This site is a reference work about search engines themselves — who runs them, how they find pages, what they cost to operate, and the question almost no other list answers: whose index are you actually reading when you type into the box?
That question sounds pedantic. It is the most useful fact there is about any search engine, and it is routinely omitted. A search engine is really three separate machines bolted together. The crawler is a program that fetches pages and follows the links it finds. The index is the engine's own stored, searchable copy of what the crawler fetched. The ranking system decides which of those stored pages to show you, in what order. Building and running the first two is enormously expensive and largely invisible; building the third is where the visible product lives. So a great many companies build only the third — a search box, an interface, a privacy promise, an AI answer layer — and buy the links from somebody who already owns an index.
None of that is dishonest, and this site does not treat it as a scandal. Startpage relays your query to Google and Bing on purpose and says so; that anonymising relay is its product. But it changes what an "alternative" means. If you leave Google because you dislike how Google ranks the web, and you move to an engine that re-serves Bing, you have changed the interface, the privacy policy and the advertising, and you have not changed who decides what the web looks like. If you moved to escape Bing, and you land on something else fed by Bing, you have not moved at all.
What this site does differently
Most "list of search engines" pages are a list of logos. This one carries, at the top of every engine's page and on every card, a plain statement of where the results come from, in one of five states: own index, hybrid index, resold results, not a web index, or no longer available. Fifty-eight engines are profiled — the ones people actually search for, from Google down to a one-person Swedish crawler, plus the twenty-three that no longer work and are still typed into address bars every day.
The finding, once you assemble it, is stark. Genuine general-purpose web indexes are rare. Google and Microsoft Bing each run one. Yandex, Baidu, Naver, Seznam and Cốc Cốc run regional ones. Brave has run its own since April 2023, when it removed the last of its Bing API calls; Mojeek has run its own since the mid-2000s and has never resold anyone; Marginalia is a working index built by one person. Perplexity and OpenAI have built retrieval indexes for their answer engines. That is close to the whole list. Almost everything else on any "alternatives" page is downstream of one of them — including Yahoo, which has had no crawler of its own since 2010.
Who it is for
- People changing their default engine. The useful comparison is not which logo you prefer but which engines are genuinely different underneath, what each one gives up in coverage, and who is paid when you search.
- People who want to know how search works. How crawling and indexing actually happen, who was first, why AltaVista lost, what a metasearch engine was for, and why "own index" is a harder line to hold in 2026 than it was in 1998.
- Researchers and journalists who need a citable fact. Ownership, launch dates, supplier contracts and market share, each attributed to a named source and dated, with conflicts printed rather than smoothed over.
How the facts here are handled
Anything that can move carries a date and a source. Market share is quoted as a measurement by a named measurer, not as a property of the engine: StatCounter's figures count page views on sites carrying its tracker, which is a proxy and not a census, and it undercounts engines whose users never leave the engine. Where two credible sources disagree, both appear. South Korea is the clearest case — as of the most recent figures gathered here, StatCounter put Naver at 40.9% of Korean search in July 2026 while the domestic measurer InternetTrend put it at 63.8% in March 2026, and the gap is a methodology difference, not an error. Where something could not be verified, the page says so instead of guessing.
There are no scores, no rankings and no verdicts on quality. "Resold results" is a neutral structural fact, not a criticism. What the site will do is say plainly when an engine's marketing and its architecture disagree, and when a claim rests on a company press release that nobody has independently checked.
It is also a moving subject, which is why the dates matter. Microsoft retired its public Bing Search APIs in August 2025, cutting off the cheapest route to reselling web results. Ask.com — running in some form since 1997 — closed in May 2026. Ecosia and Qwant, both long dependent on other people's indexes, jointly built one and began serving traffic from it in August 2025. Any statement here about who supplies whom is true as at its stated date and no longer.