GA4 saw 12 of my 23 Google clicks. Where did the rest go?
Quick answer
From 19 August to 15 September 2026, Google Search Console counted 23 clicks from Google search to this blog. Google Analytics 4, for the same 28 days, reported 12 sessions from google / organic. So GA4 showed about half of the visits that Google itself says it sent.
Most of that gap is by design. The site uses Consent Mode v2 with analytics denied until the visitor accepts. Google can fill the gap with behavioral modeling, but only above thresholds that a small site does not come close to. Below them, GA4 simply doesn't report visitors who decline. My property's settings say so directly: modeling isn't available for it.
The rest of this post is the two numbers, the configuration behind them, and what I now trust each tool for.
The two numbers
Both reports cover the same days. The Search Console number is filtered to the main domain, with final data only. The GA4 number is the Traffic acquisition report by session source and medium, with no filters.
| Source | 19 Aug – 15 Sep 2026 |
|---|---|
| Search Console, clicks from Google search | 23 |
GA4, google / organic sessions | 12 |
| GA4, all Organic Search sessions (Google, Bing and others) | 15 |
| GA4, all sessions from every source | 229 |
Twelve out of 23 is 52%. The numbers are small, and on a small site they always will be, so I don't read anything into the decimals. But half is not a rounding error. Only 3 of those 12 GA4 sessions counted as engaged, so the visible part is thin as well as small.
A click and a session are not the same thing. Search Console counts clicks on a Google result. GA4 counts a session when the Google tag runs on the page and is allowed to report it. Everything between those two events is where the other 11 visits went.
What the consent setup does
This is the start of the analytics bootstrap in the site's _app.tsx. Every storage type that can identify a visitor starts denied:
gtag('consent', 'default', { ad_storage: 'denied', ad_user_data: 'denied', ad_personalization: 'denied', analytics_storage: 'denied', functionality_storage: 'granted', security_storage: 'granted', wait_for_update: 500});
The Google tag then loads for everyone, whether or not they have answered the banner. That's what Google calls the advanced implementation. While analytics_storage is denied, the tag sends cookieless pings instead of normal hits, with no client ID that would tie two page views to the same person. When a visitor accepts, the banner calls gtag('consent', 'update', …), and the answer is stored and replayed on their later visits.
So the site does send something to Google for every visitor that runs the tag. The question is what GA4 does with the pings from people who never accepted.
Why GA4 doesn't fill the gap
Google's answer is behavioral modeling for consent mode. It trains a model on the visitors who accepted and uses it to estimate the ones who didn't. The help page lists the prerequisites, and two of them are about volume:
- at least 1,000 events per day with
analytics_storagedenied, for at least 7 days; - at least 1,000 daily users with
analytics_storagegranted, for at least 7 of the previous 28 days.
It also says what happens below that line: "events triggered by users who decline consent aren't reported."
This blog had 229 sessions in 28 days across every source, which is about 8 a day. The threshold is 1,000 consenting users a day. It will not get there, and I checked instead of assuming. In GA4, under Admin, Data display, Reporting identity, the property is set to Blended. The line for modeled data says modeling is not available for this property, and that it will turn on by default if that changes.
That leaves the reports with observed data only, which means visitors who accepted. The cookieless pings from everyone else are collected and never shown.
What else is in the gap
Consent is the biggest documented cause, but it isn't the only one, and GA4 can't tell them apart. Every one of them looks the same from inside GA4, a visit that isn't there:
- Ad and tracker blockers stop
gtag.jsfrom loading at all. Consent never comes into it. - Quick exits. The tag is injected after hydration, so someone who hits back within a second or two can count as a click and never become a session.
- Clicks that aren't unique people. A visitor who clicks the same result twice is two clicks in Search Console and may be one session in GA4.
The honest conclusion is an upper bound. At most about half of the visitors Google sent were visible to GA4, and I can't split the missing half between consent and blockers with the tools I have. Separating them needs a count that sees every request whatever the browser allows, which is the next thing I'm setting up on this site. I'll report it when there are 28 days of it.
What this changes on this blog
The view counter was reading the smaller number. Each post shows a view count that comes from the GA4 Data API. It only ever counted page views from visitors who accepted cookies. It counts the readers who consented, and for Google traffic that was about half.
Search questions go to Search Console. When I decided which kind of post to write next, I used Search Console clicks, not GA4 sessions: one investigation post was taking most of the blog's clicks while the error-fix posts took none. Had I used GA4, I'd have been deciding on about half the visits, from samples already too small to lose half of.
GA4 keeps one job. For the people who accepted, it still shows what they read and how long they stayed. That's useful, as long as I remember it describes a subset that chose to be counted.
Check your own site
This takes ten minutes and needs no code:
- Pick a 28-day window that ended at least three days ago, so Search Console's data is final.
- In Search Console, Performance, filter to your own hostname and write down the clicks.
- In GA4, Reports, Acquisition, Traffic acquisition, switch the dimension to session source / medium, set the same dates, and write down the sessions for
google / organic. - In GA4, Admin, Data display, Reporting identity, expand Blended and read whether modeling is available.
If the GA4 number is well under the Search Console one and modeling is not available, your reports are showing the visitors who said yes, and nobody else. That's fine if you know it. It's a problem if you're making decisions as if it were everyone.