AI sources are human sessions that arrived from an AI product. They are not the crawler reads on Bots & AI crawlers.
ConsoleWeb analyticsSourcesWhat the report shows
Rows are first grouped by channel type. Rules look at UTM, the referring host, and click IDs. Rules you add for the site win over the defaults. By default, visits with gclid, msclkid, wbraid, gbraid, yclid, or bd_vid are paid search. A referring host that is an AI product, or a base channel of AI, is an AI assistant. A hop from your own site is internal. No source at all is direct.
The tabs below are Channels, Referrers, Search engines, AI sources, UTM, and Search terms. UTM splits into source, medium, campaign, term, and content. Each row has visitors, sessions, pageviews, bounce, and duration. The channel table also has conversions. Ad click IDs are a separate table: type and hash only. The raw click ID is not shown. Search terms are split into organic terms, paid terms (utm_term), and on-site search terms.
How to use it
Click a row to add that channel, source, or click-ID type as a filter. Other reports narrow with it. See Filters, compare, share & export. Channel rules, and recomputing history, are under ConsoleConfigurationWeb settings.
Limits
Organic search terms exist only when the referrer actually carried the keyword. Common browsers do not pass search terms to a third party, so this column is often thin. That is expected. Full queries need a connected search account. See SEO & content decay. Paid terms and on-site search parameters are captured in full.
FAQ
If my site links to itself, does that become a referral? No. Those hops are internal. With no external source they fall through to direct, so an in-site click is not treated as acquisition.
Why is AI sources zero while the bots page shows crawls? A crawl is a program reading the page. AI sources counts people sent back, and the two need not move together. With only the page script, crawlers that do not run JavaScript are invisible. See Server events & log import.