Optimise CTR: the click-through rate in search results, upstream of conversion

Your impressions are rising, your clicks are not following. Before asking what a page converts, you need to know how many people reach it: that is what the click-through rate in search results measures, and it measures nothing else.

This page is for the executive or marketing director who wants to optimise CTR on their pages: what to compare a click-through rate with, which pages to rework, what to rewrite and how to check that a gain is real. What becomes of the click afterwards belongs to a good conversion rate.

What CTR measures, and how far it measures

CTR stands for "Click Through Rate". It tells you whether what you display in a search result makes people want to click.

This indicator sets the number of clicks on a piece of content against the number of impressions, that is, the number of times the link was seen. In SEO, the clickable element is your page as it appears in the SERP: its title, its URL and its description. Those three lines are your first lever; structured markup, covered below, can add format elements to them.

With 10,000 impressions and 100 clicks, the CTR is 0.01, or 1%. In Search Console, there is no deduplication by visitor: a person who sees your result three times and clicks twice produces three impressions and two clicks. It is a measurement of display yield, not of audience.

That places CTR in the chain that leads from a search to a conversion. Each link depends on something else:

• Impressions depend on position and search volume.
• Clicks depend on what the result displays.
• Conversions depend on what the page delivers once the visitor has arrived.

CTR stops at the second link: it is measured before arrival on the site, whereas the conversion rate sets conversions against sessions and is measured after. A page can convert well without receiving clicks, another can capture every click and convert nothing: two problems, two levers.

For an e-commerce site, a CTR of 5% together with a conversion rate of 10% is better than a CTR of 30% with no sale at all (fictitious example): the first result promises what the page delivers, the second does not.

What happens after the click is handled with conversion rate optimisation and the page levers it covers, and the choice of pages to rework with the method for optimising the conversion rate. Here, we stop at the click.

What a SERP click-through rate is compared with

"What is a good CTR?" The usual answer, a market average, is the least useful: in search results, a click-through rate cannot be read without its rank. A page in seventh position at 4% and a page in first position at 12% are not judged by the same yardstick. The decline of the click-through rate by rank gives that benchmark, position by position:

These aggregated averages are a benchmark for placing yourself, never a target or a threshold. A rate below the average for its rank is not a fault to correct. It may come from several causes:

• a less commercial query;
• a particular device mix or country;
• the result format displayed around your link;
• the absence of brand queries on your page, where the average does contain them.

A comparison is therefore worth something only if it holds the rank, the share of brand queries, the device, the country and the SERP format constant. Failing that, the benchmark serves to place yourself, never to diagnose.

Two variables move that benchmark. First, the type of result displayed around you: an AI overview, a featured snippet or a local pack pushes your link towards the bottom of the screen without your position moving. Then intent: a query that the results page answers directly produces fewer clicks than a query that requires visiting a site.

Reputation escapes any rewrite: a well-known brand logically has a CTR above average, because the reader recognises the domain before having read the title. Hence a division that is better stated plainly:

• You have a hold on the title, the description and the markup of your pages.
• You have no hold on your position on the day, the reputation of your brand or the type of result the engine chooses to display.
• Everything that follows bears on the first category.

What generative answers do to the click-through rate

The share of informational searches giving rise to an AI overview is 58% (SEO.com, 2026): on those queries it is the ordinary case, and that explains a volume of clicks falling at unchanged position.

The click-through rate for the first position in the presence of an AI Overview is 2.6% (Squid Impact, 2025). It cannot be compared directly with the first-position average given in the table above: not the same source, nor the same year, nor the same sample. Each value holds for the condition it describes and is not subtracted from the other.

Work on position keeps its point, and this is what is forgotten fastest. 99% of AI Overviews cite the organic top 10 (Squid Impact, 2025): citation is taken from the top of the results, where the click is also won.

