Across 33 websites we own and operate, 81.1% of Google Search clicks over 90 days came from queries that Search Console did not show, and the median site hid 83.2% of its clicks. The sites earned 19,509 clicks and 2,794,976 impressions in web search between August 17 and November 14, 2026. Every visible query row we could pull through the Search Console API added up to 3,680 clicks and 1,490,974 impressions. The other 15,829 clicks, and 46.7% of impressions, had no query attached.
That gap is not a bug in our pull. Google documents it. Search Console leaves out queries that too few people searched, and its API caps how many rows it returns. What the documentation does not tell you is how big the gap gets on a real portfolio of small and mid-size sites. On ours, the query report describes about one click in five. We call this the 80/20 visibility gap: roughly 80% of your search clicks arrive on queries you cannot see, so any report built on query rows is a report on the other 20%.
What we measured and how.
- Sites. 33 Search Console properties we own and operate: content, review, ecommerce and service sites. No client sites are in the data. 31 of the 33 returned at least one query row.
- Window. August 17 to November 14, 2026, a 90-day window.
- Source. The Search Console API, search type web. Site totals came from a request with no dimensions. Visible rows came from a request with the dimensions query and device, paginated until the API returned no more rows. That pull returned 116,155 rows.
- Hidden share. Site total clicks minus the sum of clicks on visible query rows, divided by site total clicks. The same math for impressions.
- Brand exclusion. For the device, position and query-length sections, we removed each site's own brand terms so branded navigation searches do not inflate CTR. The headline hidden share uses all queries.
- Per-site figures. We report the per-site hidden share only for the 14 sites with at least 50 clicks in the window. Below that, one or two clicks swing the percentage too far to mean anything.
Two caveats matter. First, the hidden share has two possible parts: anonymized queries, and rows beyond the API's export limits. Google's own deep dive on performance data limits (October 2022) defines anonymized queries as "those that aren't issued by more than a few dozen users over a two-to-three month period" and says they are included in chart totals "unless you filter by query." The same post puts the API export ceiling at "50,000 rows per day per site per search type" and the interface export at 1,000 rows. Our two largest sites returned 68,189 and 9,787 query rows, so some of their hidden share could be rows the API did not hand back. For the 12 sites under 7,000 rows, row limits are not a factor, and the hidden share there is anonymized queries.
Second, this is one operator's portfolio. Many of our sites are young, long-tail content sites, which is exactly the kind of site whose traffic is spread across thousands of rare queries. Your share may be lower. Read the numbers as what happens on sites like these, not as a web-wide rate.
Hidden clicks by site.
The portfolio figure is weighted toward the largest site, which earned 13,149 of the 19,509 clicks. The per-site view shows the gap is not one site's quirk. Twelve of the 14 sites with 50 or more clicks hid more than half of their clicks, and nine hid more than 80%.
| Site | Clicks (90 days) | Clicks with no visible query |
|---|---|---|
| Site A | 13,149 | 83.1% |
| Site B | 1,763 | 83.4% |
| Site C | 1,190 | 86.1% |
| Site D | 1,050 | 53.0% |
| Site E | 627 | 78.6% |
| Site F | 347 | 83.9% |
| Site G | 342 | 72.5% |
| Site H | 236 | 66.9% |
| Site I | 221 | 97.3% |
| Site J | 168 | 88.7% |
| Site K | 90 | 91.1% |
| Site L | 75 | 4.0% |
| Site M | 74 | 81.1% |
| Site N | 61 | 86.9% |
| Median of these 14 sites | 19,393 combined | 83.2% |
| All 33 sites combined | 19,509 | 81.1% |
The range runs from 4.0% to 97.3%. The 4.0% site is small, and nearly all of its clicks landed on queries that cleared Google's threshold for showing a query. At the 97.3% site, almost none did. The spread is the useful part. A site's hidden share tells you how concentrated its demand is. A low number means a few head terms carry the site. A high number means it lives on the long tail, and its query report will undercount it badly.
