← Back to blog
CONSENT MODE V2

My own website was running Basic Consent Mode and I never chose it

I write about Consent Mode for clients, and my own site wasn’t in the mode I would have said it was. I found out by running a test that I recommend to everyone and that almost nobody runs on their own site.

The test

A fresh window, no prior cookies or storage, with DevTools open on the Network tab. I load the site and accept nothing. I wait a few seconds and look at what goes out.

Here’s what I saw:

  • The dataLayer starts with gtag('consent', 'default', …) set to denied for all four signals, and it’s the first thing that runs. Correct.
  • gtm.js loads and the container initialises.
  • gtag/js is not loaded and no request goes out to /g/collect. Not even a cookieless ping.
  • No cookie is created.

On a visit that does accept, the request appears with gcs=G111, the container responds, and the event reaches GA4.

What it means

That’s the behaviour of Basic Consent Mode: while analytics consent is denied, the tag sends nothing. The container loads, but it doesn’t measure.

My own export confirms it. In the BigQuery tables for the days with data, there isn’t a single event without a user_pseudo_id, meaning without a cookie. In Advanced mode, a visitor who rejects leaves pings with no identifier; here there are none.

It’s not a bug. It’s a valid configuration and, in Spain, the most conservative one. But I hadn’t chosen it: it was simply what came out of how the tag had been set up.

Where the setting was

In GTM, on the Google tag, inside Consent settings. There are two blocks:

  • The built-in checks (ad_storage, ad_user_data, ad_personalization, analytics_storage), which the tag already has by default.
  • The additional checks: “The tag requires additional consent to fire: analytics_storage”.

That second line is what stops the tag from firing while analytics is denied. That’s how the container was set up, without anyone having decided which mode was being chosen.

An effect I didn’t expect: the landing page was lost

When I repeated the test accepting cookies on the same page, nothing went out. Not even the page_view. Consent switched to granted, but the tag had already tried to fire on load, had been blocked, and GTM doesn’t fire it again just because consent changes. Measurement started on the next page.

In other words, a badly built Basic setup loses the landing page of every visitor who accepts. If someone reads a single article and leaves, they don’t exist in your data.

The fix was small:

  • The banner adds a consent_source parameter to the consent_update event: banner when the decision was just made, and stored when it’s restored from a previous visit.
  • In GTM, the Google tag gets one more trigger: the consent_update event with consent_source equal to banner.

It keeps the analytics_storage requirement, so if the visitor rejects, nothing goes out. And the parameter avoids duplicates for returning visitors, whose consent is restored before GTM loads and whose tag already fires normally.

I checked it in production, in a clean window:

Case Result
I accept on the same page 1 page_view, gcs=G111
I return with consent already stored 1 page_view, no duplicate
I reject, with no prior cookies 0 requests and 0 cookies

Why choosing on purpose matters

  • Basic: you only measure people who accept. There’s no data from people who reject and no conversion modelling. It’s simpler to explain in your cookie policy.
  • Advanced: you send cookieless pings from people who reject, and Google models the gaps. You collect more, but you have to declare it in the cookie policy and it fits worse with the more cautious readings of the regulation.

Either one is defensible. What isn’t defensible is not knowing which one you have. And the only way to know is the test above, run in a clean window without accepting anything.

My decision

I’m staying on Basic, now on purpose and with the landing page fixed.

Advanced would give me little. Google’s modelling needs a minimum volume: according to its documentation, at least 1,000 events per day with analytics_storage denied for 7 days and at least 1,000 daily users with analytics_storage granted on 7 of the previous 28 days, plus an advanced implementation. My low-traffic sites don’t reach it; without modelling, all I’d gain is a rough count of visits from people who reject. In exchange, I’d have to explain in the cookie policy that I send pings before consent, and those pings carry IP and user agent to my server. For a site whose message is “nothing goes out without your permission”, Basic is consistent and can be demonstrated with the test above.

For a client with a lot of traffic and heavy Ads spend, the maths would be different. That’s why the question isn’t “which one is better”, but “which one do you have, why, and have you checked”.