Guide · anonymized queries

Why do a Search Console query table's impressions add up to less than the whole site? What are “anonymized queries”?

Updated 2026-11-13 · kagentx

Short answer

Because Google gives no rows for anonymized queries but the site total counts them. Over the 25 days from 2026-10-17 to 11-10, MochiStamp had 5,518 impressions on the whole site and the queries shown in words add up to only 1,013 (18%); Gamma AI Slide over the same period had 719 on the whole site and 311 (43%) in words. A smaller second kind of gap comes from our dashboard storing only the top 25 queries a day (4% for MochiStamp); keep the two apart. The sample is only two new sites, so do not copy the ratios.

Why are the query table's impressions lower than the whole site's?

Because Google gives no rows for anonymized queries, but the site total counts them. Over the 25 days from 2026-10-17 to 11-10, MochiStamp had 5,518 impressions on the whole site, and the queries Search Console showed in words add up to only 1,013 (18%); the other 4,505 (82%) are anonymized queries, with no row you can see. Gamma AI Slide over the same period (10-19 to 11-10) had 719 on the whole site, 311 from queries shown in words and 408 (57%) from queries not shown.

Site (date range)Site impressionsQueries shown in wordsShareNot shown by Google
MochiStamp (10-17 to 11-10)5,5181,01318%4,505 (82%)
Gamma AI Slide (10-19 to 11-10)71931143%408 (57%)

Clicks are the same: over these 25 days MochiStamp had 178 clicks on the whole site, only 38 (21%) from queries shown in words and the other 140 (79%) from queries not shown. Source: local dashboard, read 2026-11-13, Google Search Console.

How does Google explain it officially?

Google's Search Console help says that, to protect user privacy, some queries are left out of the report and are called anonymized queries; they are included in chart totals unless you apply a query filter; and the table does not show every query other than the anonymized ones either. A Search Central blog post of 2022-10-19 defines an anonymized query as one not issued by more than a few dozen users over two to three months, and states that summing the clicks of all rows will not give the chart total.

These are the meanings of the original text as we checked it. We found no figure published by Google for what share anonymized queries usually take, so the 82% and 57% above describe only these two small sites over this period; both sites are new and most of their queries were searched only a few times, so they are more likely to count as anonymous.

Does the gap shrink as the site grows?

In MochiStamp's data it does: the share of site impressions from queries shown in words rose from 6% in mid-October to 24% in November. Probably because the same query is searched by more people and passes the “few dozen users” line. This is one site's observation, not a rule.

MochiStamp, by weekSite impressionsQueries shown in wordsShare
10-17 to 10-186346.3%
10-19 to 10-25584437.4%
10-26 to 11-011,19813911.6%
11-02 to 11-082,40450320.9%
11-09 to 11-121,84344724.3%

The last row includes 11-12, whose Search Console data was still incomplete, see How many days is Search Console data behind? Why are the latest days low?. Gamma AI Slide shows no such upward trend: 47.5%, 55.1%, 48.2%, and 30.5% in the last three days, because it suddenly got a batch of new visitors and new queries, and the sample is small too.

Our dashboard has a second kind of gap. How do you tell them apart?

Our data dashboard keeps only the 25 queries with the most impressions each day and folds the rest into one “other” row, keeping the numbers but not the text. So the gaps here are of two kinds with different causes:

  • Anonymized queries Google hides: site impressions minus “top 25 queries plus ‘other’”; Google gives no rows for these at all. For MochiStamp over the 25 days it is 4,505 (82%).
  • Caused by the dashboard storing only the top 25: that is the “other” row, 239 (4%) for MochiStamp and 8 (1%) for Gamma AI Slide. Google did give these; we just did not keep the text.

To tell which kind it is, check whether the “other” row is 0. From October 17 to 25, MochiStamp's “other” was 0 impressions, meaning each day had fewer than 25 queries, the dashboard cut nothing, and the whole gap then was hidden by Google; only in the week of 11-02 to 11-08 did “other” show a noticeable number (120 in total).

The page dimension does not have this problem: by page Google gives rows for all pages, so our page totals are 5,656 impressions and 178 clicks, in line with the site (impressions are 2.5% above the site total, which we have not verified). So to find out “which pages the traffic lands on”, the page table is more reliable than the query table; for an example see MochiStamp's LINE sticker size guide.

When the query table does not add up, how should you use the data?

  1. 1. Take the site total as the total first

    The site's impressions and clicks include anonymized queries and are the most complete numbers.

  2. 2. Treat the query table only as a “sample”

    Do not use it to infer the site's click rate or position distribution, because the missing part may be distributed differently from the visible part. For example the ranking bands in Impressions but no clicks: what to do with pages ranking 8 to 20 describe only the queries shown in words.

  3. 3. To see the whole picture, use the page dimension

    Page numbers have no anonymity problem and suit seeing which pages have impressions and which get clicks.

  4. 4. Do not add a query filter and then compare with the site

    The official help says anonymized queries are not counted in the total once a query filter is applied, so the filtered numbers will naturally be smaller than the site's.

  5. 5. Know the row limit when exporting through the API

    The API's rowLimit defaults to 1,000, is at most 25,000, and does not guarantee returning all rows. Once a site has many queries, an API export will be missing some too.

How a new site's impressions and clicks get started: How long until a new site gets its first 100 search clicks? Four sites, real numbers and How do you get a new website indexed by Google? How many indexing requests can you send a day?.

What are the limits of these numbers?

  • Only two sites and under 4 weeks; both are new and their queries are scattered, so the anonymous share may be higher than for a mature site.
  • We read the Search Console API data and did not check the sum of the query table item by item in Google's web interface.
  • The page dimension's impressions total is 2.5% above the site's; the reason is not verified.
  • Google does not say what words the anonymized queries are, so we cannot analyze them.

How average position is calculated, and click-through by position band, is in What does average position in Search Console mean? Position 8 but no clicks: how to read it.

FAQ

What are “anonymized queries” in Search Console?

Queries Google leaves out of the report to protect privacy. The Search Central blog defines them as queries not issued by more than a few dozen users over two to three months. They count in the chart totals but have no row.

Why do the query table's impressions add up to less than the whole site?

Because anonymized queries have no row but are in the total. In our MochiStamp data, over 25 days the queries shown in words are only 18% of site impressions.

How do you know whether the gap is hidden by Google or your tool stored only the top few?

Check whether your tool folds everything beyond the top few into “other”. Our dashboard has an “other” row: on days it is 0, the whole gap is hidden by Google; when it has a number, that part is what the tool cut.

Does the anonymized share fall as a site grows?

In MochiStamp's data it does: the share of queries shown in words rose from 6% to 24%. That is one site's observation; Google has not published such a rule.

To see where all the traffic lands, which table should you use?

The page table. Google gives rows for all pages, with no anonymity problem; our page totals match the site's clicks.

Sources

kagentx

Do AI answers
mention you?

An evaluation asks ChatGPT, Perplexity and Claude your buyers' questions (with web search on) and records who they recommend, which pages they cite and whether you are among them. Basic $1, results usually in 2–3 minutes.