For context, Ahrefs' 2022 study of hidden terms looked at one month of data across 146,741 websites and nearly 9 billion clicks and found 46.08% of clicks had no associated query. Our portfolio sits far above that. We cannot separate the causes cleanly, but a sample of small, long-tail sites over a 90-day window is the setting where you would expect the gap to be widest.
The desktop impression gap.
The visible rows hold a second distortion. When we split non-brand query rows by device, desktop carried three quarters of the impressions and a third of the clicks.
| Device | Clicks | Share of clicks | Impressions | Share of impressions | CTR |
|---|---|---|---|---|---|
| Mobile | 2,230 | 64.2% | 342,392 | 23.1% | 0.65% |
| Desktop | 1,159 | 33.4% | 1,110,375 | 74.9% | 0.10% |
| Tablet | 82 | 2.4% | 30,719 | 2.1% | 0.27% |
| All devices | 3,471 | 100% | 1,483,486 | 100% | 0.23% |
The gap holds at every page-one position. Desktop CTR on positions 1 to 10 ran from about 0.06% to 0.17% depending on the position. Mobile ran from about 0.3% to 2.4%. Real desktop searchers do click, so a desktop CTR that low at good positions suggests many desktop impressions were never seen by a person.
Our likely explanation, which we cannot prove from Search Console data, is automated traffic. Rank trackers, AI visibility tools and scrapers run searches on a schedule, and if they load results pages as desktop browsers, every result they load can count as a desktop impression with no click behind it. We documented one visible form of this in Ghost Prompts, where templated AI prompts earned 1,965 impressions and 1 click across 25 sites. The desktop gap is the same effect seen from a wider angle.
The zero-click top 3.
Ranking in the top 3 is supposed to be where clicks happen. In our visible non-brand data, queries with an average position under 3.5 earned 66,222 impressions and 261 clicks, a 0.39% CTR. Then we looked at which queries those impressions belonged to.
Most top-3 impressions sat on queries that never produced a single visit in three months. A dashboard that celebrates "impressions in the top 3" is mostly celebrating these. Some are real people who got their answer on the results page. Many, judging by their length and wording, are not people at all.
Query length vs CTR.
Length separates the two groups. Short queries in the top 3 got clicked at a normal-looking rate. Queries of five or more words, which held most of the top-3 impressions, barely got clicked at all.
| Query length | Clicks | Impressions | Share of top-3 impressions | Share of top-3 clicks | CTR |
|---|---|---|---|---|---|
| 1 to 2 words | 110 | 5,134 | 7.8% | 42.1% | 2.14% |
| 3 to 4 words | 97 | 22,785 | 34.5% | 37.2% | 0.43% |
| 5 to 6 words | 44 | 31,285 | 47.3% | 16.9% | 0.14% |
| 7+ words | 10 | 6,932 | 10.5% | 3.8% | 0.14% |
Share of impressions uses the 66,136 impressions that fell into a length bucket. Queries of 1 to 2 words were 7.8% of top-3 impressions and 42.1% of top-3 clicks. Queries of 5 or more words were 57.8% of impressions and 20.7% of clicks. A 2.14% CTR against 0.14% is a 15-fold difference at the same ranking band.
For scale, branded queries at position 1 on our sites had a 3.90% CTR (77 clicks on 1,972 impressions). That is a small sample from young sites with little brand demand. We include it only as a reference point for the non-brand rates above.
Where clicks come from by position.
The usual assumption is that the top 3 positions drive most organic clicks. In our visible non-brand data, positions 4 to 10 drove two thirds of them.
| Average position | Share of non-brand clicks | CTR |
|---|---|---|
| 1 to 3 | 7.5% | 0.39% |
| 4 to 10 | 66.8% | 0.28% |
| 11 to 20 | 20.9% | 0.22% |
| 21+ | 4.7% | 0.06% |
Two things drive this. Query-level average position blends every position a query held over 90 days, so a query that bounced between 2 and 9 lands at 5 or 6. And the top-3 bucket is full of the zero-click long queries above, which drags its CTR close to the 4 to 10 bucket. The practical point: on long-tail sites, the queries that pay are often the ones sitting on the bottom half of page one and the top of page two, which are also the cheapest to improve.
