Full Answer

GA4 measures engagement through session duration, pages per session, and bounce rate. Every one of those metrics assumes a human browsing pattern: longer sessions signal interest, more pages signal intent, and a single-page exit signals disengagement. AI agents break that assumption completely. An agent that loads a product page, reads the structured data, confirms price and availability, and completes a purchase in under ten seconds looks like the worst possible visitor in your analytics — while being your best customer.

The distortion compounds as agent traffic grows. Site-wide bounce rate climbs even as conversion revenue holds steady or increases, because agent sessions are short and shallow by design. Pages-per-session averages drop, making it look like your content is underperforming when the reality is that a growing share of visitors doesn't need to browse. Teams optimising for engagement metrics start changing pages that are already converting — fixing what isn't broken, guided by numbers that no longer mean what they used to.

The only reliable fix is cohort separation before the data reaches your reporting. Tagging agent sessions at the server lets you split your GA4 reports into human and agent views, restoring the interpretive value of every engagement metric. For more on the gap between GA4's count and your actual AI visitor volume, see GA4 Says 66 AI Visitors, Your Server Logs Say 680.