Search Console in GA4: which organic numbers you can trust
Plenty of properties have a Search Console link that nobody ever uses. Someone connected it once, went looking for the reports, failed to find them, and assumed the link was broken. It is not broken. Google keeps the whole Search Console collection unpublished until an editor publishes it.
The people who do find the reports run into a second problem. Clicks and sessions never agree. One number says 8,200 clicks from Google, the other says 6,900 organic sessions, and the difference of 1,300 has no label on it. That gap is not an error. It has a small set of documented causes, and once you know them the argument about which number is right becomes a sentence about which number answered which question.
What the link actually adds
Connecting a GA4 web data stream to a Search Console property makes two reports available. The first is Google Organic Search Queries, which lists the queries that produced impressions for your site and attaches four Search Console metrics to each: clicks, impressions, click through rate and average position. The second is Google Organic Search Traffic, which lists landing pages with both Search Console metrics and ordinary GA4 metrics such as active users and sessions, and can be broken down by country and device.
Neither report replaces the Performance report in Search Console, and Google says as much in the documentation: the data can also be found in Search Console. What the link buys you is proximity. Impressions and average position sit next to engagement data on the same screen, so the question "did that page rank well and then fail to hold people" stops requiring two browser tabs and a spreadsheet.
There is a hard limit on the history. Search Console keeps 16 months of data, so the reports in Analytics carry a maximum of 16 months as well. A three-year comparison of organic demand is not going to come out of this link.
The setup rules that catch people out
The link lives under Admin, then Product links, then Search Console Links. To create one you need the Editor role on the GA4 property and you need to be a verified owner of the Search Console property. Ownership can be verified during the linking step, which saves a detour. Deleting a link needs Editor as well.
Four rules decide how the thing behaves afterwards:
A web data stream can be linked to only one Search Console property, and a Search Console property to only one web data stream. An Analytics property can have only one data stream linked. If you run several domains, you are choosing one pairing per property rather than connecting everything to everything.
Links cannot be edited. To change the property or the stream you delete the link and create a new one.
That delete has a side effect worth knowing before you do it on a Friday. If the new link uses a different web data stream, all date ranges in those reports switch to the new stream. You do not get a clean break at the deletion date. The reports re-read history through the new stream.
Data starts when both halves exist. If the stream was created first and site verification came later, data is available from the date of verification. If verification came first, it is available from the date the stream was created. Either way, the earlier of the two periods is simply absent from the reports, and it looks like a tracking failure when it is a sequencing fact.
Why the reports look missing
The Search Console collection is unpublished by default. It sits in the Library, and it stays there until somebody publishes it. Once published, the two reports appear in the left navigation under Search Console. Editors and above can also add a removed report back if a colleague has been tidying the menu.
Search Console data also arrives late. Google states that the data is available in Analytics 48 hours after Search Console collects it. A link created this morning shows nothing for the last two days, and someone will call that a broken integration before lunch.
What you cannot do inside GA4
The Search Console metrics are picky about their neighbours. They are compatible only with Search Console dimensions and three Analytics dimensions: landing page, device category and country. You cannot slice impressions by session source, by campaign, by a custom dimension or by audience.
The Queries report carries Search Console metrics only. You cannot add key events to it, so "which query produced the enquiries" has no answer inside that table. The Traffic report does carry Analytics metrics on landing pages, which is the closest the link gets to joining demand with outcome.
And the reports do not support time series charts. You get tables. Trend lines stay in Search Console or move to Looker Studio, where Google points people with its own Search Console connector.
Why clicks never equal sessions
Five documented reasons explain almost every gap you will meet.
The counting unit is different. Search Console counts clicks on a search result. GA4 counts sessions on your site. A click that lands, followed by a trip back to Google and a second click on your site within the session window, is usually still one session. Search Console, aggregating at property level, also counts one click, because it de-duplicates repeat clicks from the same property in the chart. Match those two and you are still comparing a moment in the results page with a 30-minute activity window.
Rare queries are hidden, but their clicks are not. Search Console omits anonymised queries, meaning terms searched very few times, from the query table. Those clicks still sit in the chart total. Filter the table by a query and the match plus does-not-match figures can fail to add up to the unfiltered total for the same reason.
The table stops at 1,000 rows. Search Console displays a maximum of 1,000 rows. On a site with real long-tail search demand, the sum of the table is below the total in the chart, and the missing part is the tail. This is the single most common source of "the numbers in this export do not match the screen" complaints.
Time is measured differently. Search Console data is usually available 2 to 3 days after the fact, and daily figures are labelled in Pacific Time. GA4 uses the property time zone. A report run on a Monday morning compares a complete GA4 Sunday with a partial Search Console Sunday, and the last day always looks like a collapse.
Only one side needs JavaScript. Google counted the click on its own servers. GA4 recorded the session in a browser, and only if the tag ran and consent allowed it. Between the click and the session sit page speed, script blockers, a consent banner, a container that failed to load, and a redirect chain that dropped the campaign parameters.
The consent gap is the one that keeps growing
For UK sites running a consent banner with analytics storage denied by default, the fifth cause does most of the work. The click is a fact recorded by Google. The session is a measurement your tag was not allowed to make. When a third of visitors decline, the organic session count sits visibly below the click count, and no amount of container debugging closes it.
This is the point where a client asks whether tracking is broken. It usually is not. Pull the consent rate for the same period before you pull anything else. A gap that tracks the decline rate is behaviour, not a defect. A gap that appears in a week when consent did not move is worth investigating, and the order to check is tag firing, redirects, then filters.
A weekly routine that uses both numbers honestly
Use Search Console for demand and visibility. Queries, impressions, average position, click through rate. Remember the row cap and the hidden rare terms before you add anything up by hand.
Use the GA4 traffic report for what happened after the click. Landing pages, sessions, engagement, key events per session. This is where a page that ranks well and converts badly shows up as a number rather than an opinion.
Reconcile direction, not digits. Clicks up 20 percent and sessions up 18 percent is a healthy pair. Clicks flat and sessions down 30 percent is a measurement event, and it usually dates to a release, a consent change or a container edit.
Keep one honest exception list. Query strings on landing pages, trailing slashes, capitalisation and the www to non-www redirect all produce rows that look like separate pages. A landing page with a query string is a different row from the same page without one, and that is by design: the dimension is page path plus query string.
Before you blame the link
- Is the Search Console collection published from the Library?
- Does the linked stream cover the same pages as the Search Console property? Subdomains and legacy paths break this quietly.
- Do you hold Editor on the property and verified ownership of the Search Console property?
- Is the date range inside the 16-month window, and does it end two days before today?
- Are you comparing clicks with sessions, or clicks with users? Both are legitimate, and they answer different questions.
- Does the query table total match the chart total? If not, the row cap and the hidden rare terms explain it.
- What was the consent rate over the same period?
Two tools, two definitions, one site. The link is worth having because demand and outcome belong on the same screen, and it is worth understanding because the two numbers will never agree exactly. Say which question each panel answers and the monthly report stops being an argument.
An organic report that contradicts itself is usually two correct measurements of different things. Book a free tracking audit and we will check the link, the consent rate and the export against each other, then tell you which number to put in front of a client.