Search engine market share, measured and dated

Share of what, and measured by whom

Search engine market share describes how search activity is distributed between engines. The phrase hides a question that decides the answer: share of what? Share of queries, share of users, share of search advertising revenue and share of outbound referral traffic are four different quantities, and they do not agree with one another.

The figures with the strongest provenance are not the ones most often quoted. They come from competition regulators and from a federal court, bodies that can compel a company to produce its own internal query logs and then publish findings resting on them. Those are set out first on this page. The figure almost everyone quotes comes second, because of what it is. StatCounter Global Stats measures referral traffic — visits arriving at a website from a search engine's results page — by counting page views on sites carrying its tracking code and attributing each visit to whichever engine referred it. That is not a census of searches performed. It is the best free public proxy available rather than ground truth, and Comscore, Similarweb and Sistrix publish rival estimates built on different mechanisms that produce different numbers.

Every figure below carries the period it was measured and the source that measured it. Market-share pages go stale faster than anything else written about search, and an undated percentage is worse than no percentage — it gets quoted for years afterwards as though it were current.

What a court and two regulators found

The most authoritative search shares in existence were produced by litigation and by regulatory investigation, because in both settings the numbers were compelled rather than sampled.

The United States court findings, 2020 data

The memorandum opinion of the US District Court for the District of Columbia in United States of America v. Google LLC, No. 1:20-cv-03010-APM, issued in 2024, records that by 2020 Google handled nearly 90 percent of all US general search queries and 94.9 percent of queries on mobile devices. The court's table of the remainder reads Bing 5.5 percent, Yahoo 2.2 percent, DuckDuckGo 2.1 percent and all other rivals together 0.9 percent, so every competitor combined accounted for less than 11 percent. These are findings of fact drawn from Comscore data and from Google's own internal figures rather than from a vendor panel, which is why DuckDuckGo's 2.1 percent is more citable than any estimate of the same quantity. Two limits belong with the numbers: the shares describe 2020, and the judgment is on appeal.

The United Kingdom regulatory findings, 2024 and 2025 data

The UK Competition and Markets Authority, in its 2025 final decision on the strategic market status investigation into Google's general search services, found that Google Search accounts for more than 90 percent of all general search queries in the United Kingdom. The same decision records that more than 200,000 UK firms collectively spent over £10 billion on Google search advertising in 2024. The CMA's query-share measure is defined to include search-grounded AI queries, so it counts activity that a referral-based tracker cannot see at all. This is a legally binding determination backed by data the regulator compelled from Google.

Ofcom measures a different quantity again. Its Online Nation 2025 report puts Google Search's reach at 82 percent of UK online adults in May 2025, down from 83 percent in 2024 and 86 percent in 2023, and estimates approximately 3 billion UK web searches per month. Ofcom uses Ipsos iris passive device-level measurement, so this is observed behavior rather than survey recall. Reach counts users, not queries, so the 82 percent figure is not comparable with the CMA's 90 percent and the two should never be placed in the same series. The three-year decline is nonetheless the kind of movement a headline share table conceals entirely.

Worldwide search engine market share, August 2026

StatCounter Global Stats, all platforms (desktop, mobile, tablet and console combined), August 2026. StatCounter describes the sample as based on over 3 billion monthly page views. Retrieved 17 September 2026.

• Google — 91.1%
• Bing — 4.5%
• Yahoo! — 1.23%
• Yandex — 0.99%
• DuckDuckGo — 0.7%
• Baidu — 0.62%

Those six account for roughly 99.1% of the measured total; the residue is spread across engines too small to be broken out. Two things in that list are worth pausing on. Baidu at 0.62% worldwide is a Chinese-market engine appearing through a panel that barely reaches China, and Yandex at 0.99% is a Russian-market engine whose worldwide figure is dominated by one country. Neither number describes a global consumer proposition; both describe a national one seen from outside.

The other thing worth noticing is what is absent. There is no row for Brave Search, Ecosia, Startpage, Mojeek, Kagi or Perplexity in the worldwide top six. Every independent alternative engine put together is a rounding error at global scale, and any page that implies otherwise is selling something.

Adjudicated against panel-estimated: the same engines, different numbers

Set the court's findings beside StatCounter's United States table for August 2026 — Google 86.1%, Bing 8.92%, Yahoo! 2.74%, DuckDuckGo 1.7%, Yandex 0.34%, Ecosia 0.12% — and the discrepancy is plain. The court found Google at nearly 90 percent of US general search queries; StatCounter reads 86.1% of US search referrals. The court found DuckDuckGo at 2.1 percent; StatCounter reads 1.7%. Bing runs the other way, 5.5 percent in the court record against 8.92% in the panel.

