Click-through rate is the only number on your dashboard that three different trades can push up without improving anything.

Organic search does it by rewriting a title to promise more than the page delivers. Paid advertising does it by narrowing targeting until it only reaches people already convinced. Email does it by removing the people who no longer open.

All three move the ratio within a day. None of them produces one more customer.

A second problem sits underneath, and it is less often stated. At the volumes of a small company’s site or campaign, most of the month-to-month variation you observe is not variation at all. It is fluctuation, and the decisions taken on it are arbitrary.

This page covers both: what raises the ratio without changing anything, and from what point a gap deserves your attention.

One formula, three trades, three different things

Clicks divided by impressions, times one hundred. The formula does not change between contexts. What it measures does.

In organic search. The denominator is the number of times your page appeared on a results page. You control neither the queries that trigger it, nor your position, nor what the engine displays above you.

In paid advertising. The denominator is the number of impressions the algorithm decided to serve, on an audience you defined. You control the targeting, so you partly control the denominator.

In email. The denominator is either the number of messages delivered or the number opened, depending on the definition used, which produces two very different numbers for the same campaign.

Why that forbids the comparison. A 2% click-through rate in search, in advertising and in email describes three unrelated phenomena. Pages that line all three up in one benchmark table are comparing incomparable quantities, and that is nearly every page on the subject.

The practical consequence you will actually meet. Someone compares an ad campaign’s click-through rate to an email benchmark, concludes they are underperforming, and changes what was working.

Click-through rate applied to three distinct contextsDiagram showing that the click-through rate formula, clicks divided by impressions multiplied by one hundred, does not change from one context to another while what it measures changes entirely. In organic search the denominator is the number of times the page appeared on a results page, and the advertiser controls neither the queries that trigger it, nor the position, nor what the engine displays above it. In paid advertising the denominator is the number of impressions the algorithm decided to serve on an audience the advertiser defined, so the advertiser partly controls that denominator through targeting. In email the denominator is either the number of messages delivered or the number opened depending on the definition used, which produces two very different numbers for the same campaign. It follows that a click-through rate of two percent in search, in advertising and in email describes three unrelated phenomena, so pages lining these three universes up in a single benchmark table are comparing incomparable quantities. The most frequent practical consequence is that a manager comparing the click-through rate of an advertising campaign to an email benchmark wrongly concludes that performance is poor and changes what was working.One formula, three denominatorsclicks ÷ impressions × 100Identical everywhere. It is the denominator that changes nature.Organic searchAppearances on aresults page.You control nothing.Paid advertisingImpressions served on anaudience you defined.You control the targeting.EmailMessages delivered, oropened, by definition.Two possible numbers.Lining the three up in a benchmark table compares incomparable things.The case you will actually meet: someone compares an ad campaign’s CTR to an email benchmark,concludes they are underperforming, and changes what was working.
A 2% click-through rate in search, in advertising and in email describes three unrelated phenomena. Almost every page on the subject lines them up anyway. Source : MASTRATOS (2026)

Three ways to raise it without improving anything

This is the heart of the problem, and no page on the subject states it.

Narrow the targeting. In paid advertising, restricting the audience to people who already know you makes click-through rate jump. You are no longer reaching anyone new, your volume collapses, and your dashboard shows your best number of the year.

Prune the list. In email, removing inactive subscribers mechanically raises the ratio. The number of clicks is unchanged, the denominator drops. Nobody read more.

Promise more than you deliver. A headline that overstates gets the click and loses the visitor. In organic search it moves the problem to the return-to-results rate, which you do not measure.

What makes these three insidious. None is cheating. Restricting an audience, cleaning a list and working on a headline are three good practices in other contexts. They simply raise this ratio for reasons unrelated to performance.

The question that exposes all three. Did the absolute number of clicks go up? If the rate rises while clicks fall, you have not improved anything. You have reduced the denominator.

