Click-through rate is the proportion of times an advertisement, search result or link is shown that ends with someone clicking it. Abbreviated to CTR, it is calculated by dividing clicks by impressions and expressing the result as a percentage: five clicks on 100 impressions is a 5% CTR, the worked example Google uses in its own Ads documentation. The metric exists because the click was the first thing the web could count that print and television could not. A magazine reader leaves no trace. A browser that follows a link leaves a server request, and that record turned the click into a unit of pricing, then into an input for ranking advertisements, and eventually into a probability estimated by machine learning systems billions of times a day.

How a click becomes a number

Neither half of the fraction is as simple as it looks. For most of the industry, the numerator is governed by the Click Measurement Guidelines that the Interactive Advertising Bureau (IAB) published with the Media Rating Council (MRC) on May 12, 2009. They describe a four-stage click-referral cycle. An initiated click is the user's action on the ad. A measured click is counted by the publisher or its ad server, typically when it successfully sends an HTTP 302 redirect. A received click is logged when the request reaches the advertiser's server, and a resolved click when the destination page actually loads, usually signalled by an HTTP 200 response. Traffic leaks at every stage through latency, abandoned loads and broken pages, and the guidelines state that pre-filtration measured clicks should always equal or exceed received clicks. That gap is one reason an ad platform and a site analytics tool rarely agree.

Two counting rules are permitted. Under the one-click-per-impression method, a repeat click on the same impression in the same session is ignored. Under the multiple-click method, repeat clicks count once a publisher-defined refractory period has passed, which the document concedes can produce click-through rates above 100%. A separate impression-staleness window stops clicks on long-abandoned ad renders from counting. Both parameters are kept confidential to protect fraud filters.

The denominator shifts by platform. Google Ads divides by impressions. AdSense also reports an ad request CTR, and a four-slot Multiplex grid counted as one request could register several clicks and exceed 100% until Google changed the counting basis on June 23, 2026. In organic search, Search Console's Performance report pairs clicks with impressions that register only when a result appears on a page the user actually views. Meta's Ads Manager separates a CTR (all), which counts any interaction with an ad, from a link click-through rate limited to clicks towards a destination. Two figures labelled CTR can therefore describe different events.

From report column to auction input

CTR matters most where it sets prices. GoTo.com, which introduced pay-per-click search in February 1998, ranked listings by bid alone. Google's AdWords Select, launched in February 2002, introduced cost-per-click pricing, and quality-based ranking followed that weighted bids by historical clickthrough rate. An advertiser bidding $1.00 per click with a 1% CTR is worth $10 per thousand impressions to the search engine; one bidding $0.50 with a 5% CTR is worth $25. Ranking by bid multiplied by CTR maximises revenue per results page and gives a relevant low bidder a route above a richer one.

Google folded that weighting into Quality Score in 2005. As of September 2026, its help centre lists three components: expected clickthrough rate, ad relevance and landing page experience, each rated above average, average or below average against other advertisers whose ads showed for the exact same search over the previous 90 days. Expected CTR is a prediction rather than the CTR column. According to Google, it is adjusted to remove the influence of ad position and assets, then recalculated at auction time using the search terms, device and other signals. A small US experiment that began on June 24, 2026 draws on the same components to label some ads "Strongest match".

Programmatic buying uses the same probability differently. Exchanges sell impressions priced per thousand (CPM), so a demand-side platform (DSP) pursuing a cost-per-click (CPC) goal has to translate. Google's Display & Video 360 API lists CPC among the performance goals its automated bidding can target. With a $2.00 click target and a predicted CTR of 0.1%, an impression is worth $2.00 CPM (2.00 x 0.001 x 1,000); if the model predicts 0.2%, the bid doubles. In a paper presented at the KDD data mining conference on August 11, 2013, Google engineers led by H. Brendan McMahan described CTR prediction as a massive-scale learning problem central to online advertising and set out the online learning methods behind a deployed system.

Origins and decline

The founding figure is anecdotal. On October 27, 1994, HotWired ran an AT&T banner asking readers whether they had ever clicked their mouse there. Joe McCambley, who worked on it at Modem Media, later wrote: "Of those who saw the ad, 44% clicked." Digiday's oral history notes that HotWired launched ads from 12 brands simultaneously, so even the claim to be first is contested, and no audit of the 44% exists.

Rates fell as novelty wore off. In October 2009, comScore and Starcom USA reported that the share of US internet users clicking a display ad in a month had halved, from 32% in July 2007 to 16% in March 2009, and that 8% of users produced 85% of clicks. "Marketers who attempt to optimize their advertising campaigns solely around the click are assigning no value to the 84 percent of Internet users who don't click on an ad," said Linda Anderson, comScore's vice president of marketing solutions. The IAB guidelines arrived the same year, first released for public comment on February 26, 2009, in response to growing cost-per-click spending and to click fraud.

Why it matters

In search, CTR has become the clearest measure of what generative answers take from websites. SISTRIX found that the CTR at organic position one in Germany falls from 27% to 11% when an AI Overview appears, across more than 100 million keywords. Ahrefs measured a 58% reduction for top-ranking pages in December 2025 data. Seer Interactive recorded organic CTR down 61% and paid CTR down 68% on informational queries carrying the feature, from 1.76% to 0.61% and from 19.7% to 6.34% across 3,119 queries at 42 organisations. The strongest evidence is experimental: a randomised study of 1,065 Chrome users put the outbound organic click loss at 39.8%, while clicks on sponsored results stayed flat.

