ToolsThatRank / measured data
First party measurement
What pages below position 30 actually earn
Published CTR curves stop at the bottom of page one and estimate the rest. We own 34 Search Console properties that live down there, so we read it instead. Two bands, fifty ranking positions apart, came back almost identical.
Verified 2026-08-15 against the Search Console searchAnalytics endpoint. Window 2026-05-15 to 2026-08-12.
Try it on your own numbers
Band reading tool
Type the average position a page of yours records and the impressions behind it. You'll get what that band actually recorded across the fleet, pooled and with the biggest site taken out.
Nothing is fitted between bands. Position 14 returns the 11 to 20 figure and says so. A decimal position is rounded first, the same rule the dataset uses, so 30.6 reads as 31. Anything outside 1 to 100 returns no reading.
The measurement
Every band, with the counts behind it
Each rate sits next to the clicks and impressions it came from, because a rate on its own hides how thin it is. One property, belikenative.com, supplies 87.46% of the fleet's impressions, so both views ship. Read them together.
| Band | All 34 properties | Largest property removed | ||||
|---|---|---|---|---|---|---|
| CTR | Impressions | Clicks | CTR | Impressions | Clicks | |
| 1-10 | 0.317% | 1,387,666 | 4,399 | 0.7092% | 46,811 | 332 |
| 11-20 | 0.2559% | 415,841 | 1,064 | 0.5722% | 99,790 | 571 |
| 21-30 | 0.3088% | 62,831 | 194 | 0.4537% | 29,755 | 135 |
| 31-50 | 0.1108% | 56,871 | 63 | 0.1083% | 40,619 | 44 |
| 51-100 | 0.1119% | 34,866 | 39 | 0.1066% | 33,757 | 36 |
Highlighted rows are the two deep bands the page is about.
What the numbers do
Two things worth saying out loud
The deep end flattens. Take the biggest site out and the 21 to 30 band reads 0.4537%, while 31 to 50 reads 0.1083%. That's a fall of about four times across one band edge. Below it, the curve stops moving.
The shallow end isn't tidy either. Pooled, the 21 to 30 band sits above 11 to 20, which is backwards. Drop the dominant property and the order returns to normal, so that inversion belongs to one site rather than to the fleet.
We're not going to tell you why. AI Overview presence, query mix and intent could all be driving it, and this export measured none of them. Reporting the shape is a measurement. Explaining it would be a story.
Method
How we measured this
We measured these bands on 2026-08-15, over 2026-05-15 to 2026-08-12, against the Google Search Console Search Analytics API on 34 verified properties the operator owns. Two routes through the same endpoint had to agree before any figure here was allowed to render.
Route one
Query the endpoint with no dimensions and take the property level totals.
Route two
Query it grouped by page, put each page in a band by its rounded 90 day average position, then sum impressions and clicks inside the band.
When they disagree
The field is held and ships as a visible placeholder. We never publish the nicer of two numbers.
What a band means
Pages grouped by their average topmost position across all their queries. That isn't the same thing as a published CTR at rank curve.
The bug this caught
The first version of the export asked for one page of rows per property and took what came back. The largest property ran past the row cap, its rows were cut without an error, and the two routes then disagreed on band impressions by as much as forty five percent. Paging through every result set brought them back into agreement on every field. The cross check existed to catch exactly that kind of quiet failure, and it did, before anything was published.
Context
Where to get numbers that do represent the web
Our fleet isn't a sample of the web, so here's where to go instead. Backlinko's CTR study covers Google's first page organic results and draws on its own sample of pages and queries. Pew Research Center found that visits with an AI summary produced a click on a traditional result less often than visits without one, using browsing data from US adults over March 2025.
Impression counting itself moved recently. After Google stopped serving the num=100 search parameter in September 2025, an analysis of several hundred properties found a majority of sites lost impressions in Search Console. Our window sits after that change, which is one more reason to read these bands against each other rather than against a curve published before it.
Can you check this yourself
Not the raw export. Search Console is authenticated and the properties belong to the operator. What you can check is the method, the endpoint and every source cited here.
Limits
What this page won't tell you
It won't tell you what to do about a page sitting at position 40. Reporting that two bands came out close together is a measurement. Telling you to abandon or rewrite a page is business advice resting on one operator's 34 sites, and the pipeline doesn't ship that.
No ranking, traffic, or revenue outcome is promised.
The operator sells the software described here and owns every site in the dataset.
These figures come from 34 small sites in mixed niches, not from a sample of the web.
Search Console is authenticated, so no third party can verify these readings from outside.
Figures are read on the stamped date and change as Search Console reprocesses data.