Three methods that raise click-through rate without real improvementDiagram setting out three methods that raise a click-through rate without any performance having improved, which is the central problem posed by this metric. The first is narrowing the targeting in paid advertising, where restricting the audience to people who already know the company makes the click-through rate jump while volume collapses and nobody new is reached. The second is pruning the email list, where removing inactive subscribers mechanically raises the ratio because the number of clicks stays identical while the denominator falls, and nobody has read more. The third is promising more than the page delivers, where an overstated headline earns the click and loses the visitor, moving the problem in organic search toward the return-to-results rate, which is not measured. These three cases are insidious because none of them is cheating, restricting an audience, cleaning a list and working on a headline being three good practices in other contexts. The question that exposes all three is whether the absolute number of clicks went up, because a rate that rises while clicks fall signals a reduced denominator rather than an improvement.Rising without progressingNarrow the targetingThe rate jumpsVolume collapsesYou reach nobody new, and you post your best number of the year.Prune the listThe rate risesClicks do not moveSame numerator, smaller denominator. Nobody read more.Promise more than you deliverThe rate risesThey leave at onceThe problem moves to a metric you do not measure.The question that exposes them: did the absolute number of clicks go up?None of the three is cheating. That is what makes them hard to spot in a report.
The question that exposes all three: did the absolute number of clicks go up? If the rate rises while clicks fall, you reduced the denominator. Source : MASTRATOS (2026)

In email, one of the two denominators has stopped existing

This is the clearest case of a destroyed denominator, and many companies still manage on it.

The two competing definitions. A send’s click-through rate is calculated either on messages delivered or on messages opened. The second, usually called click-to-open rate, produces a much higher number, and it is the one vendors like to present.

What happened to the second denominator. Apple describes Mail Privacy Protection as hiding your IP address from senders and “privately download remote content in the background when a message is received (instead of when you view it).” The open is therefore recorded whether or not the recipient read anything.

The arithmetic consequence. The number of opens no longer measures opens. It mixes real reads and automatic fetches, in a proportion you do not know and which depends on how much of your list uses Apple Mail.

What it does to the ratio. The denominator inflates, click-to-open collapses, and nothing changed in your recipients’ behaviour. Conversely, a company whose list shifts toward other mail clients will see that ratio recover without having improved anything.

The rule that follows. Click-to-open is no longer comparable over time, nor between two lists. Click-through on delivered messages stays sound: its denominator is a count of successful sends, which nothing inflates.

One benchmark, read with its date. Mailchimp publishes 2.62% click rate across all industries and 2.78% for business and finance, explicitly measured on successfully delivered emails rather than on opens, and only for campaigns sent to at least 1,000 subscribers. The page states the data was last updated in December 2023, and Mailchimp itself warns that open rates are affected by Apple’s privacy changes.

The reflex to have in a meeting. Always ask which denominator produced the number on screen. A vendor who cannot answer does not know what they are showing you.

From what point does a gap mean something?

This is the question the market never asks, and it decides everything else.

The principle. A click-through rate is a proportion measured on a sample. Like any proportion it fluctuates from one measurement to the next with no cause whatsoever. To claim a gap is real, you need enough observations.

The order of magnitude, and it surprises people. Detecting a move from 3% to 4%, already a large gap in relative terms, requires roughly 5,300 impressions per variant at 95% confidence and 80% power. That is over 10,000 impressions in total to compare two ads. The formula is the standard two-proportion sample size calculation, published with a worked example by Select Statistical Services.

What that means for a small company. A modest-budget campaign often takes several weeks to reach that volume, and a page receiving a few hundred impressions a month in search results will never reach it over an interpretable period.

The direct consequence. Most of the click-through rate variation discussed in monthly review meetings is not variation. It is fluctuation. The decisions taken on it are arbitrary, and they sometimes stop the better version.

The rule that follows, and it is simple. Never compare two click-through rates without looking at the volumes underneath them. A three-point gap on two hundred impressions says nothing. The same gap on twenty thousand says something.

Impression volume required to interpret a click-through rate gapChart indicating the impression volume required for a click-through rate gap to be interpretable. A click-through rate being a proportion measured on a sample, it fluctuates from one measurement to the next without any real cause, so claiming that a gap is real presupposes enough observations. Detecting a move from three percent to four percent, a considerable gap in relative terms since it represents a third of an increase, requires roughly five thousand three hundred impressions per variant at ninety-five percent confidence and eighty percent statistical power, which amounts to more than ten thousand impressions in total in order to compare two advertisements against each other. A modest-budget campaign often takes several weeks to reach that volume, and a page receiving a few hundred impressions a month in search results will never reach it over an interpretable period. It follows that most click-through rate variations discussed in monthly review meetings are not variations but fluctuations, the decisions taken on that basis being arbitrary and sometimes leading to stopping the better version. The practical rule is never to compare two click-through rates without looking at the volumes underneath them, since a three-point gap on two hundred impressions says nothing whereas the same gap on twenty thousand impressions says something.When does a gap mean something?To detect a move from 3% to 4%, at 95% confidence and 80% power.200 impressionsSays nothing. Pure fluctuation.1,000 impressionsStill not interpretable.5,300 impressions per variantThe threshold.20,000 impressionsComfortable.Comparing two ads takes more than 10,000 impressions in total.A page at a few hundred impressions a month will never get there over a readable period.
Order of magnitude for detecting a move from 3% to 4% at 95% confidence and 80% power. Most monthly variations never reach it. Source : MASTRATOS from Select Statistical Services (2026)