Three things separate the two columns, and all three matter. They measure different quantities, queries against referrals. They describe different years, 2020 against August 2026. And they rest on different evidence, compelled internal data tested at trial against a sample of the web's outbound referral flow. The gap is not a sign that one source is dishonest. It is the normal distance between an adjudicated figure and an estimated one, and a page that quotes only the estimate without showing that distance is understating how uncertain the estimate is.

Query diversity is not query share

Share counts how many searches an engine handles. Query diversity — how much of the space of distinct questions an engine has ever seen — is a different quantity, and it is the one that compounds.

The trial record in United States v. Google LLC contains a measurement of it. Across a seven-day sample of 3.7 million unique search phrases, 93 percent were seen only by Google, while just 4.8 percent were seen only by Bing. An engine holding a twentieth of the volume does not see a twentieth of the questions. It sees most of the common ones and almost none of the rare ones, and rare queries are precisely where an engine has the least evidence about which result is good. The scale gap therefore does not close by winning users at the margin, because the marginal user brings common queries.

This figure is reproduced in a University of Chicago Law School course supplement, Picker, Antitrust Fall 2025, Supplement 4, which reprints the court's opinion. Where a citation of record is required, cite the primary docket filing rather than the supplement.

The device split changes the answer materially

A single worldwide number conceals the largest structural fact in search: mobile is a near-monopoly and desktop is merely dominated. Same source, same month — StatCounter Global Stats, August 2026, retrieved 17 September 2026.

Worldwide, desktop only, August 2026

• Google — 86.05%
• Bing — 8.84%
• Yahoo! — 2.15%
• Yandex — 1.1%
• DuckDuckGo — 0.7%
• Baidu — 0.34%

Worldwide, mobile only, August 2026

• Google — 95.59%
• Yandex — 0.9%
• Baidu — 0.84%
• DuckDuckGo — 0.69%
• Bing — 0.6%
• Yahoo! — 0.47%

Bing's share falls from 8.84% on desktop to 0.6% on mobile — a factor of roughly fifteen. That is not a statement about how good Bing is. It is a statement about defaults: Bing is the out-of-the-box engine in Microsoft Edge on Windows PCs, and it has no equivalent position on any phone. Google is the default in Chrome, in Safari under a long-standing paid arrangement with Apple, and on Android. Where a default is not set by Microsoft, Bing largely does not appear.

The CMA reached the same structure from the other direction, using Google's own query logs. Annex B to its strategic market status investigation, Market Outcomes, records that mobile devices rose from roughly 50 to 60 percent of Google's UK search queries in 2017 to roughly 70 to 80 percent in 2024, and that Google handled approximately 15 to 20 billion UK queries in December 2024 alone. Those are bands, not point estimates: the CMA publishes them as redacted ranges to protect Google's confidential data, and they should never be quoted as single numbers. Even as bands they are unusual, because a genuine desktop-versus-mobile query split and an absolute monthly query count for one country cannot be derived from referral measurement at all.

This is why the single most common market-share error is quoting the wrong denominator. "Bing has about 4%" and "Bing has about 9%" are both defensible — the first is worldwide all-device, the second is worldwide desktop, and neither is wrong so long as you say which.

How StatCounter measures share, and what it structurally misses

StatCounter's method is simple and its blind spots follow directly from it. A visitor arrives at a site carrying StatCounter's code; the referring URL identifies the engine; the page view is counted. Four consequences fall out of that design.

• It counts referrals, not searches. A search that answers the user's question on the results page itself — a weather box, a currency conversion, a featured snippet, an AI summary — produces no outbound referral and is invisible. As results pages answer more queries directly, the entire measure drifts further from search volume. This is the gap the CMA's query-share definition closes by counting search-grounded AI queries and StatCounter's cannot.
• It systematically undercounts walled-garden engines. An engine that answers queries inside its own properties generates few outbound referrals by design. This is the mechanism behind the Naver discrepancy discussed below, and a milder version of it affects Seznam in Czechia and Yahoo! Japan.
• The panel is uneven between countries. Global coverage is good; coverage of specific national markets varies enormously with how many local sites carry the tracker. Some country tables are robust and some, notably China, are not.
• It cannot see in-app or assistant search at all. Queries typed into an app, a voice assistant or a chat interface leave no referral trail of the kind StatCounter reads.

