Skip to main content
Blog
Back to Blog

How to identify Google Search traffic in your server logs

On November 12, 2026, Google published a Search Central page on how to recognize visits from Google Search in your own web analytics or server logs. That matters if you publish pages for search and want to know whether Google sends real people to them, especially when you run no analytics script or many visitors block it. This guide is for a founder or small team with their own site: the Referer values to match, a script that counts Google Search visits per page from a standard access log, how to read the gap against Search Console, and a worked example. It is useful without a Soloop account and is not a statement from Google.

The short answer

A visit from Google Search is a request for one of your pages, not an image or script, whose Referer header is https://www.google.com/, a local Google domain such as https://www.google.co.uk/, or the Google app's android-app://com.google.android.googlequicksearchbox value. Google sends only the scheme and host, never the path or the search query.

Set aside requests whose URL carries gclid, gbraid or wbraid: those are Google Ads clicks with auto-tagging, not organic search. Expect your count to be lower than Search Console's clicks, because browsers and privacy tools often drop the Referer. Track the ratio between the two over time rather than expecting them to match.

What Google published on November 12, 2026

Google's new page, "Identify Google Search traffic in your web analytics or server logs", sits in the Monitoring and debugging part of Search Central. It lists the Referer values Google Search sends, explains why the Referer is sometimes blank, says to count only landing-page requests and gives reasons your numbers will not match Search Console. It explains how to read traffic you already receive; it is not a new Search Console report.

The Referer values to match

Match the host, not a full URL. These are the values Google lists, by where the person clicked. The Google app on Android can send android-app://com.google.android.googlequicksearchbox/https/www.google.com.

Referer values Google Search sends, by surface
DimensionReferer valueWhat to know
Google Search on the webhttps://www.google.com/ or a local Google domain, such as https://www.google.co.uk/ or https://www.google.co.jp/Match every country domain, not only google.com, or you will miss visitors outside the US.
Google app on AndroidThe Google app's android-app value shown above, or the web valueThe android-app value appears when the app opens the link in a browser tab.
Google app on iOShttps://www.google.com/ or a local Google domainIndistinguishable from a web click.
DiscoverThe same value as the surface where the person saw DiscoverYou cannot separate Discover from Search with the Referer alone.
Google Images and other services on the same originsThe same values as Google SearchIncluded when you match these values; other Google hosts, such as mail.google.com, are not Search.

Count Google Search visits per page

Most web servers write an access log with one line per request. On a typical Linux server with nginx it is /var/log/nginx/access.log; with Apache it is often /var/log/apache2/access.log or a file your hosting panel lets you download. The script below reads the "combined" format, which is nginx's default and a common Apache setting, and also reads rotated .gz files.

Copy the script, paste it into a new file named google_search_visits.py, read it, then run python3 google_search_visits.py access.log access.log.1 > google-visits.csv. It needs only Python 3, reads the files on your computer and sends nothing anywhere. The CSV lists each page with its Google Search, Google Ads-tagged and ChatGPT-tagged visits; a one-line summary with the number of days and all page views by source, including blank-referer visits, prints separately.

Download Markdown

What the script counts and skips

  • Counts only successful (2xx) GET or HEAD requests, so a redirect and the page it leads to are not counted twice.
  • Skips images, stylesheets, scripts, fonts, feeds and paths in SKIP_PREFIXES; they carry your own page as the Referer anyway.
  • Skips requests whose user agent looks like a bot, crawler, monitor or command-line tool. Anyone can fake a Referer, and spam bots do.
  • Matches the Referer host exactly against Google's domains, so lookalikes such as google.example.net and other Google hosts such as mail.google.com are not counted.
  • Counts a Google visit whose URL has gclid, gbraid or wbraid separately as google_ads_tagged.
  • Counts a visit whose URL has utm_source=chatgpt.com separately as chatgpt_tagged.
  • Ignores the query string when grouping, so /pricing and /pricing?ref=x are one page.

Separate paid clicks and ChatGPT visits

Clicks on Google Ads also arrive from Google's domains. With auto-tagging on, Google adds gclid to the landing page URL, and on some iOS clicks gbraid or wbraid instead. If auto-tagging is off and you run ads, part of what the script calls google_search may be paid; Search Console counts only organic clicks.