A high click-through rate that earns nothing

This is the most common case in B2B advertising, and it is almost always misdiagnosed.

The symptom. The ad beats the benchmark, traffic arrives, and nothing happens afterwards: no enquiry, no form, no call.

The usual reading, and why it is wrong. People blame the landing page. Sometimes that is fair. More often the problem is upstream: the message made the wrong people click, and the page can do nothing about it.

What produces this case. A promise of free access, a headline asking a question everyone answers yes to, a visual that catches the eye without saying what is on offer. All three raise clicks and select badly.

How to tell the two causes apart. Look at bounce and time on page. If visitors leave within seconds, they were not the right people: that is the message. If they stay, read, and do not convert: that is the page.

The consequence for management. Click-through rate must never be read alone. It is read next to what happens after the click, or it points you at the wrong correction.

In organic search, the average hides what it aggregates

This is an aggregation trap, and it regularly produces inverted conclusions.

What Search Console shows you. An average click-through rate across all your queries for the period. It is weighted by impressions, and it mixes queries that have nothing to do with each other.

Why the average moves when nothing does. A very broad query on which you rank poorly brings many impressions and almost no clicks. If you lose ground on that query, your impressions fall, your average click-through rate rises, and your traffic does not move. The dashboard shows an improvement where only a disappearance occurred.

The reverse move is just as misleading. An article starting to rank on broader queries brings in many poorly qualified impressions. The site’s average click-through rate falls, and that is good news.

What makes the trap hard to see. The two figures that explain it, average position and impression count, sit right next to it. But they are averages themselves, and a stable average position can conceal a collapse on one query offset by a gain on another.

The reading that corrects it. Never look at click-through rate at site level. Go down to the query, or to the group of queries describing one intent. That is the only level at which the number means anything.

The link with the rest of this page. It is the same mechanism again: the ratio moves because the population composing it changed, not because behaviour changed. In advertising you cause that change through your targeting. In organic search the engine imposes it on you.

What AI summaries changed, and why the studies disagree

The measurements exist, they are recent, and they do not agree. The disagreement is methodological, not noise, and understanding it is more useful than picking the largest number.

What real browsing shows. The Pew Research Center tracked 900 US adults across 68,879 unique Google searches in March 2025, of which 12,593 produced an AI summary. A traditional result was clicked on 8% of visits where a summary appeared, against 15% where none did. Only 1% of visits produced a click on a link inside the summary itself, and 26% of pages carrying a summary ended the browsing session, against 16% without.

What ranking data shows, and how fast it moved. Ahrefs compared 300,000 keywords, half triggering an AI Overview and half an informational control, on desktop Search Console data. In April 2025 it reported 34.5% lower click-through rate in position one. In February 2026, the same design reported 58%.

What a same-keyword comparison shows, and it points the other way. Semrush compared the same keywords before and after they began showing AI Overviews, across more than 200,000 keywords, and found zero-click rates slightly decreasing, from 38.1% to 36.2%.

Why they disagree. Pew measures what humans did, at visit level, on a panel. Ahrefs measures the click-through rate of ranking pages, on desktop only, through aggregated Search Console. Semrush holds the keyword constant and looks before and after. These are three different questions, and pages with AI summaries skew informational to begin with, which biases any comparison that does not control for it.

What to take from the spread. The direction is consistent across the panel-based and ranking-based work: informational queries lose clicks. The magnitude is not established, and anyone quoting a single figure without its method is selling you a number rather than a finding.

What it means for your own dashboard. If your organic click-through rate fell without your positions moving, the most likely explanation is not your title. It is that the engine now answers above your link. And comparing a 2026 click-through rate to a 2024 one on the same page no longer compares the same thing: an impression today carries less attention than it did two years ago.

