Identify Google Search traffic in GA4 and server logs
Google Search visits can be identified in your own analytics or server logs by checking the HTTP Referer header: on the web and iOS it is https://www.google.com/ (or a country domain like https://www.google.co.in/), and from the Google app on Android it can start with android-app://com.google.android.googlequicksearchbox/. Google spelled this out in a new Search Central page published on 12 November 2026. It matters because many Indian sites see organic traffic leak into "Direct" or "Referral" in GA4, and this is now the official reference for fixing it.
What Google published
The page, first spotted by Search Engine Roundtable, covers four things: which Referer values Google sends, how Google sets them, why the Referer is sometimes blank, and why your numbers won't match Search Console.
The key technical detail is that Google Search uses the origin referrer policy. The header carries only the scheme and host, with no path and no query. In Google's words, the Referer "doesn't include the search query", so keyword data still has to come from Search Console.
Referer values by surface
| Where the click came from | Referer you will see |
|---|---|
| Google Search in a browser | https://www.google.com/ or a local domain such as https://www.google.co.in/ |
| Google app on Android (opens in a browser tab) | android-app://com.google.android.googlequicksearchbox/https/www.google.com or the plain https://www.google.com/ form |
| Google app on iOS | https://www.google.com/ or a local domain |
| Discover | Same as the surface it was viewed in (Google app or web) |
Google Images and other services served from the same origins send the same values, so they are counted together. The page does not mention AI Mode or AI Overviews separately, so don't assume you can isolate those clicks with the Referer alone.
Why GA4 might misfile Google traffic
GA4 puts a session in Organic Search when the source matches its built-in list of search sites or the medium is exactly "organic", according to GA4's traffic-source definitions. Out of the box, Google traffic should land there. When it doesn't, Google's new page points to three usual causes:
- Custom channel rules that override the default grouping.
- Referral exclusions or filters that strip or reclassify the source.
- Your own redirects. A JavaScript or meta-refresh redirect replaces the Referer with your own URL. Redirects that drop the query string also lose
gclid, which hurts paid tracking too.
In India we see the redirect problem most on sites that bounce mobile users from example.com to m.example.com, or send visitors through a language or city picker first. Every one of those hops is a chance to lose the Referer.
How to find Google visits in server logs
Analytics tools miss visits when consent banners, ad blockers or privacy settings block the tag. Server logs don't depend on JavaScript, so they are a useful cross-check.
Google's page does not give a regex. These are our own patterns, so test them on your logs before relying on them:
- Web and iOS:
^https://www\.google\.[a-z.]+/$ - Android Google app:
^android-app://com\.google\.android\.googlequicksearchbox/
The first pattern is broader than Google's listed domains, so spot-check what it catches. Also note a detail from the doc: browsers send a Referer for images, CSS and JavaScript too, with your own page as the value. Count only requests for landing pages, not assets, or your totals will be inflated.
Why the numbers won't match Search Console
Expect a gap, and don't spend a week chasing it. Google lists these reasons:
- Blank Referers. Private browsing, anti-fingerprinting protections, some new-tab behaviour, native search widgets that open an external browser, and privacy extensions can all suppress it.
- Timing. Search Console counts a click when the result is tapped. Analytics counts a visit only after your page loads and the tag fires. On slow mobile connections, that difference is real.
- Paid versus organic. Search Console excludes Google Ads clicks. Your analytics view may not.
- Consent and blockers stop the tag from firing at all.
For a site where most traffic is Android users on 4G, the timing gap alone can be noticeable. If your landing pages are slow, the shortfall between clicks and sessions is partly a speed problem.
This also matters if you are tracking the effect of a recent update. If you are comparing traffic before and after the November 2026 core update, make sure a tracking change isn't hiding inside the ranking change. The same applies after infrastructure moves; see our note on CDN migration and SEO.
What to do this week
- [ ] In GA4, open Traffic acquisition and filter for session source containing "google" with medium "referral". Any sessions there point to a classification problem.
- [ ] Check Admin → Data streams → Configure tag settings → List unwanted referrals for entries that shouldn't be there.
- [ ] Review any custom channel group rules that touch "google".
- [ ] Map every redirect between your search landing pages and the final URL. Replace JavaScript or meta-refresh redirects with server-side 301s where you can.
- [ ] Pull one week of server logs and count landing-page hits matching the patterns above. Compare with GA4 Organic Search sessions for the same week.
- [ ] Note the ratio. A stable gap is normal. A sudden change means something in your tracking moved.
FAQ
Can I see the search keyword in the Referer?
No. Google uses the origin referrer policy, so only the scheme and host are sent. Use the Search Console Performance report for queries.
Is google.co.in counted as Google Search?
Yes. Google says local domains are sent as the Referer, and GA4's search-site list treats them as search sources.
Can I separate AI Mode or AI Overview clicks with the Referer?
Not based on this guide. It does not list separate values for those features, so treat them as part of Google Search traffic unless Google documents otherwise.
Why does GA4 show fewer Google visits than Search Console clicks?
Blank Referers, blocked tags, consent choices and the timing gap between click and page load all reduce the analytics count. Some difference is expected.