Tutorial: isolating AI referral traffic in GA4 and BigQuery

On October 16, 2026, a post on DEV Community described a way to separate AI-referred traffic from everything else that lands in the Referral bucket. The approach starts in GA4, where a custom channel group is created for AI assistants such as ChatGPT, Perplexity, and Gemini, so their visits are no longer lumped in with other referring websites.

From there, the post turns to BigQuery. A query is run against the GA4 export to pull out AI-referred sessions and landing pages, which are then joined against citation data to compare what was cited with what was actually clicked. The post frames this as a follow-up question to AI visibility metrics: those metrics show how often assistants mention and cite a site, but they do not answer whether any of that turns into traffic. Part of that answer, the post notes, is already sitting in analytics data, mixed into Referral alongside every other website.

As of this report, the method stops at that point: a GA4 custom channel group to isolate the sources, a BigQuery query over the GA4 export to surface AI-referred sessions and landing pages, and a join against citation data to compare cited with clicked. The post does not include specific query statements, data volumes, or performance numbers.