← All posts

Consent Mode and how GA4 fills missing data with behavioral modeling and modeled conversions

Nov 14, 2025 • Written by Hricha Shandily

For a long time, website owners could collect as much data as they wanted without asking anyone. Nobody had to give consent, nobody questioned tracking, and tools like Google Analytics worked perfectly.

That time is gone. People say no to tracking, browsers block cookies, and privacy laws demand real consent. When a visitor rejects tracking, GA4 loses almost all data. To help patch this loss, Google created something called the Consent Mode.

Let’s break down what it is, why Google introduced it, what happens behind the scenes, and how Plausible takes a very different approach where you don’t have to lose data in the first place.

1. What is Google Consent Mode?
   1. How Consent Mode works?
   2. Setting up Consent Mode
      1. The reality: you still lose data
2. Why Consent Mode creates a legal gray area
3. What this means for your GA4 reports
4. Do I need Google Consent Mode?
5. GA4 is reconstructing data, not recovering it
6. Plausible as a privacy-first, accurate alternative to GA4
   1. What you do not need with Plausible
   2. What GA4 still cannot rebuild even with modeling
   3. The complexity you avoid by choosing Plausible

What is Google Consent Mode?

When you implement a cookie consent banner on your website, normally what you’d expect to happen is this: 

Consent given: track. Consent not given: do not track.

And because a major chunk of site visitors deny the cookie banners and do not like to give away their data, GA script gets blocked from loading on a site and in return, GA4 (and Google Ads) lose a lot of valuable data to show on their dashboards. In fact, about 50% of the data is known to be lost due to this very reason.

Also, GA4 depends heavily on identifiers to build its reports. It tries to:

• connect sessions
• create user journeys
• map attribution
• link behaviour from web to app

All this breaks the moment cookies or identifiers disappear. If someone says no to tracking, GA4 loses the ability to understand who did what.

Enter: Consent Mode. Google designed the consent mode to help you reconstruct some of that lost data (through anonymized data collection and data modeling). Also, Google Ads cannot optimize well when conversions vanish. So Google built modeled conversions to fill the gaps.

Basically, Consent Mode is Google’s patch to keep the ecosystem running.

How Consent Mode works?

Consent Mode is like the bridge between cookie banners and GA script. It helps your site or app tell GA script whether a user has agreed to cookies or tracking. So when someone gives or refuses consent, Google tags can accordingly change how they work.

There are two implementation options available: a “basic” mode where tags are blocked until consent is given. It’s simple in the sense that Google tags are completely blocked from firing when the user doesn’t consent. And if the user does consent, everything works normally.

And there’s an “advanced” mode where tags load with default denial, send limited “cookieless pings” and only send full measurement data when consent is granted. These pings do not include:

• User identifiers
• Cookie values
• Cross page state
• Any personally identifying data

So GA4 cannot link one page to another. A single user moving around your site may appear as ten separate events.

The thing to note is that even if consent is not given, Google tags still fire but dynamically adapt and anonymize the customer data. So some information is still collected.

To explain: say a visitor rejects the tracking cookies through your consent banner. You’d think now no tracking would happen but actually, the google tag still fires, except that the tag dynamically adapts and anonymizes the customer data this time.

Such anonymized data plus the patterns observed from people who did consent are used to estimate what the non-consenting users probably did. This is called behavioral modeling. Google also creates modeled conversions for Google Ads.

Consent Mode is a whole system that tries to rebuild the data GA4 couldn’t exactly collect: by reconstructing, reinforcing the patterns of behaviour from people who did consent and browsed the website.

Example: Imagine you run an online store and 100 people click your Google Ads campaign. Only 60 of them give consent, so GA4 can fully track those users, and you see 5 real purchases from them. The remaining 40 users do not give consent, so GA4 only receives limited, cookieless pings with no information about what they actually did.

Without modeling, your GA4 report would simply show 5 conversions and nothing from the 40 untracked users.

With Consent Mode modeling enabled, GA4 looks at how the consenting users behaved and uses that pattern to estimate what the non consenting users might have done. Based on the model, GA4 may decide that around 3 additional conversions likely happened among the 40 users who rejected tracking.

Your report now shows a total of 8 conversions instead of 5. GA4 mixes the 3 modeled conversions with the 5 real ones, and you cannot see which is which. The final numbers look complete, but some of them were predicted rather than observed.

Setting up Consent Mode

Google explains the full setup here. But the general process is:

1. Update the default consent state to “denied”
2. Load your tags but keep them blocked until the visitor chooses
3. When someone accepts, update the consent state to “granted”
4. GA4 and Ads start collecting full data again
5. If they reject, only pings are sent

It is not a plug and play feature. It requires Tag Manager configuration, banner integration, and usually some developer help.

If you want modeling to appear in your GA4 interface, you must switch your reporting identity to “Blended” from your settings.

This mixes real, observed events AND modeled, predicted events. GA4 does not tell you which is which in the final reports.

To even activate modeling, Google requires thresholds like:

• at least one thousand daily events where analytics storage is denied for at least seven days
• at least one thousand daily events where analytics storage is granted in the past twenty eight days

Small sites often cannot meet these requirements. So for many websites, modeling never happens.

