What eight Search Console accounts say about click-through rate

Every so often a finding turns up in client data that we can't check against anything published, because nobody publishes the thing we'd need. So in August 2026 we sat down with the Search Console exports for every account we could legitimately look at, eight of them, and ran the same question across all eight at once.

The question was simple. Does organic click-through rate follow how visible a site is, or does it follow how specific the searches are?

We're publishing what came back because the raw data isn't available anywhere else, and because one of the eight results is embarrassing for us. What we're not doing is presenting it as a rule. Eight accounts we happen to have access to isn't a study, so there's a section below on what this can't tell you and it's worth reading before you quote anything from the table.

The finding: click-through rate tracked the specificity of the search, not the volume of impressions. The accounts serving the broadest questions converted worst, however many impressions they had collected.

What we looked at, and what we didn't

The method, so you can judge it or repeat it.

The sample. Eight Google Search Console properties: sites we currently work on, sites we used to work on, and our own. Not all of that work is SEO. They span the UK and the US, e-commerce and local services, healthcare and trades, and businesses from a few months old in search to several years in. We did not select them for their numbers. They're the accounts we have access to, which is the single biggest weakness in what follows.

What none of it is. None of these figures is a claim about our work. Several of the windows include long stretches before we were involved at all, and one account we haven't worked on for months. The comparison between the accounts is the point, not credit for any of them.

The period. Twelve months to 6 August 2026, except the heating and plumbing account, where sixteen months were available and we used all of it. Every other figure on this page, including all three accounts we break down below, covers the same twelve months.

The metric. Account-level clicks divided by impressions, taken from the Performance report with no filters applied: all queries, all countries, all devices, brand searches included. That's a deliberately crude measure and we've kept it crude so the comparison stays like for like.

What we did not do. No control group, no matched pairs, no adjustment for market or competition, and no attempt to separate our work from everything else happening at each business. This is a description of eight accounts at one moment, not an experiment.

The eight accounts

Accounts are anonymised except our own. We don't put client names next to numbers unless the result is theirs to be proud of and they've agreed to it, and applying that rule selectively would be worse than not applying it at all.

What the table says

The gap between the top and the bottom of that column is roughly seventy to one, and it doesn't line up with effort, spend, or how long we've worked on each account.

It lines up with how specific the searches are. Take two of them. The equipment hire business at the top took 1,027 clicks from 8,121 impressions across the year. The home improvement account took 484 clicks from 203,697 impressions. The second site was seen twenty-five times more often and got less than half the clicks.

Nothing about the second site is broken. It's being served against people at the reading-and-comparing stage, and being served against a great many of them. The first site is being served against people who have decided what they want and are looking for somewhere to get it. Impressions of the second kind are cheap to accumulate and mostly don't convert into anything.

The middle of the table is where it gets more interesting. The e-commerce account carries one of the largest impression counts in the sample, 541,225 of them, and still returns 1.47% and 7,943 clicks in the year. Volume didn't hurt it, because the volume is people wanting to buy the product it sells. Whatever's driving the spread, it isn't impressions on their own.

The account at the bottom is ours

consultico.co.uk has the worst click-through rate of the eight at 0.18%.

We've written before about six near-identical pages of ours where one earned 5,926 impressions and another earned 6, and about what our own query data showed when we split it by intent. This is the same problem measured a different way. We rank for comparison and research terms, on the most saturated search results pages there are, where our result is one of ten agencies and the person searching is browsing rather than buying.

That's a diagnosis of our own content strategy, and it's the reason the terms we go after are changing. It would be strange to publish this table and leave ourselves off it.

Where the pattern gets messy

Three things in this data cut against a clean story, and they're worth more than the headline.

A low rate doesn't mean a bad account. The heating and plumbing account at 0.22% was one of the best engagements we've run, judged on booked jobs rather than clicks. The clicks it did get were people with a broken boiler, and enough of them turned into work to fill a diary. If we'd managed that account on this metric we'd have made it worse.

