What was the first search engine?

The short answer

The first internet search engine was Archie, which went live on 10 September 1990. The first search engine that indexed the full text of web pages — which is what almost everyone means by the phrase today — was WebCrawler, launched on 21 April 1994 by Brian Pinkerton at the University of Washington.

Both are true at once because they answer different questions. Archie predates the public World Wide Web and never indexed a web page: it catalogued the filenames sitting on public FTP servers. WebCrawler was the first tool that would find a page because of the words inside it. Between them sit four more claimants, each called the first search engine somewhere reputable, and each in fact first at something narrower:

• First internet search engine — Archie, 10 September 1990.
• First web robot — the World Wide Web Wanderer, June 1993.
• First searchable index of web resources — W3Catalog, 2 September 1993.
• First tool to combine crawling, indexing and searching — JumpStation, 12 December 1993.
• First engine built from publisher-submitted metadata — ALIWEB, announced late 1993, launched May 1994.
• First full-text web search engine — WebCrawler, 21 April 1994.

Anyone who gives you a single name without a qualifier has picked a definition without telling you which one.

Why the question has six answers

A modern search engine does four separable things. It discovers documents, usually by following links from documents it already knows. It indexes them — builds its own stored copy, so queries are answered from local data rather than the live network. It searches that index. And it ranks the matches. In 1990 no tool did all four; in 1993 several did two or three each, and picked different two or three. People arguing about which came first are not disagreeing about the record, which is well documented, but about the definition.

Three distinctions do most of the work. Crawling is not indexing for search: the Wanderer crawled in June 1993, but was built to measure the web's growth. Indexing is not full-text indexing: JumpStation stored titles and headings, Lycos stored the title, headings, the hundred most heavily weighted words and the first twenty lines — neither stored the document. Searching is not ranking: JumpStation returned matches in whatever order the index yielded them. And one distinction matters more than all of them — the internet is not the web. The internet carried FTP, Usenet, email and Gopher long before it carried HTTP, and the web became publicly available only in 1991.

Archie, 10 September 1990 — the first internet search engine

Archie was built at McGill University in Montreal by Alan Emtage, with Bill Heelan and Peter J. Deutsch credited alongside him. By 1990 hundreds of public anonymous-FTP servers held software and documents, with no way to find which one had the file you wanted short of connecting to each in turn.

Its mechanism was deliberately narrow. Archie kept a list of known FTP hosts, connected to each roughly monthly and downloaded full recursive directory listings — hostname, path, filename, size, date — merged into one database of strings and matched by exact filename, substring or regular expression. It read nothing inside the files, summarised nothing and ranked nothing; if you did not already know roughly what a thing was called, it could not help you. A companion database, Whatis, held human-written descriptions of software packages, and was the only part that told you what something was rather than where it was. You queried it by a local archie client, by Telnet, or — the standard method on a mail-only connection — by email, sending find <string> and getting results back by post.

The 1989 versus 1990 conflict

Sources genuinely disagree on the year. Wikipedia's timeline of web search engines gives 10 September 1990 for Archie going live, the most specific dated claim available. McGill's own bicentennial page and the Internet Hall of Fame both say 1989, dating the creation rather than the public service. Wikipedia's Archie article separately says the implementation was written in 1990 and that Archie first appeared in 1986, a figure that appears to describe precursor work. Reading 1989 as the work and 10 September 1990 as the live service reconciles most of it — but no source states that reconciliation, so treat the year as contested.

Version 3.5 in 1996 was the last release. The final legacy production server, at the University of Warsaw, ran until 2023, and the preservation project The Serial Port launched a public instance from its recovered source on 11 May 2024 — but as of 19 August 2026 that instance is offline.

1993 — four claimants in seven months

By mid-1993 the young web was growing fast enough that several people independently decided it needed indexing. Four tools followed within seven months, and three of them are called the first web search engine somewhere reputable. They divide neatly by what they refused to do.

The World Wide Web Wanderer (June 1993) is the one that crawled. Matthew Gray, at MIT, wrote a program that fetched a page, extracted its links and followed them, generating an index called the Wandex later that year and charting the web's growth until late 1995. But Gray built it to measure the web, and Wikipedia's article states plainly that its purpose was not to be a search engine. The disclaimer is the author's own, which is why the Wanderer is credited as the first web robot rather than the first search engine.

W3Catalog (2 September 1993) is the one that did not crawl and shipped first anyway. Oscar Nierstrasz, at the University of Geneva, exploited the fact that several good hand-maintained lists of web resources already existed: he mirrored those pages, broke them into individual entries and put a searchable Perl interface over the result. Wikipedia's timeline calls it the world's first web search engine. It is a genuine searchable index delivered before anything else — but an index of other people's link lists, not of the web. It ran until 8 November 1996.

ALIWEB refused to crawl on principle. Its name stands for Archie-Like Indexing in the WEB, the title of Martijn Koster's own conference paper, and it has no single launch date: Nexor, where Koster worked, says he built it in 1992; it was announced in October or November 1993; and it was presented at the First International Conference on the World Wide Web at CERN in May 1994, the date usually cited as its launch. Site administrators wrote a small structured index file — conventionally site.idx at the server root — listing each resource with a title, description, keywords and URI, and registered its location once; ALIWEB fetched those files on a schedule and merged them. What you searched was a pile of self-descriptions, not page text. Koster argued that crawlers imposed considerable network overhead, could overload servers and destroyed the structure of what they gathered — and it is no coincidence that the same man wrote robots.txt in 1994. He built the discipline that made large-scale crawling acceptable, and crawling then destroyed his engine.