These records allow four readings, and no more:

• The click falls: the click-through rate of position 1 falls clearly when an overview is displayed.
• Position remains useful: the top 10 is where both the click and the citation are won.
• Citation exposes without a click: whether that exposure makes up for the loss of clicks remains to be established site by site.
• Slipping is not measured: the record says where the cited pages come from, not what becomes of a page that drops out of the top 10.

The method does not change, the reading does. A fall in clicks at constant position, on informational queries where an overview has settled in, is not a fall in performance: a share of the value moves from the click to the citation. Showing it means separating in the record the queries that trigger an overview from those that do not. That is the first breakdown to set up.

Spotting the pages to handle: high impressions, low rate

Prioritise the URLs that have many impressions and a click-through rate far from the average for their position: that gap designates a page to examine, never an established fault. Then target the queries where you are already on page 1, where a snippet rewrite reads faster. What remains is producing the list.

The report is taken from Search Console, in search performance, in four steps:

• Display the four columns together: clicks, impressions, CTR, average position.
• Read by page, then by query: the first view designates the URLs, the second says what each one promises and provides the material for the rewrite.
• Set an impression floor and filter before sorting.
• Consolidate by type of SERP and by intent: with or without an AI overview, informational or transactional.

The window must absorb weekly variations without diluting average position. For the click-through rate, three rolling months suit most sites: it is a window specific to CTR, which does not replace the reading cadence of the conversion rate. Always compare windows of the same length.

Below a few hundred impressions, a click-through rate is not readable: one click more or less moves it by several points, and you would be rewriting titles on noise. That floor is a practitioner's order of magnitude, to be calibrated on your volume. Consolidating by type of SERP, for its part, will let you say whether a gain comes from your rewrite or from the disappearance of an overview.

Each signal recorded translates into a decision, and it is better to have written it down before starting.

The second line is the one handled worst: a page whose click-through rate is normal but which receives few impressions does not have a snippet problem, it has a visibility problem.

What gets rewritten: title, meta description, structured data

Two levers act on what is displayed, and they are not of the same nature. They are handled in the order of the table:

The title and the meta description

Optimising CTR sits within an SEO strategy. By offering content that matches the search intent of the user, you are acting on what depends on you; the improvement in the click-through rate, for its part, is observed after the fact and is not promised.

Four moves concentrate most of the gain available:

• add a clear benefit to the title;
• use a differentiating element: a figure, a year, a method;
• align the wording with the intent as your entry queries reveal it, not as you imagine it;
• rewrite the meta description as a promise, a proof and a small call to action.

The interrogative form deserves to be tried when the query is itself a question: the average CTR for a title containing a question is +14.1% higher (Onesty, 2026).

An example makes the move concrete. Before: "CTR Optimisation | Site name". After: "Optimise CTR in the SERP: benchmarks by position and recording method". The first announces nothing the URL does not already display; the second names the subject, says what will be found on the page and carries a differentiating element.

Note the initial wording, the publication date and the average position of the moment before changing anything: without that dated before-state, there will be nothing to compare. Three limits then frame any result announced:

• Google rewrites the titles and meta descriptions it displays frequently, from the content of the page and the query: what you write is a proposal.
• Rich results depend on an eligibility that the engine judges alone: valid markup is its condition, never its guarantee.
• Search Console is not an A/B testing tool: it records a single reality over time, with a rank, a season and a competitive field that move of their own accord.

Structured data and rich snippets

The second lever does not change the text of the result, but its format. Structured markup lets the engine display, on top of the title and the description, elements that your SERP neighbour does not display:

• rating stars;
• a price;
• a list of questions;
• a breadcrumb.

These snippets take up more room on the screen and give one more reason to choose your line. The markup to put in place and what it produces are set out with structured data and the formats it triggers.

The featured snippet follows the same logic, with one reservation: it is not requested, it is earned. The click-through rate on featured snippets is 6% (SEO.com, 2026), an order of magnitude of what a result format captures on its own. Your chances of being picked come down to the content: a short, self-contained answer, below the section heading that asks the question, in the format the engine already displays.

