12,000 impressions, 40 clicks. The Search Console performance report puts those numbers side by side and explains nothing. Twelve thousand times your page appeared in the results, forty times somebody clicked: a CTR of 0.33 percent. Anyone who has sent a newsletter knows friendlier rates, so the first thought is almost always that something is broken.
Nothing is broken. There are exactly four reasons for that gap, two that let you off the hook and two that mean work, and they can be ruled out in a fixed order. This piece walks them in order and ends with a work list built from your own account instead of a benchmark table.
What Search Console actually counts as an impression
Before you interpret CTR, know what sits in the denominator. Google counts an impression when someone saw, or could have seen, your link. The "could have" is the whole point.
The rule: if your result appears on the currently displayed page of results, the impression counts, whether or not anyone scrolled down to it. Someone on a phone who reads the top three and clicks the first still handed your result at position 9 an impression. The rule only stops applying when a click is needed to reach the result: for page two, the impression counts only once somebody pages over. Inside carousels and expandable elements, your entry has to be scrolled into view or expanded.
That settles the most important thing about CTR: on page one, "impression" does not mean seen, it means potentially seen. Everyone who never scrolled down drags your CTR down without registering your result. So 0.33 percent at position 9 says nothing about your title and a lot about the screen height of your visitors. One oddity: the rows in the Queries tab never add up to your total impressions, because Google omits rare queries for privacy.
Average position 8.4 does not mean you sit at 8
Average position is the most consistently misread number in the report. It is not a ranking. It is the impression-weighted mean of the topmost position your link occupied, counting only queries where you appeared, and only your best result when several of your pages rank for one query.
A plumbing firm in Manchester gets 1,000 impressions a quarter for one query. Two distributions, same mean:
- Distribution A: 800 impressions at position 9, 200 at position 6. Mean 8.4.
- Distribution B: 400 impressions at position 3, 600 at position 12. Mean 8.4.
Two completely different situations. In A you sit stably in the back half of page one, and four positions forward is a clean gain. In B almost all your clicks come from the 40 percent of impressions in the top 3, while the other 60 percent sit on page two and contribute nothing. A calls for ranking work, B for stability, and the mean does not distinguish them.
This also explains a pattern that confuses people: thousands of impressions at a double-digit average position. If page two only counts when somebody pages over, they must come from the days and query variants where you sat on page one after all. Read the position column per query over time, never at page level.
Cause one: your position is double-digit, and then the CTR is normal
Clicks are not spread evenly down the results page, they fall off steeply. Roughly, varying by query type: position one takes around 27 percent of clicks, position three about 11 percent, position eight roughly 3 percent, position fifteen about 1 percent. The figures vary by source. The shape of the curve does not. So a thousand impressions at position eight yields about 30 clicks, at four about 100, at fifteen about 10. In all three cases the CTR is exactly what you should expect.
This is why the order of checks matters. If your average position is 11 or higher, you do not have a snippet problem, you have a position problem. Rewriting the title of a page at 14 is the most commonly wasted hour in SEO: even a spectacular headline doubles a CTR of 0.8 percent to 1.6 percent, which on 3,000 impressions is 24 clicks a quarter. The lever sits one level down, in visibility itself, and how to pull it is in increasing website traffic.
Cause two: you rank for a question your page does not answer
You only find this cause by looking away from the numbers and reading text. Open the Queries tab, sort by impressions descending, and read the top thirty rows.
A coffee roaster in Portland had 8,400 impressions in a quarter for variants of "how to roast coffee at home", CTR 0.4 percent, average position 6.1. Good position, wrong intent: Google was serving a shop category page against a how-to query. Nobody looking for instructions clicks a product listing, and whoever does is gone in seconds.
Here the right answer is explicitly not "rewrite the snippet". A better title on a query you cannot serve only buys a visitor who bounces. Two sensible reactions: ignore the row, because that demand is not yours, or answer the question for real, which means working on content rather than the snippet.
Causes three and four: the snippet and zero-click search
Cause three: the snippet does not sell. Narrow case: position between 3 and 7, the intent matches the page, and the CTR still sits well below what your other rows at similar positions bring in. Only then is the result text worth touching. What works in titles and descriptions is in writing titles and meta descriptions.
Cause four: the result page eats the click. Above your listing sit an AI Overview, a featured snippet, a map pack, an image carousel and two ads. The question is answered before anyone reaches you. The pattern is called zero-click search, and the crucial point is that the impression still counts: your result was there, further down, potentially visible.
That is why it looks like a collapse although your ranking never moved. A heating installer held roughly 6,200 impressions a quarter for a grant query at 4.8. Clicks fell from 240 to 95, CTR from 3.9 to 1.5 percent, position and impressions stayed put. That signature, position flat, impressions flat, clicks gone, is the fingerprint of zero-click search. If the impressions fall too, it is a genuine ranking loss and belongs in the diagnosis from lost rankings.
Check it two ways. Type the query yourself, ideally logged out, and see what sits above your result. And open the Search appearance tab, which lists the special result types your pages appear in. For AI Overviews and AI Mode, Google has added a generative AI performance report, which shows impressions only and is not in every account yet; those impressions were inside your totals all along. What to change so you show up there is in Google AI Overviews.
Why comparing your CTR to industry benchmarks wastes your time
The most common advice here: compare your CTR to your industry average. Well meant, useless, and the reason is sitting in your own account.
