\n

On top of that come classic training and search crawlers such as GPTBot or OAI-SearchBot. These crawlers usually do not execute JavaScript and therefore only appear in logs, not in tag-based tracking.<\/p>\n

The real problem is the third class. Agentic browsers run the tracking script, set cookies and trigger events \u2013 without behaving like a human. Precisely these sessions shift conversion rates, bounce behaviour and funnel analyses without being marked as agents.<\/p>\n

The new AI Assistant channel in GA4: what it measures and what it does not<\/h2>\n

The \u201cAI Assistant\u201d channel only captures clicks by humans with a recognisable referrer \u2013 in other words, only the first of the three classes.<\/strong> According to Google\u2019s release notes of 13 May 2026<\/a>, GA4 sets the medium to \u201cai-assistant\u201d as soon as the referrer matches a known AI assistant. These visits land in the Default Channel Group as \u201cAI Assistant\u201d, and the campaign is \u201c(ai-assistant)\u201d.<\/p>\n

In addition, the Source Group<\/strong> dimension has been available since 11 June 2026. It groups source values and, according to Google, already contains groupings for ChatGPT (OpenAI) and Perplexity \u2013 populated retroactively, so it can also be used for historical periods.<\/p>\n

Three gaps remain nonetheless:<\/p>\n