Published measurements of the effect of AI summaries on organic click-through rateTable comparing the published measurements of the effect of artificial-intelligence-generated summaries on organic click-through rate, and setting out why they disagree with one another. The Pew Research Center tracked the real browsing of nine hundred United States adults across sixty-eight thousand eight hundred and seventy-nine unique Google searches during March two thousand and twenty-five, of which twelve thousand five hundred and ninety-three produced an artificial-intelligence summary, and found that a traditional search result was clicked on eight percent of visits where a summary appeared against fifteen percent of visits where none did. Ahrefs compared three hundred thousand keywords on desktop Search Console data and reported a thirty-four and a half percent lower click-through rate in position one in April two thousand and twenty-five, then fifty-eight percent using the same design in February two thousand and twenty-six. Semrush compared the same keywords before and after they began showing summaries, across more than two hundred thousand keywords, and found zero-click rates slightly decreasing from thirty-eight point one percent to thirty-six point two percent. The measurements disagree because they answer different questions: Pew measures what humans actually did at visit level on a panel, Ahrefs measures the click-through rate of ranking pages on desktop only through aggregated Search Console data, and Semrush holds the keyword constant and looks before and after. Pages carrying summaries also skew informational to begin with, which biases any comparison that does not control for it. The direction is consistent, informational queries lose clicks, but the magnitude is not established.Four measurements, three questionsSOURCEFINDINGWHAT IT ACTUALLY MEASURESPew, 20258% vs 15% of visitsReal browsing, 900 US adults,68,879 searches, March 2025.Ahrefs, Apr 2025-34.5% CTR, position 1300,000 keywords, desktop only,aggregated Search Console.Ahrefs, Feb 2026-58% CTRSame design, two-year window,baseline already post-rollout.SemrushZero-click 38.1% to 36.2%Same keywords before and after,200,000+ keywords.The direction is consistent. The magnitude is not established.Anyone quoting one figure without its method is selling a number, not a finding.
Three different questions, three different methods. The direction is consistent, the magnitude is not established. Source : MASTRATOS from Pew, Ahrefs, Semrush (2026)

The position benchmarks are older than they look

They are everywhere, they are quoted as current reference points, and almost none of them is.

The 28.5% figure. It comes from a Sistrix analysis published in July 2020, on mobile data only. It still circulates as a 2026 benchmark.

What that same article actually argues, and nobody quotes. Sistrix’s own point is that a single average is no longer valid. On its data, position-one click-through rate ranges from 13.7% when a Shopping block sits above to 46.9% when sitelinks are present, with 18.8% under ads and 34.2% on a purely organic page. The headline number is the one part of that article worth the least.

The 27.6% figure. It comes from Backlinko, drawn from 1.3 million pages and 12.2 million queries via Semrush. The page carries a last-updated stamp but no disclosed data collection date and no date range, and it does not say whether the data is mobile or desktop. A last-updated stamp is not a collection date.

Why this matters more now than it used to. The results page changed shape in between: generated summaries, question blocks, rich snippets, carousels. A position one in 2026 does not occupy the screen a position one occupied in 2020, which is precisely Sistrix’s argument.

The advertising benchmark, and the one that does declare itself. WordStream puts search click-through rate at 6.64% across industries, on a stated sample of 13,474 US search campaigns running April 2025 to March 2026, with at least 52 campaigns per subcategory. It also states plainly that its “averages” are technically median figures.

A detail worth knowing before you cite it. The same numbers are republished on a more accessible page without that methodology block. Cite the version that declares its sample, not the one that dropped it.

What makes all these figures weakly useful anyway. None of them knows your targeting, your definition of an impression, or your sector as you understand it. A benchmark describes the dispersion of a market, not your situation.

In advertising, CTR is no longer what the algorithm pursues

This is the point the pages on this subject have not updated, and it changes how to read the number.

What changed. The dominant bidding strategies optimise conversions or their value, not clicks. You are asking the platform to find buyers, not clickers.

What that does to click-through rate. It becomes an intermediate variable the model passes through. The algorithm can deliberately serve your ads to profiles that click less and buy more, lowering your click-through rate while improving your result.

The consequence in a review meeting. A click-through rate falling while cost per enquiry also falls is not a problem to correct. It is the expected behaviour of the system. Correcting it means fighting the optimisation you asked for.

The exception that remains. On campaigns set to bid on clicks, with no conversion objective, click-through rate stays a direct indicator: it is what you are buying. The distinction is in the setting, not in the platform.

