Google Search

By Bill Hartzer · Updated 19 August 2026

What Google is, and where its results actually come from

Google Search is the world's dominant general-purpose web search engine, and it is one of a genuinely small number of engines that builds its own index — the engine's own stored copy of the web, assembled by its own crawler. When a query returns ten links, those links come from Google's index, ranked by Google's own systems. No other general search engine supplies them.

That matters more than it sounds, because most of the search market runs on borrowed results. Startpage resells Google's results as its product. Yahoo has resold results for its entire modern life — from Google between 2015 and 2018, and from Microsoft before and after. Google is on the other side of that trade: it does not take web links from anyone. It licenses and ingests structured data for specific answer boxes — flights, sports scores, financial and other licensed feeds — but the ranked organic results are its own crawl, its own index, its own ranking.

The product a reader meets in 2026 is no longer a list of ten blue links. It is an answer surface: an AI summary at the top of many queries, a knowledge panel, maps, shopping units, video, news, images and ads, with the organic links pushed down the page. The index underneath is unchanged in kind; the presentation on top of it has changed enormously.

How it gets its results: Googlebot, the index, and the ranking layer

Google documents its crawlers publicly, which means the claim that it crawls for itself can be checked rather than taken on trust. The main crawler is Googlebot, in two variants:

Googlebot Smartphone, which identifies itself as a mobile Android browser with compatible; Googlebot/2.1 appended. Google indexes mobile-first, so this variant does most of the crawling.
Googlebot Desktop, the desktop equivalent.

Both obey the Googlebot token in a site's robots.txt file — the plain-text file at the root of a site that tells crawlers which paths they may fetch. Google also runs GoogleOther, a generic fetcher used by various internal product teams, and publishes Google-Extended, which is a robots.txt token with no user-agent of its own. Google-Extended controls whether a site's content may be used to train and ground Gemini models; Google states it does not affect whether a page appears in Search. A site can therefore decline AI training while remaining indexed.

Ranking runs on hundreds of signals. Google acknowledges broad categories — link-based authority, on-page relevance, freshness, page experience, site-level quality — without disclosing weights. Core algorithm updates are announced on its public Search Status Dashboard.

The AI layer sits on top of the same index rather than beside it. AI Overviews and AI Mode are generated by Gemini models grounded on Google's index using what Google calls a “query fan-out” technique, in which one question is decomposed into several underlying searches. There is no second, separate AI crawl of the web feeding it.

Who runs it and how it makes money

Google LLC has been a subsidiary of Alphabet Inc. since the 2015 reorganisation. The company was incorporated on 4 September 1998 by Larry Page and Sergey Brin; the google.com domain was registered on 15 September 1997.

The business model is advertising, overwhelmingly. Search ads are sold through Google Ads on an auction basis, priced per click. Alphabet reports “Google Search & other” as its largest single revenue line; cloud, YouTube advertising, the Play Store, subscription tiers and hardware make up the remainder. Web search itself is free to the user, and always has been.

Google has also historically paid very large sums for default placement in other companies' browsers and devices, the payments to Apple being the best known. Those contracts were the central conduct at issue in the United States antitrust case, and they survive it in modified form.

Privacy: what is collected, and what the controls really do

Google's own position is that search history is retained in a Google Account only when Web & App Activity is switched on, that auto-delete defaults apply to new accounts, and that anyone can search signed out or in Incognito. That is accurate as far as it goes, and the account controls are real and usable.

What sits behind them is the most comprehensive commercial behavioural profile any search engine holds, by design, because the advertising business depends on it: signed-in search and browsing activity, location history where enabled, device identifiers, IP address, and cross-product signals drawn from Android, Chrome, YouTube, Gmail and the ad network. The posture is opt-out, not opt-in.

Independent scrutiny has repeatedly tested that posture. The Brown v. Google class action over Incognito mode (N.D. Cal., No. 5:20-cv-03664) reached a settlement announced in December 2023 and finalised in 2024, under which Google agreed to delete billions of private-browsing records and change its Incognito disclosures. A separate multistate settlement with US state attorneys-general over location tracking was reported at $391.5m in November 2022, with further individual state settlements in 2023.

