How much of your Search Console traffic is in hidden (anonymized) queries, and can you measure it?

Anonymized queries are searches that Google omits from the Search Console Queries table to protect user privacy; the chart totals count them unless you filter by query. The Google pages we read give no share for any site, and you can measure only an upper bound: the chart total minus your query rows, for clicks and impressions separately.

• Google says anonymized (rare) results are omitted from the Performance report's table but included in the chart totals unless a query filter is applied (Google, Search Console Help, read 30 October 2026).
• Google's October 2022 post described privacy-filtered queries as those not issued by more than a few dozen users over a two-to-three month period (Google, Search Central Blog, posted 19 October 2022, read 30 October 2026).
• The Queries table shows at most 1,000 rows, and Google says Search Console stores and shows only the most important data rows (Google, Search Console Help, read 30 October 2026).
• Google's table reference marks rare queries in the BigQuery export with an is_anonymized_query flag, while the export overview describes the data as with the exception of anonymized queries (Google, Search Console Help, read 30 October 2026).
• Penlid's hidden query share tool calls every gap between the chart total and the query rows an upper bound, even for a table under 1,000 rows (Penlid, hidden query share tool, read 30 October 2026).

What are anonymized queries in Search Console?

Anonymized queries are queries that Google leaves out of Search Console reports to protect the privacy of the people who searched, and its help pages also call them rare queries. The dimensions page says Some queries are omitted from the report to protect user privacy. The troubleshooting page says anonymized (rare) results are omitted from the table and that they are included in the chart totals unless a query filter is applied.

Google's October 2022 post on filtering and limits described the queries involved as ones that aren't issued by more than a few dozen users over a two-to-three month period. The current help pages we read give no number for what counts as rare; the troubleshooting page says only that the report omits some queries that are searched a very small number of times. Treat the 2022 figure as a description from 2022. Penlid's tool page makes the same point about the pages it cites: they give no rule by which a query is called rare, so the tool does not guess one.

"Hidden" therefore means missing from the rows, not from the unfiltered totals: an anonymized query's clicks and impressions are in the chart, and the Queries table has no row for them. Penlid's methodology page lists this among Search Console's known gaps, and the guide to why an export stops at 1,000 rows covers the other reasons a table adds up to less than the chart.

Does Google say how much of a site's traffic is in anonymized queries?

No: none of the Google pages we read gives a share of clicks or impressions in anonymized queries, for one site or on average. The 2022 description concerns how many users issue a query, not how much traffic it brings (our reasoning). Penlid's tool page explains why it uses no outside figure: No industry average is used, because the share differs from site to site.

Where are anonymized queries counted, and where are they left out?

Anonymized queries are counted in the Performance report's chart totals and left out of its Queries table, and a query filter removes them from the totals as well. The chart total and the table rows therefore answer different questions: how much traffic the site got, and how much Google lists by query.

Can you measure how much of your traffic is in anonymized queries?

Yes, but only as an upper bound: the share of clicks, and separately of impressions, that sits in no row of the Queries table. The free hidden query share tool computes it in your browser as one minus the query rows' total over the chart total, and Penlid's tool page words the result as an upper bound, with or without a full table.

The bound can be loose, because the gap holds more than anonymized queries. Google's pages say The table can display a maximum of 1,000 rows and Search Console stores and shows only the most important data rows, and, as the tool page notes, Google does not say how much it truncates. Nothing in the file splits the gap between anonymized queries and the rows Google does not show, so the tool does not try. A table of 1,000 rows or more is treated as possibly cut by the limit, and the wording says so; a shorter table is still labeled an upper bound.

1. Export with no query, page or search appearance filter. In Search Console open Performance, then Search results, choose the dates, and use Export, then Download CSV. Search type, country and device filters are fine because they apply to both tables alike. The tool reads the ZIP's filters file and refuses an export taken with a query, page or search appearance filter.
2. Keep the whole ZIP. It holds the daily table (Dates.csv, or Chart.csv in newer exports) and Queries.csv. Google's file has no total row, so the tool sums the daily rows for the chart total, and it refuses weekly, monthly and Compare tables.
3. Read clicks and impressions separately. Each is measured on its own; the two shares are usually not equal, and neither implies the other.

Can you see anonymized queries in the BigQuery bulk export?

Yes, according to Google's table reference, as flagged rows without query text, so you can see where they are but not what they were. Google's table reference says of the is_anonymized_query field in searchdata_site_impression: Rare queries (called anonymized queries) are marked with this bool. The query guidelines say An anonymized query is reported as a zero-length string in the table. and add that anonymized queries often make up the single most common query for a site. The table reference itself describes the empty query field as a zero-length string in one entry and as null in another.

The export's overview, however, describes the data with the exception of anonymized queries. Our reading is that the export holds them as flagged rows with no query text, but the pages we read do not say whether those rows hold all of their clicks and impressions. To measure the flagged share, add up clicks and impressions on the flagged rows, then divide by the export's own totals for the same dates; Google says rows can repeat a key and that you should almost always aggregate all your metrics, so sum the flagged rows. That calculation is our method, not a Google procedure, and the pages we read do not say the export's totals match the Performance chart. The chart-minus-rows subtraction also works on an API pull, as the row limit guide linked above explains.