The reading rule. Always ask which bidding strategy is running before commenting on a click-through rate. Without that, the number carries no interpretable meaning.

Real scope of the most quoted click-through rate benchmarksTable exposing the real scope of the three click-through rate benchmarks most widely repeated as current reference points. The figure of twenty-eight and a half percent for position one in organic search comes from a Sistrix analysis published in July two thousand and twenty on mobile data only, and that same article argues that a single average is no longer valid since position-one click-through rate ranges on its own data from thirteen point seven percent when a Shopping block sits above the result to forty-six point nine percent when sitelinks are present, with eighteen point eight percent under advertisements and thirty-four point two percent on a purely organic page. The figure of twenty-seven point six percent for position one comes from Backlinko, drawn from one point three million pages and twelve point two million queries through Semrush, but the page carries no disclosed data collection date and no date range and does not state whether the data is mobile or desktop, a last-updated stamp not being a collection date. The advertising figure of six point six four percent for search click-through rate across industries comes from WordStream on a stated sample of thirteen thousand four hundred and seventy-four United States search campaigns running from April two thousand and twenty-five to March two thousand and twenty-six, with at least fifty-two campaigns per subcategory, and that source states plainly that its averages are technically median figures. The same numbers are republished on a more accessible page without that methodology block, so the version that declares its sample should be cited rather than the one that dropped it.What the quoted averages actually measureFIGUREWHAT IT DESCRIBESREAL SCOPE28.5%Position 1, organicSistrix, July 2020, mobile onlySame article: the range is 13.7% to 46.9% by layout27.6%Position 1, organicBacklinko, 1.3M pages via SemrushNo collection date, no range, device not stated6.64%Paid search, all industriesWordStream, 13,474 US campaignsApr 2025 to Mar 2026, and its “averages” are mediansThe only benchmark that concerns you is your own history.Same channel, same audience, same season. It is the only one that controls the variables that matter.
Three reference points repeated as current. Only one declares its sample and period, and it says its averages are medians. Source : MASTRATOS from Sistrix, Backlinko, WordStream (2026)

What to watch instead, and what CTR is still good for

It is not to be thrown away. It is to be put back in its place, which is that of a diagnostic indicator rather than a result.

What it is genuinely good for. Spotting a break. An ad whose click-through rate collapses within days signals creative wear or an audience change. It is an anomaly detector, and that is useful.

What it is not good for. Deciding a budget, arbitrating between channels, or judging a vendor. All three require knowing what happens after the click.

The pair to watch. Click-through rate and cost per qualified enquiry, read together. When both rise, you are attracting the wrong audience. When the first falls and the second falls too, the algorithm is doing its job.

The log to keep. Record every change of targeting, creative and bidding strategy, with the date. Without that log, no click-through rate variation is attributable to a cause, and you will comment on noise for months.

Where to go next

Your organic click-through rate fell without your positions moving. Look at what is displayed above your link before you rewrite your title.

Your creative has been running a long time. Creative fatigue metrics.

You want to know which metric to run on. ROAS vs MER vs CAC vs LTV.

Your counters disagree with each other. Why GA4 and Meta conversions don’t match.

You are structuring a search account. Google Ads account structure for B2B.

Your landing page is where the clicks die. What makes a B2B landing page convert.

In short

  • The formula is identical everywhere, the denominator is not. Search, advertising and email do not measure the same thing under the same name.
  • Three methods raise it without improving anything: narrow the targeting, prune the list, promise more than you deliver. None of them is cheating.
  • The question that exposes them: did the absolute number of clicks go up?
  • Roughly 5,300 impressions per variant are needed to detect a move from 3% to 4%. Most of the variation discussed monthly is noise.
  • The AI Overview studies disagree for methodological reasons. Pew found 8% of visits clicking against 15%; Ahrefs reported 34.5% then 58% lower CTR; Semrush found zero-click slightly falling. The direction holds, the magnitude does not.
  • The position benchmarks are 2020 mobile-only data, or carry no collection date at all, on a results page that no longer exists.
  • In email, one denominator has stopped existing. Apple downloads remote content on receipt rather than on view, so click-to-open is no longer comparable.
  • The algorithm no longer optimises the click. A falling CTR alongside a falling cost per enquiry is the system working.

A click-through rate is never judged alone. Book a diagnostic, or see how we approach B2B paid acquisition.