JumpStation (12 December 1993) did all three jobs. Jonathon Fletcher, at the University of Stirling, began indexing on that date and announced it on 21 December; Wikipedia's timeline describes it as the first WWW resource-discovery tool to combine all three essential features, and the first that behaved, and appeared to the user, the way current web search engines do. Its limits were real — titles and headings only, and no ranking whatsoever. By late 1994, when Fletcher left the university having failed to persuade investors, the university included, to fund it, JumpStation held 275,000 entries across 1,500 servers. Then it stopped, which is why a tool that got the architecture right a year early is not a household name.

WebCrawler, 21 April 1994 — the first full-text web search engine

Brian Pinkerton launched WebCrawler from the University of Washington on 21 April 1994 with an index of just over 4,000 websites. Its crawler had produced a first list of the top 25 websites on 15 March 1994, and it served its one-millionth query on 14 November 1994 — a widely cited figure originating in Pinkerton's own accounts.

One design decision is why it matters. Every predecessor indexed something about a document: its filename, URL, title, headings, or a description its owner had written. Pinkerton downloaded and indexed the entire text of each page. That is what makes it possible to find a phrase appearing nowhere in a title, and it is the model every general web search engine since has used — AltaVista, Lycos, Infoseek, Excite, Google, Bing.

Its afterlife is a lesson in a brand outliving its product. AOL bought it in June 1995 and sold it to Excite in April 1997; when Excite@Home collapsed in 2001, InfoSpace acquired it and switched it to serving Excite's database. The domain still resolves in 2026, but it has not crawled anything since 2001.

The claims side by side

• Archie — 10 September 1990, Alan Emtage, McGill University. Indexed FTP filenames and directory paths. No crawl, no page content, no ranking. First internet search engine.
• Veronica and Jughead — 1991 to 1993. The same idea for Gopher: they searched Gopher menu titles, not content. First Gopher search tools.
• World Wide Web Wanderer / Wandex — June 1993, Matthew Gray, MIT. Crawled and indexed, but built to measure the web's growth; its author does not claim it was a search engine. First web robot.
• W3Catalog — 2 September 1993, Oscar Nierstrasz, University of Geneva. Mirrored existing link lists into a searchable database. No crawl. Ran to 8 November 1996. First searchable web index, by date.
• JumpStation — 12 December 1993, Jonathon Fletcher, University of Stirling. Crawled, indexed and searched; titles and headings only, no ranking. 275,000 entries across 1,500 servers by late 1994. First web search engine, by function.
• ALIWEB — built 1992, announced late 1993, launched May 1994, Martijn Koster at Nexor. Merged index files site owners wrote and registered by hand. No crawl, by design. First submission-based web search service.
• WebCrawler — 21 April 1994, Brian Pinkerton, University of Washington. Indexed the full text of pages. First full-text web search engine.

Correcting Wikipedia on ALIWEB

The English Wikipedia article on ALIWEB opens by calling it the first internet search engine. That is wrong, and it has propagated widely, because the opening sentence of a Wikipedia article is what gets copied.

Archie went live on 10 September 1990, more than three years before ALIWEB was announced — and ALIWEB's name, Archie-Like Indexing in the WEB, is a reference to Archie. A tool cannot be the first internet search engine when it is named after an earlier one; the error is self-refuting on the face of the title. The narrower claim, that ALIWEB was the first web search engine, is more defensible but still contested. Nexor's own retrospective asserts ALIWEB is acknowledged as the world's first search engine and criticises the BBC for crediting WebCrawler — but Nexor compares it only to WebCrawler (April 1994), not to the Wanderer, W3Catalog or JumpStation, two of which predate even ALIWEB's announcement.

Its real significance is not chronological anyway. ALIWEB is the origin of the model in which publishers describe themselves and the engine trusts them, and the first demonstration of why that fails: coverage was capped by how many administrators bothered to maintain a file, and self-description was unverifiable, because there was no page text to check the claim against. The same weakness later hollowed out the HTML keywords meta tag.

The verdict, and the names that do not qualify

If someone asks with no qualifier and wants one name, the least misleading answer is Archie, 1990 — followed immediately by the sentence that makes it true: it searched FTP filenames, not web pages, because the web did not yet publicly exist. If what they meant was the first thing that worked the way search works now, the answer is WebCrawler, 21 April 1994. Four other names get offered and none holds:

• WHOIS (1982) is a lookup protocol for domain and network registration records. It does not search over content or files, and Wikipedia's timeline of web search engines does not list it.
• AltaVista (15 December 1995) is misremembered as first because for many people it was the first they used. Its achievement was scale — Scooter, a multi-threaded crawler keeping many fetches in flight at once, over a full-text index large enough to sit in the memory of the AlphaServer hardware the project existed to demonstrate. Eighteen months after WebCrawler.
• Lycos is one of the earliest crawler-based engines, but three months after WebCrawler — and its dates get collapsed. The research service went online at Carnegie Mellon on 20 July 1994 with 54,000 documents, it was announced on 12 August 1994, and the commercial lycos.com site is dated 13 April 1995. Three events, not one contested date.
• Yahoo (1994) and Google. Yahoo began as a hand-built directory and did not run a crawler of its own until 2004. google.com was registered on 15 September 1997 and Google Inc. incorporated on 4 September 1998 — seven years after Archie. Google's contribution was ranking by link analysis, not search itself.

Frequently asked questions

Sources

• en.wikipedia.org/wiki/Archie_(search_engine)
• en.wikipedia.org/wiki/Timeline_of_web_search_engines
• 200.mcgill.ca/history/creation-of-the-first-internet-search-engine/
• internethalloffame.org/inductee/alan-emtage/
• en.wikipedia.org/wiki/W3Catalog
• en.wikipedia.org/wiki/World_Wide_Web_Wanderer
• en.wikipedia.org/wiki/JumpStation
• en.wikipedia.org/wiki/WebCrawler