For years, anyone trying to identify Google Search traffic relied on GA4’s defaults and a handful of aging blog posts. On November 12, 2026, Google finally wrote the rule down itself: a Search Central page called “Identify Google Search traffic in your web analytics or server logs”, listing exactly what you should see when a visitor arrives from Google (Google Search Central). Nothing earth-shattering, but it is the first official reference you can hold your setup against.
And you will need it, because the same question keeps coming back: why does Search Console report, say, 1,200 clicks when GA4 only shows 900 organic sessions? Where did the rest go? Here is what the page says, how to check it in GA4, in BigQuery and in your server logs, and how to explain the gap without shrugging.
What Google published on November 12, 2026
The page shows up as new in the Search Central changelog, and PPC Land covered it the same day. It boils down to three ideas: the shape of the referrer, the list of expected values, and the reasons your tools will never land exactly on Search Console’s numbers.
The “origin” referrer policy: no path, no query
Google Search uses the origin referrer policy. The Referer header your site receives contains only the scheme and host: https://www.google.com/, full stop. No path, no parameters, and therefore never the query the user typed. If you are still hunting for keywords in your logs, stop: they have not been there for a long time, and Google explicitly points you to the Search Console Performance report for that.
A useful mental picture: the Google referrer is the postmark on an envelope. It tells you which city the letter came from, never what is inside.
Google Search referrer values by surface
The page lists the values to match:
| Surface | Referer value |
|---|---|
| Google Search on the web | https://www.google.com/ or a local domain (https://www.google.co.uk/, https://www.google.de/…) |
| Google app on Android, link opened in a browser tab | android-app://com.google.android.googlequicksearchbox/https/www.google.com |
| Google app on Android, other cases | https://www.google.com/ or a local domain |
| Google app on iOS | https://www.google.com/ or a local domain |
| Discover | the same value as the surface where the content was seen (Google app or web) |
| Google Images and other services on the same origins | the same values as Search |
The Android row is the practical news: it is the first time Google has documented the android-app:// value that many sites see as an odd-looking source. The last two rows are the bad news: Discover and Images send the same values as Search, so the referrer cannot tell them apart. If you want a breakdown by surface, Search Console still owns it, not your analytics tool.
What the page does not cover
The page says nothing about AI Mode, AI Overviews, Gemini or Google News. I am not going to guess on Google’s behalf: without documentation, assume an AI Mode click probably arrives with a regular Google referrer, but check it in your own data before you build a report on that assumption. To measure how much of your organic traffic AI is eating, the method lives elsewhere, in the 5 SEO analyses to run across GA4 and Search Console.
How to identify Google Search traffic in GA4
According to Google, most analytics tools classify these values as organic Google traffic automatically. In GA4, a visit arriving with https://www.google.co.uk/ as referrer ends up as source google, medium organic, channel Organic Search. That is the happy path, and for most of your visitors it works.
The Android case to check
The android-app://com.google.android.googlequicksearchbox/... value is trickier. I found no official documentation stating how GA4’s default channel group classifies it today, and public reports contradict each other. Rather than assume, check: in the Traffic acquisition report, add “Session source” and look for googlequicksearchbox. If it sits under Referral, you have Google Search traffic filed in the wrong place, and a custom channel group will put it back where it belongs.
The settings that push it into Direct or Referral
Google says it plainly: if Search traffic shows up as referral or direct, check your custom channel rules, referral exclusions and filters. In GA4, that gives you three suspects:
- A custom channel group with hand-written rules and an overly strict source condition (exact match on
google, while the source is sometimes a local domain or the Android identifier). - An unwanted referrals list in the web stream settings where someone added
google.comwith the best intentions. Every Google visit then becomes Direct. - A client-side redirect on your site (a JavaScript language redirect, a consent flow that reloads the page) that replaces the Google referrer with your own URL.
The third one is the nastiest, because nothing in GA4 flags it. The visit really came from Google, but the page that loads the tag thinks it came from you.
The BigQuery query to isolate Google Search sessions
The BigQuery export lets you compare the raw referrer with what GA4 made of it. The session_start event carries the page_referrer parameter; the session_traffic_source_last_click field holds the channel GA4 finally assigned. The query below isolates sessions that arrived with a Google referrer, drops those carrying an ad click identifier, and shows how GA4 classified them.