Checking that a CTR gain is a real gain

A higher CTR does not necessarily mean more business. This is the check most often neglected, because the column has gone up and that makes a good slide. To validate an optimisation, record four indicators alongside it:

• the conversion rate of the page;
• the bounce rate or engagement;
• the value per session or per lead;
• the average position.

The objective is to increase qualified clicks, not volume alone. Record all four over the same window as the click-through rate, before the change: an initial state reconstructed after the fact compares nothing.

The comparison protocol decides what you will be entitled to conclude. It sets five conditions:

• the same page and the same queries, with brand queries excluded from the record;
• two windows of equal length;
• a comparable average position from one window to the other, otherwise the gap in rate is explained first by the rank;
• the same device and the same country, by filtering rather than aggregating;
• the same SERP format, with queries with and without an overview separated.

If one of the five conditions is not met, the comparison is not made: you wait, or you choose another page.

Average position serves as a non-regression check: a click-through rate that rises along with the position says nothing about your rewrite. Also check that the type of result displayed is the same and that no redesign has come to overlap between the two measurements.

What happens after the click is measured with other breakdowns (channel, entry page, device), which are absent from Search Console. The click-through rate tells you that people are coming in; it will never tell you what they do next.

Continue reading

Depending on what you are missing after this record, four pages take over:

• You want to produce the report described above yourself: without the four columns, all the spotting waits on Google Search Console, with its impressions, clicks and positions by query.
• You have gained clicks and have to say what becomes of them: the question moves from the click to the action, with other indicators and other breakdowns, those of the conversion rate KPIs and their segmentation.
• You have to show that these clicks produce conversions: if you cannot attach a conversion to organic or to generative search, it is attribution that is blocking, covered in what SEO does to conversion.
• Behaviour on the page escapes you: the measurement tool on the site side is missing, and what Google Analytics lets you track fills that gap.

To know whether a rewritten snippet has produced an effect or whether it is the format of the result that has changed, the Incremys reporting module tracks CTR by type of SERP (with or without an AI Overview) to steer the optimisation of your snippets.

Frequently asked questions

What is CTR and how do you calculate it?

CTR (Click-Through Rate) is the ratio of the number of clicks to the number of impressions, expressed as a percentage: (clicks / impressions) × 100. In organic results, the average click-through rate for the first position is 27.6% (Backlinko, 2026). A good CTR depends on the rank, the format displayed and the intent.

How do you optimise CTR in SEO?

You optimise CTR by rewriting the title and the meta description, then by working on the format of the result through structured data and short answers that aim at the featured snippet. The gain is never guaranteed: Google often rewrites what it displays, rich results depend on its judgement, and Search Console does not compare two versions like an A/B test.

What impact do AI Overviews have on CTR?

An AI Overview clearly lowers the click-through rate of the first position, especially on informational queries, and much less on transactional ones. Being cited in the overview produces an exposure that the click record does not show; whether that exposure makes up for the loss of clicks remains to be established site by site.

Which quick optimisations raise CTR in the SERP?

The quickest optimisations bear on pages already on page 1: a clear benefit and a differentiating element in the title, wording aligned with the real intent of the queries, and a meta description built as a promise, a proof and a call to action. Structured data comes next, with no guarantee of display.

Which indicators should you track alongside CTR to avoid false gains?

You should track the conversion rate of the page, engagement, the value per session or per lead and the average position, over the same window as the CTR and from an initial state noted before the change. A rise in the click-through rate accompanied by a rise in position proves nothing about the rewrite.

How do you prioritise the pages to optimise in order to improve CTR?

Prioritise the pages that combine many impressions with a click-through rate far from the average for their rank, starting with those already on page 1. First filter out pages below an impression floor, then separate queries with and without an AI overview, by intent, device and country: without those cuts, two pages are not comparable.