← Blog·Privacy & Consent

Privacy & Consent

Consent Mode v2: how to audit it for compliance

What Consent Mode v2 is, the four parameters and the two modes, and a step-by-step audit of defaults, firing order, regions and server-side forwarding.

David PombarSwiss army knife at Trackingplan
10 min read · 2114 words

Consent Mode v2 is Google's mechanism for telling its tags what the visitor allowed: two parameters for analytics and ad cookies, two more for sending user data to Google Ads and for personalized ads. Enabling it in the consent platform feels like checking a box, and that is where most teams stop. The configuration then quietly fails in production: defaults that fire late, a region that never gets denied, a server container that forwards every hit whatever the banner said. This guide, part of our consent management series, explains how Consent Mode v2 works and walks through the audit that finds those failures.

Consent Mode is a layer between your consent banner and Google's tags. The banner records the visitor's choice; Consent Mode translates it into parameters that GA4, Google Ads and Floodlight tags read before they do anything. Version 1 had two parameters. Version 2, required since March 2024 for anyone using Google Ads features such as remarketing and conversion modeling for EEA traffic, added two more:

ParameterControlsWhat breaks if it is wrong
analytics_storageGA4 measurement cookiesGaps in sessions and events, or cookies set without consent
ad_storageAdvertising cookiesBroken remarketing audiences, or ad cookies set without consent
ad_user_data (new in v2)Sending user data to Google for advertisingMissing conversion signals, or user data sent without consent
ad_personalization (new in v2)Personalized advertisingCompliance violations, or remarketing that should not run

The official reference for the four is Google's consent mode documentation; the Google Ads side is in the consent parameters help.

Infographic of consent parameters and their functions
The four Consent Mode v2 parameters and what each one controls

Basic and advanced mode

In basic mode no Google tag loads until the visitor grants consent; before that, Google receives nothing. In advanced mode the tags load before the choice and send cookieless pings with the consent state; Google uses them to model the conversions and sessions it could not observe. Advanced mode recovers data, and it is the one a legal team needs to review, because those pings still transmit behavioral signals. Which mode you run changes what a correct audit result looks like, so establish it before the first test.

Two behaviors matter for everything that follows: the default state fires before any interaction, so a wrong default skews data from the first page view; and an update has to change the parameters at once, not on the next page load.

Run the steps in order, in a clean incognito window, once per region you serve.

  1. Verify the default state. Open Google Tag Assistant, load the site, and read the Consent tab before touching the banner. All four parameters must appear with their default values. An empty tab means the default never fired.
  2. Confirm the defaults fire before the tags. In the Tag Assistant timeline the consent default must precede every GA4 and Google Ads tag. If a tag fires first, the default is late and the first hit went out with no consent state.
  3. Check the GA4 stream. In Admin > Data Streams, open the stream and confirm it receives gated events, and that the consent settings reported by GA4 match what you saw in Tag Assistant.
  4. Read Google Ads Diagnostics. In Google Ads, Tools > Diagnostics lists consent warnings, including missing ad_user_data and ad_personalization signals that GA4 never shows.
  5. Run the decline and the accept paths. Decline, and inspect the Network tab: in basic mode, no Google requests; in advanced mode, cookieless pings with gcs and gcd showing denied. Accept, and verify that cookies and the full tags fire. Reset state between tests.
  6. Look for duplicate updates. Several gtag('consent', 'update', ...) calls can override each other. Count them in the Tag Assistant log. More than one per banner answer is worth a look: check which one wins and whether they disagree.
  7. Repeat for the server. If you run a server-side container, confirm that the consent parameters reach it and that every server tag respects them. The server only knows what the browser sent; it does not inherit the banner's state. The server-side tracking pillar covers that path.

The GA4 audit checklist holds the checks that pair with these; if you run paid campaigns, the Google Ads audit adds the conversion side.

A consent audit is only as good as its scenarios. Test the happy path and you will pass. Test the visitor who closes the banner without answering, the one whose CMP loads late on a slow connection and the one who returns with last month's cookie, and you will find the gaps.

The steps above audit one site in one region. An agency with a client portfolio, or a brand with a site per market, needs the same audit repeated and a record that lets the sites be compared side by side.

Start with an inventory of every site in scope and group them by what can make them behave differently: consent region, CMP and its version, tag container and its version, and page templates. Test one representative site from each group first, to find the failures early, then test the rest before calling the portfolio consistent. A sample helps you prioritize; it does not show that the untested sites pass.

On every site, run the same scenarios in a clean session: the state before any choice, accept, decline, and a later change or withdrawal of the saved choice. Repeat them on the templates that matter for that site, such as a landing page, a lead form and the checkout. Add a template when its implementation differs from the others or a test on it fails. Same scenarios everywhere, or the rows can’t be compared.