Jurisdiction matters too. Google is a United States company and is subject to US legal process, including subpoenas, National Security Letters and FISA 702 collection. In Europe it is subject to the GDPR, and Google Search is a designated core platform service under the EU Digital Markets Act — Alphabet was designated a gatekeeper on 6 September 2023 across eight services. Microsoft Bing, its nearest rival, is not designated, so Google carries obligations on self-preferencing, choice screens and data use that Bing does not.

The honest summary: Google is not a privacy-preserving search engine, does not claim to be one, and should not be treated as one.

AI Overviews and AI Mode

AI Overviews are the AI-written summaries that appear above results. They began as the Search Generative Experience in Search Labs, announced at Google I/O in May 2023, rolled out to all US users in May 2024, reached the UK, India, Japan, Indonesia, Mexico and Brazil in August 2024, and passed 100 countries on 28 October 2024.

AI Mode is the fuller conversational surface, with follow-up questions and multimodal input, introduced in March 2025 as a Search Labs experiment for US subscribers and widened through that year. On 1 December 2025 Google put Gemini 3 Pro into AI Mode behind a model selector for subscribers and expanded the surface to nearly 120 countries and territories in English. On 27 January 2026 Gemini 3 became the default model behind AI Overviews globally, and “Show more” on an Overview began handing off directly into AI Mode's chat interface on mobile.

Two limits are worth stating precisely, because both are commonly overstated. First, as of 19 August 2026 AI Mode is a surface alongside classic results, not a replacement for them; nothing published says it has become the default search experience. Second, there is no official global setting that turns AI Overviews off. Appending &udm=14 to a search URL, or using the “Web” tab that Google added in May 2024, produces a plain-links view — but that is a per-search filter, not an account preference. The contrast with DuckDuckGo, which publishes a permanently AI-free endpoint at noai.duckduckgo.com, is real.

Gemini at gemini.google.com is a separate assistant product, not Search, and the two are routinely confused.

The antitrust cases, and what actually changed

The detail here matters, because the headline version is usually wrong. Three European Commission fines landed between 2017 and 2019: €2.42bn over Google Shopping (June 2017), €4.34bn over Android (July 2018) and €1.49bn over AdSense (March 2019).

The US case began on 20 October 2020, when the Department of Justice and eleven state attorneys-general filed United States v. Google LLC in the District of Columbia. On 5 August 2024 Judge Amit Mehta ruled that Google “is a monopolist, and it has acted as one to maintain its monopoly,” finding its exclusive default-placement contracts unlawful. The same ruling put Google's share of general search at roughly 90%, and around 95% on mobile — a court finding rather than a vendor estimate.

The remedies decision followed on 2 September 2025. Per the Congressional Research Service analysis, it:

rejected structural relief — no Chrome divestiture and no conditional Android divestiture;
banned exclusivity — Google may not hold exclusive distribution agreements for Search, Chrome, Google Assistant or Gemini, but may still pay for default placement under non-exclusive deals of a year or less;
ordered data sharing — Google must supply search-index and click-and-query data to qualified competitors, described by the court as “the raw material that Google uses to improve search”;
ordered syndication — five-year licences of search results and text ads to qualified competitors, initially capped at 40% of queries;
required ad auction transparency, the final judgment running six years.

The remedies took effect on 3 February 2026. Google filed a notice of appeal that January challenging the data-sharing requirements, and the DOJ cross-appealed seeking the structural relief it did not get. Both appeals sit with the US Court of Appeals for the D.C. Circuit, the DOJ brief filed in July 2026, and no decision had issued as of 19 August 2026. A separate ad-tech case in the Eastern District of Virginia produced an April 2025 finding by Judge Leonie Brinkema that Google unlawfully monopolised parts of the ad tech stack.

What it is good at, and what it is bad at

Nothing else comes close on breadth and freshness of index, local and map-linked queries, navigational queries, language coverage, entity understanding or long-tail recall. If the question is whether a page about something exists anywhere on the web, Google remains the best instrument for finding out.

It is bad at giving a clean list of links. On commercial queries the page is dense with ads, AI summaries and Google-owned modules, and the organic results sit well down it. It is also, structurally, bad at not profiling the person using it: signed-in search feeds an advertising profile and the controls are opt-out. Publishers have a reasonable complaint that AI Overviews answer a query using their material without sending the click.