SELECT
session_traffic_source_last_click.cross_channel_campaign.default_channel_group AS ga4_channel,
session_traffic_source_last_click.cross_channel_campaign.source AS ga4_source,
session_traffic_source_last_click.cross_channel_campaign.medium AS ga4_medium,
COUNT(*) AS sessions
FROM `my-project.analytics_123456789.events_*`,
UNNEST([STRUCT(
(SELECT value.string_value FROM UNNEST(event_params) WHERE key = 'page_referrer') AS referrer,
(SELECT value.string_value FROM UNNEST(event_params) WHERE key = 'page_location') AS page_location
)]) AS p
WHERE _TABLE_SUFFIX BETWEEN '20261015' AND '20261113'
AND event_name = 'session_start'
AND (
REGEXP_CONTAINS(p.referrer, r'^https?://www\.google\.[a-z.]+/?$')
OR STARTS_WITH(p.referrer, 'android-app://com.google.android.googlequicksearchbox')
)
AND NOT REGEXP_CONTAINS(IFNULL(p.page_location, ''), r'[?&](gclid|gbraid|wbraid|gad_source)=')
GROUP BY ga4_channel, ga4_source, ga4_medium
ORDER BY sessions DESC;
The ideal output is a single row: Organic Search, google, organic. Anything else needs an explanation. Two honest limits: the query only sees consented sessions, and it cannot split Discover or Images from Search, since the referrer does not allow it. If you see the opposite, sessions with gad_source classified as organic, the problem runs the other way, and it is covered in the gad_source trap that files your Google Ads traffic as organic.
Finding Google organic traffic in server logs
Logs have one advantage GA4 will never have: they see every request, consent or not, ad blocker or not. They are the closest thing to Search Console clicks, provided you count properly.
The Referer patterns to match
In a combined log format (Nginx, Apache), the referrer is the quoted field right after the response size. The two patterns to recognize:
^https://www\.google\.[a-z.]+/$
^android-app://com\.google\.android\.googlequicksearchbox/
Count landing pages only, not assets
This is the classic mistake. When a page loads its images, CSS and scripts, those requests carry your own URL as referrer, not Google’s. Google spells it out: only landing page requests count as Search visits. So filter on HTML GET requests with a 200 status, and exclude asset extensions and bots:
grep -E '"GET [^ ]+ HTTP/[0-9.]+" 200 [0-9]+ "(https://www\.google\.[a-z.]+/|android-app://com\.google\.android\.googlequicksearchbox/[^"]*)"' access.log \
| grep -vE '\.(css|js|png|jpe?g|webp|svg|gif|ico|woff2?)(\?| )' \
| grep -viE 'bot|crawler|spider' \
| wc -l
If your logs already live in BigQuery, the logic is identical, and the table structure from the guide to blocking AI crawlers from BigQuery logs works by swapping the user-agent filter for a referrer filter.
Logs vs JavaScript tags: why logs count more
The server records the request before any script runs. A visitor who declines cookies, blocks GA4 or leaves before the page finishes loading shows up in the logs but not in GA4. So expect logs above GA4, and Search Console slightly above both. If the order is reversed, something is off in your filter.
GA4 vs Search Console clicks: the four official reasons
Google’s page warns that the numbers often will not match one to one, and gives four reasons. None of them is new, but having them in Google’s own words helps close the debate in a meeting.
Consent, ad blockers and stripped referrers
Extensions, ad blockers and consent choices can stop the tag from running, strip URL parameters or suppress the referrer. Add private browsing, anti-fingerprinting protections in some browsers, and handoffs from a native widget to the browser: those visits arrive with no referrer and end up as Direct in GA4. To size the share that vanishes, the diagnosis follows the same logic as tracking down missing data in GA4.
Redirects and an overly strict Referrer-Policy
A client-side redirect replaces the referrer with your own URL. A server-side redirect that drops the query string loses the gclid, and a paid click turns organic. A Referrer-Policy: no-referrer header applied too broadly (on a CDN, for instance) can also strip the header on your own pages and muddy the full path.
A click is not a session
Search Console counts a click when the user selects the result. GA4 counts a visit once the page has loaded and the tag has fired. In between sit quick exits, slow pages and flaky mobile connections. The heavier your page, the wider the gap.
Paid counted as organic, and the reverse
Search Console reports organic clicks only. Your analytics totals may include Google Ads traffic depending on configuration, and as we saw, a lost gclid slides paid traffic into organic. Always compare organic with organic, never a total with a total.
QA checklist
| Check | Where | Expected result |
|---|---|---|
googlequicksearchbox source | GA4, Traffic acquisition | Not in Referral, or fixed by a channel group |
google.com in unwanted referrals | GA4, web stream settings | Absent |
| Custom channel group rules | GA4, Admin | Conditions cover local Google domains |
| Classification of Google-referred sessions | BigQuery, query above | Organic Search, google, organic |
| Client-side redirects on landing | Site, developer tools | None, or referrer preserved |
Referrer-Policy | Site and CDN HTTP headers | No global no-referrer |
| Log count | Server logs | HTML pages only, assets and bots excluded |
| Volume order | Search Console, logs, GA4 | Search Console above logs, logs above GA4 |
Google’s page does not change the game, but it gives you an official baseline to audit how your organic traffic is classified. Block out an hour this week: run the BigQuery query, see where googlequicksearchbox lands, and check your unwanted referrals. If everything comes out as Organic Search, you will know the gap with Search Console comes from consent and page loading, not from your setup. And that is a sentence that holds up very well in a steering committee.