Behavioral Modeling

Behavioral modeling is GA4's use of machine learning to estimate the users and sessions it could not observe directly because those visitors declined consent.

When Consent Mode is active and a share of your visitors reject cookies, you lose the ability to identify them. Rather than leave a hole, GA4 studies the behavior of consenting users who look similar and projects that pattern onto the unconsented cookieless pings, filling the gap in your users, sessions, and conversion counts.

Say 40 percent of your traffic declines cookies. GA4 looks at how the consenting 60 percent move through your site, then models the likely behavior of the missing 40 percent from their anonymous pings. Your reports show a modeled total instead of only the observed slice, and a "modeled" indicator appears on the affected metrics.

Modeling only activates when you clear Google's traffic thresholds. A property needs enough daily volume, historically around 1,000 daily events with denied consent and 1,000 with granted consent over a period, or the model never turns on and you are left with raw undercounted data. Small sites get none of this, which is exactly why their numbers look worse after adding a consent banner.

It fills the gap with math, not with people who actually visited.