None of this makes StatCounter useless. It is consistent over time, it is published free at monthly granularity, and its long series are genuinely informative about direction. It is a thermometer, not a scale — good at telling you which way something is moving, poor at telling you exactly what it weighs.

Share of revenue is a different ranking again

Query share and revenue share are not the same distribution, and the revenue gap between Google and its nearest competitor is far wider than the share gap suggests. Both numbers are audited and company-reported, which makes this one of the few comparisons on the page with no methodology dispute attached.

Alphabet's fourth quarter and fiscal year 2025 results, released in 2026, put Google Search and other revenues at $63,073 million in the fourth quarter of 2025 alone, up 17 percent year over year, within total Alphabet revenues of $402,836 million for the full year. Microsoft's 2025 annual report on Form 10-K reports Search and news advertising revenue of $13,878 million for the whole of fiscal 2025, up from $12,306 million in fiscal 2024 and $12,125 million in fiscal 2023, and growing 20 percent excluding traffic acquisition costs on higher search volume and higher revenue per search. One quarter of Google Search revenue is more than four times Microsoft's entire search and news advertising year. Note that Alphabet breaks Search out on a quarterly basis only in the earnings release; the full-year Search line requires the Form 10-K.

For the size of the market those two are dividing, the Interactive Advertising Bureau and PricewaterhouseCoopers report US search advertising revenue of $114.2 billion in 2025, 38.8 percent of a $294.6 billion total US internet advertising market. PwC compiles the series from audited revenue data submitted directly by the sellers, so it is closer to a census than to a model. The growth rate is the part worth recording: 11.0 percent year over year in 2025, against 15.9 percent in 2024. That deceleration is a measured fact; what caused it is not settled, and AI answer products, advertising market cycles and the size of the base all have claims on it.

A longer independent series exists in federal statistics. The US Census Bureau's Service Annual Survey, Table 1, shows revenue for NAICS 519, Other Information Services — the sector containing web search portals — rising from $104,441 million in 2013 to $347,673 million in 2022, a 3.3-fold increase in nine years. NAICS 519 is a parent category and is not search engines alone, and the 2022 NAICS revision reorganized these codes, so the series is not perfectly comparable across its full length. It is still the only decade-long revenue measure of the industry that comes from a statistical agency rather than a vendor.

Where the sources disagree — printed, not resolved

Two conflicts in the current data are large enough that publishing a single number would be a distortion. Both are recorded here as conflicts.

Conflict 1: Naver's share of South Korean search

StatCounter puts Naver at 43.71% of South Korean search for August 2026, behind Google at 46.6%. InternetTrend, the Korean domestic measurement standard, puts Naver at 63.8% against Google's 28.7% for a March 2026 monthly average, as reported by The Korea Times on 16 April 2026. The gap is roughly 20 percentage points and the two sources place different engines first.

The reason is methodological, not arithmetic. Naver is a portal: an enormous share of Naver searches are answered by Naver's own Blog, Cafe, Knowledge iN and AI Briefing surfaces and never leave naver.com. Those searches generate no outbound referral, so a referral-based measure cannot see them. StatCounter structurally undercounts exactly the behavior that makes Naver distinctive. InternetTrend measures query share from Korean panel data and does capture them. Do not average the two. Note also that the measurement months differ, so they are not strictly comparable even before methodology.

Conflict 2: StatCounter's China panel is not credible

StatCounter's China table for August 2026 reports Baidu 59.26%, Bing 18.92%, Yandex 11.41%, Haosou 6.48%, Sogou 2.25% and Google 1.41%. An 11.41% Yandex share of Chinese consumer search is not a plausible description of how people in China search, and Bing at 18.92% is difficult to reconcile with any other account of that market. The likeliest explanation is panel contamination: StatCounter's Chinese sample is small, weighted toward internationally facing and expatriate-oriented sites, and distorted by VPN traffic.

The correct handling is to state the figures as what StatCounter reports, name StatCounter, and say plainly that the panel is unrepresentative — not to quote the Baidu row as a clean fact while quietly dropping the Yandex row. If the Yandex number is an artifact, the same artifact is depressing Baidu's. Chinese domestic estimates put Baidu materially higher than 59.26%, and all methodologies agree the direction over the last decade is downward.