The reality: you still lose data

Consent Mode is not a magic fix. Google is very clear that:

• Some data will always be missing
• Modeling does not fill everything
• Reports will still be incomplete

Even the raw data you can export to BigQuery is mostly empty for non consenting visitors. You get event timestamps, but no identifiers, no session linking, no user counts, nothing that helps you understand journeys.

It is only useful for basic things like:

• total event counts
• timestamps
• very simple aggregations

And you need SQL skills to work with it.

Why Consent Mode creates a legal gray area

Consent Mode does not observe non-consenting users directly, but indirectly without clearly communicating it to the end-user. Instead it uses limited pings and the behaviour of consenting users to predict how rejected sessions might have played out.

This reconstruction is triggered by a mathematical model, not by real data. Legal experts highlight that predicting behaviour after a user rejects tracking can be problematic. It does not go with the spirit of privacy-friendliness. This is why Consent Mode is seen as a gray area instead of a clear privacy solution.

Some might argue that such data doesn’t go with personally identifiable information like complete IP Addresses (semi-anonymized IP Addresses still register), so it should be okay.

But site owners are still sending some sort of data about the user (while the user thought you weren’t) to Google servers before actually anonymizing it, processing it and modeling it. The thing is: We can never know how that data is really processed by the servers, before being anonymized. Since Google is a closed-source and proprietary entity, there’s no way to find out either.

A complete nightmare for a company if found indulging in such practices, usually without even being fully aware of it.

What this means for your GA4 reports

To put it plainly:

• Your reports may look complete, but parts of them are guesses
• You cannot tell what is real from fake
• Understanding your own data gets harder
• Debugging becomes a challenge
• BigQuery exports can be confusing

Many businesses look at their reports and do not realize how much is modeled.

…plus, you’ll still need to invest in legal consulting.

Do I need Google Consent Mode?

This is the big question most website owners have, and the honest answer is: it depends on your setup and on what you expect from GA4. Here is the straightforward breakdown:

You need, or rather might want to explore, Google Consent Mode if:

• You use GA4 with cookies
• You show a cookie banner
• You want GA4 to legally respect the visitor’s choice but want more data
• You want GA4 to fill the gaps with modeled data
• You rely on Google Ads and need modeled conversions

Without Consent Mode, GA4 will simply stop collecting data whenever a visitor declines tracking. You will lose a lot of information, and nothing will be reconstructed.

You do not need Consent Mode if:

• You stop using GA4
• You use an analytics tool that does not rely on tracking identifiers
• You do not want modeled/inaccurate data in your reports
• You prefer simple, reliable, human readable analytics over predictive ones
• You are operating on a smaller scale
• You want to avoid dealing with consent banners, GTM settings, and the legal gray area

In other words, Consent Mode is required only if you want GA4 to keep “fully” functioning in a privacy regulated world. If you want to keep using GA4 and want your numbers to look somewhat complete, Consent Mode is basically unavoidable.

GA4 is reconstructing data, not recovering it

This distinction matters. Recovery means the data existed and we got it back. Reconstruction means the data did not exist and we estimated it.

GA4’s modeled metrics are reconstruction. They are not real events. They are predictions. Once modeling is active, GA4 mixes real observed data and predicted modeled data.

GA4 does not mark which is which. You cannot separate them in your reports. This makes analytics harder to trust because you do not know how much of the dashboard is based on actual activity versus machine learning.

Plausible avoids this because it never needs to guess.

Plausible as a privacy-first, accurate alternative to GA4

You do not need Consent Mode at all with Plausible because we do not rely on cookies, identifiers, or personal data in the first place. This changes everything about how analytics works in a privacy-first world.

What you do not need with Plausible

With Plausible, you skip all the complexity that Consent Mode tries to solve:

• No consent banner required with Plausible, because we are cookieless
• No legal battles (although please check it with your legal advisor for your specific case) because we’re privacy-friendly by design
• No reconstructed or stitched sessions
• No modeled behaviour
• No hidden guesses in reports
• No BigQuery
• No machine learning

The numbers you see in your dashboard are based entirely on real events and trustworthy, accurate analytics.

What GA4 still cannot rebuild even with modeling

Google’s modeling system only fills in the broad strokes. Even with all modeling features enabled, GA4 cannot reconstruct:

• user journeys
• full session behaviour
• what pages a single non consenting user visited
• time spent
• accurate source attribution
• whether someone is new or returning

This is why GA4 reports can still feel misleading or incomplete. You avoid this problem entirely with Plausible.

The complexity you avoid by choosing Plausible

Consent Mode is a workaround that creates a long chain of technical requirements. When you switch to Plausible, you skip all of this:

With GA4 + Consent Mode	With Plausible
Cookie banner must load before all other scripts	No banner needed in most cases
Scripts must wait for the correct consent state	Scripts load normally
GTM requires special configuration	No GTM dependency
Reporting identity must be set to Blended	No reporting identities to manage
Google modeling thresholds must be reached	No modeling required
BigQuery exports vary depending on consent	No BigQuery setup needed
GA4, Ads, and Tag Manager must stay in sync	Nothing to sync

Instead of adding more layers to fix a broken tracking model, Plausible works cleanly with how the modern web operates.

We have created Plausible to be a simpler, accurate, privacy-friendly alt. to Google Analytics. You can take a look at the full comparison hereand start a free trial here.

Written by Hricha Shandily