Brand searches flatter some of these numbers. In the healthcare practice, brand queries produced 67% of the clicks from 4% of the queries. Strip brand out and its non-brand rate is 0.25%, roughly a third of what the table shows. We didn't filter brand out of any account, so every row here is a little generous in the same direction.

Position is a competing explanation and we can't rule it out. One of the low-rate accounts watched its average position slide from 20.8 to 39.3 over the period. A site sitting three or four pages deep will have a poor rate whatever the searches are about, so for that account rank and intent can't be told apart. The same objection applies more weakly across the sample, because the accounts answering narrow buying questions also tend to rank better for them. Eight accounts isn't enough to hold position still and look at intent on its own, and that's the real limit of this exercise.

There's one more thing worth recording. In the healthcare practice there's a page holding an average position of 1.0, served over a thousand times, with no clicks whatever. We'd found the same thing on our own site and assumed it was a quirk of ours. Seeing it on an unrelated site, in an unrelated sector, was the point at which we stopped assuming that. We can't yet tell a genuine click-through failure from a reporting artefact here, because a result counted as position 1.0 inside an AI summary or an entity panel would look identical in this report, and that ambiguity is itself worth knowing about.

Does anyone else's data agree?

Partly, and the part that agrees is the part that matters.

Sistrix analysed more than 80 million keywords and published the best-known curve of organic click-through rate by position, putting the top result at 28.5% (Sistrix, 2020). That figure is mobile only, the study was published in 2020, and Sistrix doesn't state which market it covers, so we wouldn't hold anyone to the number itself.

Their conclusion is the useful bit. They found the rate for position one ranged from 13.7% to 46.9% depending on what else was on the results page, and wrote that global averages across all types "aren't very useful" because search intention defines the layout and therefore how many organic clicks are available at all. That's a far larger dataset than ours, from a different direction, arriving at a compatible idea: what people are searching for does more to your rate than where you sit.

What eight accounts can't tell you

This is a convenience sample. The accounts are the ones we can see, which means they share an agency, a set of habits, and a rough profile of business. That's the definition of the kind of sample you can't generalise from.

Beyond that: eight is a small number, one account ran sixteen months against twelve for the others, the businesses differ in country, sector, age, budget and how long we've worked on them, and we've made no attempt to control for any of it. Search Console's own measurement is loose, as anyone who has compared it to analytics will tell you. And "narrow" and "broad" in that table are our judgement of each site's footprint, not a measured quantity, which is the softest thing on this page.

So the honest version of the claim is this. In eight accounts we happen to have access to, click-through rate tracked the specificity of the searches rather than the number of impressions, we can think of a mechanism for why that would be true generally, and rank is tangled up in it in a way we can't unpick at this sample size. That's a hypothesis you can test against your own data in an afternoon. It isn't a benchmark, and if someone quotes these figures at you as an industry average, they've misread them.

Run the same check on your own account

Twenty minutes in Search Console, no tools required.

1. Open Performance and set the range to the last twelve months. The account-level rate is your baseline.
2. Filter out your brand name with a "queries not containing" filter. The gap tells you how much of your performance is people who already knew you existed.
3. Sort queries by impressions and read the top thirty. Ask of each whether the person typing it is deciding what to buy or where to buy it.
4. Then find the pages with the most impressions and fewest clicks. They're earning visibility against questions they don't answer.
5. Compare the pages at either end. Not their word counts. What the reader had already decided before they searched.

If most of your impressions come from step three's first category, a low rate is a description of your keyword targets rather than a fault in your pages, and rewriting the titles won't move it.

What we'd do differently next time we run this

More accounts, and ideally some we don't work on, so the sample isn't selected by who hired us. Brand filtered out of every account rather than left in. A consistent window. And a proper measure of footprint instead of our own read on it.

We'll rerun it when we've got enough accounts for the answer to survive one of them being unusual. Until then this is eight exports, one afternoon, and a finding we're publishing mainly so somebody can check it against their own.

If you want to know which of those two groups your site is in, our SEO work starts with a look at exactly this data, and what SEO costs in the UK covers what happens after that. Or tell us what you're seeing and we'll give you a straight read on it.