OpenAI says ChatGPT adds utm_source=chatgpt.com to links it sends people to from search results, so those visits can be counted even when the Referer is blank. Other assistants are not covered by the script; add their tags or hosts only from their own documentation.

Compare with Search Console

Open Search Console's Performance report, choose the same dates as your log and filter by page. Compare a whole week rather than single days, so differences in time zone and logging delays even out. Then divide your log count by Search Console's clicks for the same page.

Common gaps between your log and Search Console, and what to check
DimensionLikely reasonWhat to check
Log count is lower than Search Console's clicksBrowsers, private modes, new tabs and privacy extensions drop the Referer; those visits show as blank.Expected. Check that the ratio stays roughly the same from week to week.
A page has Search Console clicks but almost no log visitsA CDN served cached copies without your server seeing the requests, the visits are logged under an old URL that redirects, or a redirect sent a stricter Referrer-Policy.Check whether your CDN keeps its own logs, look for the visits under the old URL and check the Referrer-Policy header on redirects.
Log count is higher than Search Console's clicksPaid clicks without auto-tagging, reloads by the same person, or bots with a fake Referer.Turn on auto-tagging if you run ads, and look at the user agents and IP addresses behind the extra lines.
Analytics shows Google visits as direct or referralCustom channel rules, referral exclusions or filters reclassify them, or a JavaScript or meta-refresh redirect replaced the Referer with your own page.Review the channel grouping and referral exclusions in your analytics tool, and use server-side redirects.

What the Referer cannot tell you

  • The search query. Use Search Console's Performance report for that.
  • Whether the click came from an AI Overview or AI Mode: Google's page lists no separate Referer values for them.
  • Whether the visit came from Search, Discover or Google Images; they share the same values.
  • How many different people visited. One person reloading a page appears twice.
  • Where blank-referer visits came from. Some are from Google; you cannot tell which.

Worked example: one guide over two weeks

This example is invented to show the workflow. A founder publishes a pricing guide on a site with no analytics script and runs the script over each week's log.

  1. Week 1: Search Console shows 120 clicks to /blog/pricing-guide. The script counts 84 google_search visits to the same page. 84 / 120 = 70%.
  2. Week 2: Search Console shows 130 clicks; the script counts 91. 91 / 130 = 70%. The ratio held, so the founder can use the log as a weekly check between Search Console updates.
  3. The week 1 CSV also shows 6 google_ads_tagged visits to /pricing from a small Google Ads test and 4 chatgpt_tagged visits to the guide. Those 6 + 4 = 10 visits stay out of the google_search column.
  4. In week 3 the founder turns on full-page caching at their CDN. Search Console shows 125 clicks, but the log counts 12 visits: 12 / 125 is about 10%. Most requests are now answered from the CDN's cache, so the founder runs the same matching rules on the CDN's logs instead.

Where Soloop fits

Soloop does not read your server logs, analytics or Search Console. Use the script and Google's tools above for measurement.

Soloop helps with the work that brings people to those pages: according to the Soloop manual, it can research a market with sources, write and publish X posts and replies through a connected account, send outreach through a connected Gmail account and prepare Google Ads search campaigns after setup and payment, with actions waiting for your approval by default.

Sources, limitations and refresh

Google's Search Central page and Google Ads auto-tagging help were read on November 17, 2026. OpenAI's help center blocks automated reading, so its utm_source=chatgpt.com statement is taken from its publisher FAQ as shown in search results. The script was tested on a sample combined-format log covering web, country-domain and Android-app Referers, assets, bots, redirects, Google Ads and ChatGPT tags, lookalike and other Google hosts, blank Referers, malformed lines and .gz files. The tables and worked example are editorial guidance, and the example is invented.

This guide will be reviewed by February 17, 2027, or sooner if Google changes the page, adds Referer values for AI features or changes its referrer policy.

Product capability

The Soloop manual describes research with sources, X posting through a connected account, Gmail outreach and Google Ads search campaigns after setup and payment; approval comes first by default. It does not read logs or analytics.

Soloop product manual