Record the same fields for every site. In your spreadsheet each site is one row; this is what one record looks like:

FieldExample: shop.example.com
RegionEEA
CMP and container, with versionsCMP 4.2 · GTM v118
ModeAdvanced
Defaults before tagsAll four denied, set before every tag: pass
Accept, decline, withdrawPass · pass · fail (the Meta pixel still fires after withdrawal)
Server-sideConsent forwarded and honored: pass
Templates testedHome, product page, checkout
Evidence and ownerHAR file and Tag Assistant export · web team

Then compare the sites, not just each one against the checklist: a region whose defaults differ, a parameter that never updates on one CMP version, a server container that forwards consent on one site and ignores it on another. Investigate every failed or unknown cell before calling the portfolio consistent, and never average a failed site into a portfolio-wide pass. After a CMP or container release, retest the affected sites and keep the new evidence in their record.

The table records technical behavior. Whether a consent design meets legal requirements is a separate review for the privacy or legal owner. For the implementation details of defaults, updates and regional behavior, Google’s Consent Mode setup guide is the reference.

A portfolio audit is also where a manual process runs out of time: every new site, market or release adds rows to retest. Trackingplan checks every real hit on the websites of an account and on their server-side destinations, and flags a vendor that fires before the visitor chooses or keeps collecting after a decline, with the share of hits and sessions affected, the destination and the cookie. The table stays current between audits, and a site that drifts is flagged the day it does. Consent checks are part of the optional Compliance module.

The common misconfigurations

For the debugging side of each one, how it shows up in Tag Assistant and in the gcs and gcd parameters, see how to debug Google Consent Mode v2 without manual checks.

Analyst manually tests website consent compliance
Manual testing finds the obvious errors; production finds the rest
IssueWhat teams doWhat they should do
Missing EEA defaultsOne global default, often granteddenied defaults for the EEA, UK and Switzerland, set by region
Late defaultsThe CMP sets the default after GTM loadsConsent default in the page <head>, before the container
Only analytics_storage configuredThe two v2 parameters left unsetAll four parameters, in the default and in every update
Uncertified CMPA banner that does not implement v2A Google-certified CMP, or a verified custom implementation
Server-side gapConsent assumed on the serverConsent parameters forwarded and honored by every server tag
IframesEmbedded third-party content outside the CMPInventory the iframes and gate or document each one

Two of these deserve a note. Region-specific defaults are the one most often missed: GDPR applies to EEA visitors, and a single default for the world means either EEA visitors get granted or every visitor loses data. And the server-side gap is invisible in the browser: the vendor request leaves the server, so no extension will show you that it ignored the opt-out.

For the regulatory side, the privacy compliance guide for analytics covers the obligations behind these settings, and Google's own troubleshooting guide lists its diagnostics.

After the audit: keeping it true

Fixing the configuration is a point in time. The configuration degrades: the CMP vendor pushes an update, a new tag lands in the container without a consent check, a market opens with its own rules. Four habits keep the audit result valid:

Then measure it where it happens. Scenario tests validate the setup you tested; monitoring real traffic tells you whether it stayed valid for the visitors you did not. Trackingplan watches the consent state of every hit as it leaves the browser, the app or the server, and reports the vendor that received data before consent, the parameter that went missing after a release, and the market where the defaults changed, to the team that owns them. It ships with the digital analytics tools and the consent platforms you already use; the privacy hub has the wider picture.

Frequently asked questions

All four: analytics_storage, ad_storage, ad_user_data and ad_personalization. Check the default state of each one before any interaction and its updated state after accept and after decline, and confirm that the default fires before any Google tag.

In basic mode no Google tag loads until consent is granted. In advanced mode the tags load first and send cookieless pings carrying the consent state, which Google uses to model what it could not observe. Advanced mode needs a legal review; basic mode needs less, but Google gets no pings from your site and its conversion modeling falls back to a general model.

Quarterly, and after every change to the container or the CMP, since both can silently break the configuration between reviews. Monitoring real traffic in between is what catches the regressions a scheduled audit would find weeks later.

Group the sites by region, CMP, container and page template, test one site per group first and then the rest, and run the same scenarios on each: before any choice, accept, decline and withdraw. Keep one record per site with its results and evidence, and compare the sites. A sample tells you where to look; it does not prove the untested sites pass.

No. The server only knows the consent state the browser sent with the hit, and each server tag has to respect it. A server tag that fires for every event whatever the consent is a breach nobody sees in the browser.

An empty Consent tab in Tag Assistant, parameters updating after tags have fired, duplicate update calls, Google Ads Diagnostics warnings about ad_user_data, and consent rates that drop in one region with no banner change.

David PombarSwiss army knife at Trackingplan

Read more from David, a Senior Product Strategist with 18+ years in digital product development and an atypical error detection knack.

Read next

All Blog →

Trackingplan

See everything. Miss nothing.

Your implementations audited around the clock with real-time, real-user data. Real-time alerts about errors or changes in your data, campaigns, pixels, privacy and consent. Let AI flag issues before they cost you.