Two groups should look elsewhere. Anyone whose threat model includes a US-jurisdiction company holding query logs tied to an identity — journalists protecting sources, activists, people searching medical or legal questions under adversarial conditions — is poorly served by a product built on retained, identity-linked activity. Anyone who simply wants raw links quickly will find a plain-links engine less frustrating.

Common misconceptions

“Google indexes the whole web.” It does not, and says so. Google indexes a selected subset; pages it judges low-value are crawled and discarded.
“PageRank is how Google ranks pages.” PageRank is one historical signal among hundreds. The public toolbar PageRank score was retired in 2016.
“Incognito hides you from Google.” It does not hide activity from the sites visited, and Google's own logging was the substance of the Brown v. Google settlement.
“Google penalises sites.” Manual actions exist and are shown to site owners, but most ranking movement is algorithmic re-scoring, not a penalty.
“AI Overviews can be switched off in settings.” No official global setting exists as of 19 August 2026; the udm=14 parameter and the Web tab are per-query workarounds.
“Google is being broken up.” It lost on liability in 2024, but the September 2025 remedies ordered no divestiture of Chrome or Android, and the D.C. Circuit had not ruled on either appeal as of 19 August 2026.
“Google can no longer pay Apple to be the default.” It can. The remedy banned exclusive agreements, not paid default placement; non-exclusive deals of a year or less remain permitted.

Frequently asked questions

Does Google have its own search index?

Yes. Google crawls the open web with its own crawler, Googlebot, and ranks its own index — the stored copy of the web that Googlebot builds. It takes no web-link results from any other general search engine. Google does license structured feeds for specific answer boxes, such as flights and sports data, but the ranked organic results come from its own crawl and its own ranking systems.

What is Googlebot?

Googlebot is Google's web crawler, the program that fetches pages so they can be indexed. It runs in two documented variants, Googlebot Smartphone and Googlebot Desktop, and obeys the robots.txt token Googlebot. Because Google indexes mobile-first, the smartphone variant does most of the fetching. Google also operates GoogleOther, a generic fetcher, and publishes Google-Extended, a robots.txt token that governs use of a site's content for Gemini training.

What is Google's market share?

StatCounter Global Stats put Google at 91.31% of worldwide search referrals across all devices in July 2026. That figure measures page views on tracked sites rather than users or queries, so it is a widely used proxy rather than a census. Separately, the US court in the 2024 antitrust liability ruling found Google held roughly 90% of general search and about 95% on mobile.

Can you turn off Google's AI Overviews?

There is no official account-level setting that disables AI Overviews, as of 19 August 2026. Two per-search workarounds exist: appending &udm=14 to a search URL, and using the “Web” tab that Google added in May 2024. Both give a plain-links view for that one search only. This is a point of genuine contrast with DuckDuckGo, which publishes a permanently AI-free endpoint at noai.duckduckgo.com.

Is Google being broken up after the antitrust case?

No. Google lost on liability in August 2024, but the remedies decision of 2 September 2025 rejected structural relief — no Chrome divestiture and no conditional Android divestiture. It banned exclusive default agreements, ordered search-index and click-and-query data sharing with qualified competitors, and ordered capped syndication licences. Google appealed in January 2026 and the DOJ cross-appealed; the D.C. Circuit had not ruled as of 19 August 2026.

Does Google still pay Apple to be the default search engine?

It can, and the court order does not stop it. The September 2025 remedies banned exclusive distribution agreements for Google Search, Chrome, Google Assistant and Gemini, but expressly allowed continued payment for default placement under non-exclusive agreements of one year or less. The widely repeated claim that the ruling ended the Apple payments is wrong.

Is Google a private search engine?

No, and it does not claim to be. Signed-in search activity, location history where enabled, device identifiers, IP address and cross-product signals from Android, Chrome, YouTube and Gmail feed the advertising profile the business depends on. Google's privacy controls and auto-delete defaults are real, but the posture is opt-out. Its US jurisdiction also exposes retained data to subpoenas, National Security Letters and FISA 702 collection.

Which search engines use Google's index?

Startpage resells Google's results as its stated product, and Yahoo sourced its web results from Google between 2015 and 2018. Google itself takes results from no one. The 2025 US remedies decision also requires Google to license search results and text ads to qualified competitors on five-year terms, initially capped at 40% of queries, which may widen the set of engines carrying Google-derived results.