What this means for SEO reporting.
If four in five clicks have no visible query, a query report cannot be your source of truth for traffic. It can still be your source of truth for intent, as long as you know what it leaves out.
- Report traffic from page-level data. Google says anonymized queries stay in the chart totals unless you filter by query, so page, country, device and date views keep their clicks while query rows do not. Use page totals for traffic trends and the query table for themes.
- Never sum query rows and call it total traffic. On our data that sum would have shown 3,680 clicks instead of 19,509. A keyword-level rank report that adds up its keywords has the same problem.
- Segment by device. Report mobile and desktop CTR separately. A blended CTR on our data was 0.23%, which hides a 0.65% mobile rate and a 0.10% desktop rate.
- Separate machine-like long queries. Put queries of five or more words in their own segment. Our ghost prompt filter handles the templated ones. Keep them out of the headline CTR and average position.
- Read the hidden share as a strategy signal. A rising hidden share usually means traffic is spreading across more long-tail queries. That is often good news, and the query table will under-report it.
- Tie SEO to visits and revenue. Search Console tells you about Google. Your analytics tells you what the visits did. Tag non-Google sources with UTM parameters so the two can be reconciled.
This also changes how you judge a new site. Young long-tail sites are where the hidden share runs highest, so their query reports look thin even when traffic is growing. If you are setting expectations for a new program, our guide to how long SEO takes covers what to watch in the early months.
How to measure your own hidden share.
The interface will not do this for you, because filtering or exporting by query drops anonymized queries and caps the export at 1,000 rows. The Search Console API gives the cleanest number. This is the method we used.
type to web. We used 90 days.searchanalytics.query for the property with your startDate and endDate and no dimensions field. The single row it returns holds total clicks and impressions, including anonymized queries.dimensions: ["query"], rowLimit: 25000 and startRow: 0. Repeat with startRow raised by 25,000 each time until a response comes back empty. We added device as a second dimension to split by device, which gives the same click sum.How to cite this study.
Theory Road, Search Console Hidden Queries Study, November 2026, theoryroad.com/insights/search-console-hidden-queries-study. Please cite the source and the window (August 17 to November 14, 2026, 33 websites) when you use the numbers.
Questions about hidden Search Console queries.
What share of clicks does Search Console hide?
On 33 websites we own and operate, 81.1% of Google Search clicks over 90 days (August 17 to November 14, 2026) had no visible query, and the median site with 50 or more clicks hid 83.2%. Ahrefs' 2022 study of 146,741 websites found 46.08% across one month. The share depends heavily on how long-tail a site's traffic is.
What is an anonymized query in Search Console?
Google defines anonymized queries as those not issued by more than a few dozen users over a two-to-three month period. Their clicks and impressions count in the chart totals, but the query text is not shown, and they drop out of the totals as soon as you filter by query.
Why do my Search Console query rows not add up to my total clicks?
Two reasons. Anonymized queries are counted in the total but have no row. And exports are capped: 1,000 rows in the interface and 50,000 rows per day per site per search type in the API. On small and mid-size sites, anonymized queries are usually most of the gap.
Can I see the hidden queries anywhere?
Not in Search Console. Page-level data includes their clicks, so you can see which pages earn the traffic, but not the exact query text. Google Ads search term reports and on-site search can show related wording, but they are different data.
Why is desktop CTR so low in Search Console?
In our data desktop held 74.9% of visible non-brand impressions and 33.4% of clicks, a 0.10% CTR against 0.65% on mobile. The likely cause is automated searches from rank trackers, AI visibility tools and scrapers that load results pages without clicking. We cannot confirm the source from Search Console data.
Should I stop using the Search Console query report?
No. Use it for intent and topic research, and use page-level totals for traffic. Segment by device, separate long machine-like queries, and never present the sum of query rows as total traffic.