Consent Mode V2 — The Complete 2026 Implementation Guide

What is Consent Mode V2 and do I need it?

Consent Mode V2 is Google's API for adjusting tag behaviour based on user consent status. It was required for EU/EEA/UK Google Ads conversion measurement by March 2024. If you run Google Ads to EU/EEA/UK users and haven't implemented Consent Mode V2, your conversion data is incomplete and Google Ads Smart Bidding is working from a degraded signal. Consent Mode V2 adds two new consent parameters to V1 (analytics_storage and ad_storage): ad_user_data (consent to send user data to Google) and ad_personalization (consent to personalised advertising).

Without all four parameters set correctly, Google's modelled conversion filling does not activate for non-consenting users, causing conversion undercounting that typically ranges from 20–50% in high-GDPR-sensitivity markets.

Basic vs Advanced mode: the critical choice

Basic mode

Tags do not fire until consent is granted. No data is collected from non-consenting users. No pings sent to Google before consent.

What you lose: Modelled conversions. Google cannot model what non-consenting users would have done because it receives no behavioural signal from them.

When to use Basic: When your legal team or DPA requires no data transmission before explicit consent. Higher compliance certainty, significant data loss.

Typical data loss in Basic mode (UK/EU traffic): 30–60% of sessions untracked depending on CMP accept rates. Accept rates for cookie banners in the UK average 45–65%; in Germany they average 25–40%.

Advanced mode

Tags fire with limited functionality before consent is granted. GA4 sends cookieless pings (no user identifier, no behavioural data) that allow Google to model what non-consenting users would have done. After consent is granted, full measurement resumes.

What you gain: Modelled conversions fill the gap for non-consenting users, typically recovering 50–80% of the "lost" conversions as modelled data.

When to use: For most businesses running Google Ads to EU/UK audiences. Legal basis: the cookieless pings sent in Advanced mode contain no personal data and no cookies — Google considers this compliant with GDPR. Confirm with your legal team; some jurisdictions and DPAs have stricter interpretations.

Recommendation for most businesses: Advanced mode. The data recovery from modelled conversions substantially outweighs the marginal compliance risk for most implementations.

The four consent parameters

All four must be set in V2:

Default state (before consent decision): Always set all four to denied as the default. This is the legal requirement — deny first, grant on consent.

GTM implementation: step by step

Step 1 — Consent initialization tag

In GTM, create a new tag:

• Tag type: Custom HTML (or use Google's native Consent Mode tag if available in your GTM version)
• Trigger: Consent Initialization - All Pages (this fires before any other tags)
• HTML content: the gtag('consent', 'default', {...}) code above

Step 2 — CMP update tag

When the user makes a consent decision, the CMP must update the consent state. Most CMPs provide a GTM integration that does this automatically:

Cookiebot: Install the Cookiebot CMP template from the GTM Template Gallery. It handles default + update automatically.

OneTrust: Install the OneTrust GTM template. Configure it with your OneTrust domain script URL. The template manages both default and update states.

Usercentrics: Install Usercentrics via GTM template. Configure consent group mappings to GA4/Google Ads consent types.

For manual/custom CMP implementations, fire a gtag('consent', 'update', {...}) call from your CMP's consent callback:

Step 3 — Configure GA4 for consent-aware firing

GA4 Configuration tag in GTM should have "Consent settings" configured. In the tag's Advanced Settings → Consent Settings:

• Additional consent checks: Require analytics_storage

This ensures the GA4 tag only fires with full measurement when analytics_storage is granted.

Step 4 — Verify with the `gcd` parameter

The gcd parameter appears in GA4 network requests and encodes the consent state at the time of the hit. Check it in Chrome DevTools → Network → filter for collect:

• gcd=11t — all consent granted
• gcd=11p — consent denied, tag firing in Basic mode (no data sent to Google)
• gcd=11l — consent denied, tag firing in Advanced mode (cookieless ping sent)

If you see gcd is absent from requests, Consent Mode is not initialised. If you see gcd=11p on all hits (including post-consent), the update tag is not firing correctly after user acceptance.

Verifying modelled conversions are activating

In Google Ads → Conversions → select your GA4-imported conversion → look for "Modelled conversions" column in the attribution report. If modelled conversions appear (they are labelled as such), Advanced mode is working and Google is filling the gap for non-consenting users.

In GA4, modelled conversions appear in reports with an asterisk (*) or modelled data indicator. They are not broken out separately in standard GA4 reports — they are folded into conversion totals.

Expected modelling recovery: 50–80% of "lost" conversions recovered as modelled data in most markets. German traffic typically sees lower recovery rates (25–40%) due to lower CMP accept rates reducing the signal Google needs for accurate modelling.

Consent Mode and GA4 data quality

What Consent Mode affects

• Session counts in GA4 — sessions from non-consenting users are not counted (no cookie = no session stitching)
• User counts — non-consenting users are not tracked as returning users
• Attribution — non-consenting sessions that convert are not attributed to channels
• Audience building — non-consenting users are not added to GA4 audiences

What Consent Mode does NOT fix

• Consent Mode does not make GA4 compliant on its own. You still need a valid legal basis, a compliant CMP, proper privacy notices, and data processing agreements.
• Consent Mode does not recover user-level data for non-consenting users — it models aggregate patterns, not individual journeys.
• Consent Mode does not fix attribution for 100% of lost conversions — modelled conversions are estimates, not actuals.