What the numbers do not yet cover: AI answer engines

There is no equivalent public series for ChatGPT, Perplexity, Claude or Google's AI Mode. StatCounter's categories were designed for engines that send referral traffic, and a conversational assistant that answers in place sends comparatively little. Where AI products do appear in referral logs they show up as small numbers that badly understate how much question-answering they absorb.

This matters for reading the Google row. Google's 91.1% describes its share of the referral traffic that search engines still send. It does not describe whether the total volume of that traffic is growing or shrinking, and the second question is now the more interesting one. A rising share of a shrinking pool and a rising share of a growing pool look identical in a table of percentages. The CMA's decision to define query share so that it includes search-grounded AI queries is a regulator solving exactly this problem, and it is one reason its 90 percent figure and a referral-based figure are not interchangeable.

Anyone who tells you they know AI search's market share to a decimal place is estimating. Treat named figures in that area as vendor claims or modeled estimates unless the methodology is published.

Reading and citing a market-share figure without being wrong

Five rules cover most of the mistakes.

• State the source. "Google has 91%" is unciteable. "Google had 91.1% of worldwide search referrals in August 2026 per StatCounter Global Stats" is.
• State the period. These series move monthly and the number you read in a blog post is usually one to three years old. Court and regulator findings carry a period too, and it is often several years before publication.
• State the geography and the device. Worldwide all-device, worldwide desktop, US all-device and US desktop are four different numbers for the same engine.
• Say whether the figure is adjudicated or estimated. A share found by a court on compelled data and a share sampled from a referral panel are different classes of evidence, and putting them in one sentence without saying so misleads the reader about how firm the number is.
• Do not mix vendors within one comparison. A Comscore figure and a StatCounter figure in the same sentence produce a trend that does not exist.

For a national breakdown, including the markets where Google does not lead, see the country-level figures. For what sits behind the shares — which of these engines actually runs a crawler and which resells another engine's results — the structural comparison is the more useful document, because share and independence are unrelated. Yahoo! has more measured share than Brave Search and no index of its own; Mojeek has almost no share and a complete one.

StatCounter Global Stats publishes the underlying series free, at monthly granularity, with the country and device filters used throughout this page.

When this page was last checked

All StatCounter figures on this page were retrieved directly from gs.statcounter.com on 17 September 2026 and are for the August 2026 reporting month, which was the most recent complete month published at that time. The InternetTrend figure is for March 2026 as reported in April 2026.

The court and regulator figures are not monthly and do not move with the calendar. The US court findings describe 2020 and the judgment is on appeal; the CMA decision and its Annex B were published in 2025 and describe 2024 data; the Ofcom reach figure is for May 2025; the Alphabet, Microsoft, IAB and Census figures carry the fiscal or survey year stated with each.

StatCounter publishes a new month roughly a fortnight into the following month. If you are reading this well after September 2026, the shares above are historical: treat them as a dated snapshot and check the current month at source. That is not a disclaimer added for form — it is the single most important thing to know about any market-share page, including this one.

Frequently asked questions

Sources

• US District Court for the District of Columbia — Memorandum Opinion, United States of America v. Google LLC, No. 1:20-cv-03010-APM (Redacted Public … (2024, PDF)
• UK Competition and Markets Authority — Final Decision: Strategic Market Status Investigation into Google's General Search Services (2025, PDF)
• UK Competition and Markets Authority — Annex B: Market Outcomes, Strategic Market Status Investigation into Google's General Search Services (2025, PDF)
• Ofcom — Online Nation 2025 Report (2025, PDF)
• US District Court for the District of Columbia, via University of Chicago Law School — United States v. Google LLC, Memorandum Opinion, reproduced i… (2025, PDF)
• Interactive Advertising Bureau with PricewaterhouseCoopers — IAB/PwC Internet Advertising Revenue Report: Full Year 2025 Results (2026, PDF)
• Microsoft Corporation — Microsoft 2025 Annual Report (Form 10-K) (2025, DOCX)
• Alphabet Inc. — Alphabet Announces Fourth Quarter and Fiscal Year 2025 Results (2026, PDF)
• US Census Bureau — Service Annual Survey, Table 1: Estimated Revenue for Employer and Nonemployer Firms, 2013 Through 2022 (2024, XLSX)
• gs.statcounter.com/search-engine-market-share
• gs.statcounter.com/search-engine-market-share/desktop/worldwide
• gs.statcounter.com/search-engine-market-share/mobile/worldwide