The ratio also moves with audience breadth, not just quality. YouTube staff described thumbnail CTR as a figure that falls predictably as impressions widen beyond core fans.

Limitations and disputes

The central criticism is that a click is not an outcome. comScore's 2009 work argued that most display value accrues to people who never click, a position now embedded in view-through measurement. A high CTR can equally reflect curiosity, misleading creative or a placement built to be hit by accident.

Invalid activity inflates the numerator. The MRC's definition of invalid traffic is wide enough to cover the second half of an accidental double-click as well as bots and competitor clicking. Lunio measured invalid rates across retail search accounts rising from 4.18% in the fourth quarter of 2025 to 5.54% in the second quarter of 2026. The 2009 guidelines themselves concede that not all fraudulent clicks are knowable or discoverable.

The denominator can be reshaped too, and not only by advertisers. Mike Ryan of Smarter Ecommerce, analysing 175 billion impressions from mid-2025 to mid-2026, found median Google Shopping impressions falling from about 1.85 million to 1.4 million while median CTR rose from 1.20% to nearly 1.55%. He theorised that Google serves AI Overviews on lower-probability queries, removing the impressions least likely to be clicked. If correct, the improvement reflects a platform decision rather than anything an advertiser changed. Google has not confirmed the mechanism.

Causation remains the weakest link. Seer's authors cautioned that brands cited in AI Overviews may post higher CTRs because they are already strong, not because of the citation. Panel measurement has its own gaps: BIScience's CTR benchmarking, launched on December 5, 2025, has not explained how it separates deliberate clicks from accidental taps.

Not the same as

Conversion rate divides conversions by clicks and measures what happens after the landing page loads; CTR stops at the click. Expected CTR is Google's position-neutral keyword prediction, and predicted CTR in a DSP is a per-impression probability: neither is an observed rate. View-through rate has no settled meaning, used for video completion by some platforms and post-impression conversions by others. Viewability records whether an ad met a visibility standard; CTR describes what a user did after the ad appeared.

Recent developments

Conversational advertising has made CTR a pricing question again. OpenAI added cost-per-click bidding to ChatGPT's Ads Manager on May 5, 2026. Similarweb data put ChatGPT ad CTR at 0.68% overall in May and 0.50% in a dataset published on August 17, 2026. Criteo, a commercial partner, reported rates roughly three times those of comparable formats. The methods differ and the Criteo figure is self-reported.

Automated traffic is the other pressure. DoubleVerify reported on July 29, 2026 that AI bots generated up to ten times more clicks than humans in some unprotected campaigns. Where that happens, a rising CTR signals contamination rather than interest.

Timeline

  • October 27, 1994: HotWired runs an AT&T banner later credited with a 44% click-through rate
  • February 1998: GoTo.com introduces pay-per-click search listings ranked by bid alone
  • February 2002: Google launches AdWords Select with cost-per-click pricing; bids are then weighted by historical CTR
  • 2005: Google introduces Quality Score, with expected CTR as a component
  • July 2007: comScore data shows 32% of US internet users clicking a display ad in a month
  • February 26, 2009: IAB releases Click Measurement Guidelines for public comment
  • May 12, 2009: IAB and MRC publish Click Measurement Guidelines version 1.0
  • October 1, 2009: comScore and Starcom report clickers down to 16% of users, with 8% producing 85% of clicks
  • August 11, 2013: Google engineers present "Ad Click Prediction: a View from the Trenches" at KDD
  • November 2025: Seer Interactive reports organic CTR down 61% and paid CTR down 68% on AI Overview queries
  • December 5, 2025: BIScience launches panel-based CTR benchmarking in AdClarity
  • February 4, 2026: Ahrefs reports a 58% CTR reduction for top-ranking pages with AI Overviews
  • March 1, 2026: SISTRIX reports position-one CTR in Germany falling from 27% to 11% with AI Overviews
  • April 3, 2026: Randomised study measuring a 39.8% organic click loss posted to SSRN
  • May 5, 2026: OpenAI adds cost-per-click bidding to ChatGPT Ads Manager
  • June 23, 2026: AdSense changes Multiplex request counting, capping ad request CTR at 100% per slot
  • June 24, 2026: Google begins a US test of "Strongest match" labels built on Quality Score components
  • July 29, 2026: DoubleVerify reports AI bots generating up to ten times more clicks than humans in some campaigns
  • August 17, 2026: Similarweb data puts ChatGPT ad CTR at 0.50%
  • September 7, 2026: Analysis of 175 billion impressions shows Shopping CTR rising as impressions fall

Summary

Who. Publishers and ad servers count clicks and impressions; platforms including Google, Meta, Microsoft and OpenAI report CTR and use predicted CTR to rank and price ads; DSPs estimate it per impression. The IAB and the MRC define what a valid click is.

What. Click-through rate is clicks divided by impressions, expressed as a percentage. It exists as an observed reporting metric and as a predicted probability that feeds auctions, and its value depends on which click stage and which denominator a platform counts.

When. The metric dates to the first commercial banners in October 1994, became an auction input with AdWords Select in February 2002, was standardised by the IAB and MRC on May 12, 2009, and has been repriced in 2026 by AI search features and click-based bidding in ChatGPT.

Where. It is reported across search, display, social, video, email and conversational advertising, and in organic search through tools such as Search Console, with definitions varying by platform.

Why. CTR is cheap to measure and fast to move, which made it the default proxy for relevance and a pricing lever. The same properties make it vulnerable to accidental clicks, bots and platform-side changes to the denominator, so it describes attention to an ad rather than its commercial effect.