How Consent Mode Affects Your GA4 Conversion Modeling

What consent mode actually sends to GA4

WithConsent Mode v2(see ourConsent Mode v2 implementation guidefor the full setup), when a user declinesanalytics_storage, GA4 can still receive acookieless ping. Google's documentation says no Analytics cookies are read or written in that state, but the ping can still include regular browser communication data such as user agent, screen resolution, and IP-based request context. If an implementation sends fields such asuser_idor custom dimensions, those may still be sent as well.

Google uses these pings alongside consented user data to train a machine learning model that estimates behavior for users who declined analytics identifiers. The output appears in GA4 reporting layers, not as a neatly labeled row called "modeled conversions" that analysts can inspect line by line.

That means some reporting views can blend directly observed activity with modeled behavior. The exact impact depends on your consent rates, property volume, and the specific report being used. This blending is also one reason teams seeGA4 numbers diverge from Google Adsfor the same campaigns.

The accuracy of modeled conversions

Modeled reporting can be directionally useful, but it should not be treated as a line-item replacement for order-system truth. It is strongest as an aggregate signal, not as proof that a specific campaign or segment produced a specific exact number of conversions. The same caution applies when interpreting modeled rows alongsideGA4 attribution modelsin standard reports.

It also has practical limits. Consent rates vary by region, device, and implementation. If those patterns differ materially between consenting and non-consenting users, modeled results may be less representative for some segments than others.

Auditing the modeling gap in your property

To understand the modeling gap in your property, compare total conversions in GA4 against your CRM or backend order system for the same period, then compare exported event totals in BigQuery where available. The goal is not to accuse one source of being wrong immediately. The goal is to separate implementation defects from expected scope differences between systems — our deeper guide onConsent Mode vs BigQuerywalks through the schema fields you need for that comparison.

Properties with low consent rates should document how consent-aware reporting works before using GA4 as the sole source for performance narratives. This is especially important when ad spend or bidding decisions depend on the reported uplift.