Your property's overall CTR blends two populations that have nothing to do with each other. First, brand queries: people typing your company name because they were coming to you anyway, whose CTR often sits past 30 percent. Second, informational queries, where you are one of ten results and land in the low single digits. A shop with a known name shows 8 percent, a young B2B site 1.2 percent, and both can have equally good snippets. The number mostly reports your share of brand queries, unknown in every benchmark table. It produces false calm or false panic.
The way out is unglamorous. Filter the brand out: open the Queries filter, pick Custom (regex), choose "Doesn't match regex", and enter every spelling of your name, like brand|brandname|brand\.com. Regex matching ignores case by default; (?-i) at the front makes it case-sensitive. What remains is your CTR in open competition. Split it into bands (position 1 to 3, 4 to 10, 11 and up) and divide clicks by impressions. Rows well below their band value are your candidates.
Building the work list from the performance report
A warning first, so you do not go hunting: Search Console has no metric filter. You cannot filter on "position between 4 and 15", however often guides say so. You sort, export, and filter in the spreadsheet.
The path: open Performance, report Search results, date range three months (six for small sites), Queries tab, then Export. You get at most 1,000 rows, taken by whichever column the table was sorted on, so sort by impressions first.
Four steps in the spreadsheet: delete the brand rows, narrow to positions 4 to 15, set an impressions floor (maybe 200 a quarter on a small site) so you are not working on noise, and add a "clicks left on the table" column: impressions times your band CTR, minus actual clicks. Sorted by that, the top row is not the biggest but the furthest from your own normal.
Then assign each row to a cause:
| What the row shows | Likely cause | What you do |
|---|---|---|
| Position 11 or worse, low CTR | Position | Visibility work, leave the snippet |
| Query does not match the page | Intent | Ignore it or write a real answer |
| Position 3 to 7, intent fits, CTR too low | Snippet | Rewrite title and description |
| Clicks gone, position and impressions flat | Zero-click | Read the SERP, check your answer |
Merging rows and keeping the 4 to 15 band in view month after month is what Traffalyzer takes over.
Where to start
- Check the average position. Above 11: stop, nothing is broken, the job is visibility.
- Open the
Queriestab, sort by impressions, read the top thirty rows. If the intent does not fit, CTR is the wrong metric. - Type two or three important queries into Google and see what sits above you.
- Only now touch titles and descriptions, and only on rows at position 3 to 7 with matching intent.
- Set up the export with the "clicks left on the table" column, then repeat in four weeks. Without a before, you know nothing after.
If individual pages have no impressions at all, CTR is the wrong topic and indexing is the right one: see page not indexed.
Frequently asked questions
Why do I have impressions but no clicks?
Usually because your average position is in double digits. Clicks fall off steeply between position one and ten and then drop to almost nothing, so a CTR below one percent there is normal, not a defect. The other three causes: you rank for a query your page does not answer, your title fails to convince, or an AI Overview answers the question above your listing. Check them in that order, because the first explains so many cases that snippet work before it is wasted.
What is a good CTR in Search Console?
There is no useful benchmark, and that is not a dodge. Your overall CTR depends almost entirely on the share of brand queries in your impressions, and that share varies enormously: a known brand lands at 8 percent, a young specialist site at 1.2 percent, with equally good snippets. Compare your own CTR per position band against itself over time instead.
Why do I get zero clicks despite a good position?
If impressions and position are stable and only the clicks are missing, something above your result is eating it: an AI Overview, a featured snippet, a map pack, ads. If the position looks good but the queries have nothing to do with what you sell, it is intent, and the missing click is correct. Remember too that "good position" is a mean: 6.0 can mean you are usually at 3 and occasionally at 12.
What does average position mean in Search Console?
The impression-weighted mean of the topmost position at which a link to your page appeared. Queries where you did not appear are not counted, and when several of your pages rank for one query, only the best counts. So the number is not a ranking: 8.4 can mean you are nearly always at 9, or that 40 percent of your impressions are in the top 3 and the rest sit on page two.
Why don't Search Console clicks match GA4?
Because the two measure in different places. Search Console counts clicks in the results, GA4 counts sessions, which only exist once the page loads and the measurement script runs. In between sit abandoned loads, redirects, ad blockers and above all people who decline in the cookie banner. A gap of around twenty percent is normal; sudden jumps deserve a look. Linking the accounts is covered in connecting GA4 and Search Console.
How do I improve my search result CTR?
Only on the rows where CTR is genuinely the problem, meaning position 3 to 7 with matching intent. What works most reliably: a concrete number in the title, a year on anything that can go stale, and a promise specific enough to sound checkable. Superlatives and exclamation marks do not. Structured data such as reviews, prices or opening hours will not change your position, but it makes your result bigger than a competitor's in the same spot.
Do AI Overviews count as impressions?
Yes, with one caveat. Links inside an AI Overview follow the standard rules, so the link has to be scrolled or expanded into view. Your classic result further down counts as always. Which is why an AI Overview looks like a CTR collapse: the impressions stay, the clicks leave. Google now breaks these out in a generative AI performance report, so far without clicks and not yet in every account.
Why are my impressions rising but not my clicks?
Usually because the extra impressions come from positions that produce no clicks. Google is testing you on more queries, you appear at position 14 to 40, impressions grow and the CTR drops as pure arithmetic. Compare two date ranges and read the new queries: if topics appear that you have no matching page for, you have just been handed a content list for free.