Google Search referrers: why traffic reports disagree
Google documents Search referrer values and their limits. Check classification, redirects and measurement before treating missing visits as lost demand.
A visit can come from Google without your analytics tool identifying it correctly. Google’s new referrer guidance helps explain that gap, but it does not provide a way to recover every missing source or turn Search Console clicks into sessions.
What the referrer tells you
Google’s documentation, updated 12 November 2026, says Search normally sends an origin-only HTTP Referer: for example, https://www.google.com/ or a local Google domain. It omits the search query. The Android Google app can also send android-app://com.google.android.googlequicksearchbox/https/www.google.com.
These signals are not exclusive to ordinary web-search results: other Google surfaces can share them. Privacy protections, browser behaviour and tracking restrictions can leave the referrer blank or prevent a visit being recorded.
Google recommends checking custom channel rules, exclusions and filters if visits are misclassified. For server-log analysis, count landing-page requests rather than associated image, stylesheet or script requests. Search Console counts search clicks; analytics sessions depend on the destination loading and measurement working. Those totals need not match.
A useful investigation starts with one landing page
Choose a product or collection page whose traffic changed. Write down the date range, timezone, country and device filters used in each report. Keep a copy of the current reporting configuration before changing any rules.
Then compare the journey in stages: search click, landing-page request where logs are available, measured visit and completed order. Ask the person responsible for analytics to trace redirects and consent behaviour along that route. A source label alone cannot tell you where a visit was lost.
For a Shopify store, you may need your analytics provider or developer to inspect the available evidence. Do not assume that a merchant account exposes raw storefront server logs. Start with the reports and configuration you can actually access.
Record any confirmed configuration error separately from unexplained differences. Fixing a channel rule can change reported source totals without creating a single additional customer. Keep orders and revenue alongside traffic when judging whether the business improved.
Do not relabel all direct traffic as AI
An unknown source is still unknown. A growth claim from another operator’s stores cannot identify the visitors in yours. Treat anecdotes as questions to investigate, rather than a percentage to apply to your own direct traffic.
For context on a separate visibility change, see our Hydrogen Search rollout coverage. A model change and a measurement discrepancy require different checks.
Background: clicks and sessions
Google Search Central’s explanation, published 29 January 2025, is background to the measurement distinction, not a demonstration of the November document. Watch the original video.
More from Marketing
See all →

