Organic CTR is clicks divided by impressions for a page or query in Google search. Published benchmarks put position 1 near 28–32%, position 2 near 15%, position 3 near 10%, and everything past position 10 under 2%. Those averages are a fine sanity check and a poor benchmark: the only curve worth measuring against is the one your own site produces.
How to calculate organic CTR
The formula is plain:
CTR = clicks / impressions x 100
An impression is counted when your result appears in the results for a query; a click when someone opens it. Both numbers come from the Search Console performance report, and both are already filtered to organic — Search Console has no other kind of traffic in it.
Two mistakes cost more than the arithmetic:
- Averaging the averages. Adding up the CTR of ten pages and dividing by ten weights a page with 40 impressions the same as a page with 40,000. Sum the clicks, sum the impressions, divide once.
- Mixing brand with non-brand. Brand queries run several times the CTR of everything else at the same position. Leave them in and your site CTR mostly measures how many people already know your name.
Organic CTR by position: what the published tables say
Large-scale studies of Search Console data broadly agree on the shape, even where they disagree on the numbers:
| Position | Typical reported CTR |
|---|---|
| 1 | 28–32% |
| 2 | 14–16% |
| 3 | 9–11% |
| 4 | 6–8% |
| 5 | 4–6% |
| 6–7 | 3–4% |
| 8–10 | 1.5–2.5% |
| 11+ | under 1% |
The shape is the useful part: the drop from 1 to 3 is far steeper than the drop from 5 to 10, which is why moving a page from 4 to 2 is worth more than moving three pages from 20 to 12.
Why your curve will not match that table
A benchmark built from millions of queries across every industry describes no single site. Six things move your curve away from it.
- SERP features. An AI Overview, a featured snippet, a local pack or a shopping row pushes the first organic result down the page without changing its reported position. The same position 1 earns very different CTR on a bare SERP and on a crowded one.
- Brand share. People searching your name click your result whatever it looks like. A site with heavy brand demand shows a flattering curve that says nothing about its non-brand pages.
- Intent. Navigational queries concentrate clicks on one result; comparison queries spread them across several, because the searcher opens three tabs on purpose.
- Query length. Long, specific queries convert to clicks at higher rates than short head terms, where the searcher is still deciding what they want.
- Device. Mobile SERPs show fewer results above the fold, so the first result takes a larger share and everything below it takes less.
- Average position is an average. This one catches people out. Search Console reports the mean position across all impressions. A page that ranks 3 on Mondays and 15 the rest of the week reports 9 — and behaves like neither. Judging it against a position-9 benchmark tells you nothing.
Build your own CTR curve from Search Console
Your own data already contains the answer. The method is short:
- Export page and query rows with clicks, impressions and position for a period long enough to be stable — eight to twelve weeks for most sites.
- Split brand from non-brand before anything else, and build two curves if brand is a meaningful share.
- Bucket by rounded position and require a minimum number of impressions per bucket. A bucket holding 30 impressions is noise wearing a percentage sign.
- Take the median, not the mean. One viral query at position 7 will drag a mean CTR up and make every honest page at position 7 look like it is failing.
- Force the curve to decline. Real data wobbles — position 6 can come out above position 5 by accident of sampling. Smoothing the curve so it never rises as position worsens removes that artefact without inventing numbers.
- Hold the page out when you judge it. If a page is large enough to shift the curve, comparing it to a curve it helped build compares the page to itself. Rebuild the benchmark without that page, then measure the page against it.
That last step is the one most CTR audits skip, and it is why big pages so often look "average" — they are being measured against their own reflection.
So what is a good CTR for organic search?
A good organic CTR is one at or above your own curve for that position, that device and that query type. There is no absolute number. In practice:
- 4% at position 8 on non-brand informational queries can be excellent.
- 12% at position 2 can be poor, if your curve says that slot usually returns 18%.
- A site-wide CTR figure on its own is close to meaningless, because it moves whenever the brand-to-non-brand mix moves.
Turning the gap into a work queue
Once a curve exists, every page gets an expected CTR, and the difference becomes a number in clicks rather than percentage points:
clicks at stake = impressions x (expected CTR - actual CTR)
Sorting by that figure puts the work in order. A page 1 point below its curve on 200,000 impressions matters more than a page 8 points below on 900. Our CTR Lens builds the curve from your Search Console data and reports the gap per page; Lost Clicks ranks the same gap by the clicks it is costing.
How to improve organic CTR once you know where to look
- Match the title to the query intent, not to the page name. A searcher comparing options wants to see that the page compares options.
- Write the description as the second half of the title, not a summary of the page. It is advertising copy that happens to be indexed.
- Keep the title short enough to survive truncation. Anything past roughly 60 characters is at risk, and a title cut mid-word reads as carelessness.
- Earn the features on that SERP. If the results are full of FAQ rich results and yours has none, you are competing with less surface area.
- Check what the SERP rewards before rewriting. If eight of ten results are listicles, a landing page will keep underperforming its position no matter how the title is phrased.
Re-measure four to six weeks after a change, against a curve rebuilt on the new period. Snippet changes take time to be recrawled, and comparing against a stale benchmark manufactures results that are not there.