Gigablast: offline since April 2023

By Bill Hartzer · Updated 19 August 2026

What Gigablast was, and why it mattered

Gigablast was an independent web search engine written from scratch in C and C++ by one person, Matt Wells, and operated from its public beta in July 2002 until it silently went dark in early April 2023. Its significance is wildly disproportionate to its traffic. For twenty-one years it was one of a genuinely tiny number of organisations anywhere — perhaps half a dozen worldwide — that crawled the open web at scale and ranked its own index (the engine's own stored copy of the web, built by its own crawler) rather than reselling Google's or Microsoft's results.

It is listed here as no longer available, and that verdict describes only its present state. The historical position needs stating separately, because filing Gigablast as "defunct" and stopping there throws away the only thing that made it notable: Gigablast was a fully independent own-index engine for its entire life. It never took a partner feed. The traffic went the other way — other engines bought results from Gigablast.

What a user actually got was a plain, fast, ad-light results page with unusually powerful operators: full Boolean algebra, site, URL and date filters, and a related-concepts sidebar called Giga Bits. Results were noticeably thinner than Google's on popular queries, but often surfaced pages Google had buried or dropped.

How it worked: Gigabot, and a hand-built C++ stack

Every layer of the engine was custom software. The crawler, the storage layer, the indexer, the query parser and the ranker were all written in C and C++ for Linux on x86 and x86-64 hardware. Wells left Infoseek in 1999 specifically in order to build the thing from nothing.

The crawler identified itself in server logs as Gigabot, commonly seen in the form Gigabot/3.0 (http://www.gigablast.com/spider.html), and site operators allowed or blocked it with a User-agent: Gigabot block in robots.txt. The name is documented in Gigablast's own open-sourced code and spider documentation and in webmaster crawler directories; the exact version string changed over the years, so treat the full string as illustrative rather than canonical.

Ranking was algorithmic and hand-built, with support for Boolean operators, phrase and field search, and the statistical related-term extraction behind Giga Bits. There was no personalisation.

On index size, the published figures conflict and this page prints both. The architecture was designed to scale to 200 billion pages, but that was a design ceiling, not an achievement. For 2015, Wikipedia records an index of "over 12 billion web pages," while Gigablast's own press release of 1 July 2015 describes "its own searchable index of over a billion pages." Those are figures for the same year that differ by an order of magnitude, and they cannot be reconciled from primary sources. The company's own wording is the more conservative and, as a formal self-description, arguably the more defensible — though "over a billion" is vague enough not to strictly contradict twelve.

Gigablast as a supplier, not just a destination

The commercially important fact about Gigablast is that its index was licensed to other search products. That is why the name turns up in the plumbing of engines that never mentioned it on the front page. Documented consumers of Gigablast results include Ixquick, Clusty, Zuula, Snap, Blingo and the Internet Archive.

The Internet Archive engagement is the largest documented deployment. Announced on 1 July 2015, Gigablast provided the search capability over the Archive's stated 400 billion-plus web documents, with Wells providing technical support for the deployment; Gigablast positioned its engine as more scalable for that job than SOLR or Elasticsearch.

The revenue lines, all of them modest, were index and API licensing to other engines and institutions; self-serve crawling and search as a service, where a customer paid to have a defined collection crawled, indexed and searched; advertising on the consumer site, which was never the core; and the Private.sh white-label arrangement described below. In its final years the front page called Gigablast "the most advanced B2B search engine ever built," which is a fair reflection of where the money was expected to come from. There is no public record of venture funding, no published revenue figure and no employee count — every credible account describes an essentially one-person operation, bootstrapped throughout.

Private.sh, and a privacy design worth understanding

Gigablast's consumer site made conventional non-tracking claims, and no independent audit of them has been published. The substantive privacy story is elsewhere: Private.sh, launched by Private Internet Access in partnership with Gigablast and announced on 14 November 2019. Its architecture is structurally unusual enough to be worth describing precisely.

The user's query is encrypted in the browser using Gigablast's public key, so the Private.sh proxy carrying it cannot read it.
The encrypted query is relayed through a Private.sh proxy to Gigablast, which decrypts and answers it but never sees the user's IP address or browser fingerprint.
Results are encrypted by Gigablast and returned via the proxy, then decrypted in the browser using a fresh temporary keypair generated for each search.

The point of the split is that neither party alone can link a query to a person: the proxy knows who you are but not what you asked, and the index knows what you asked but not who you are. That separation of roles is the whole design, which is why it matters that Private.sh was built by Private Internet Access while Gigablast merely supplied the index — it was not Gigablast's own product. Private.sh was later associated with the freenode holding company, and the gigablast.io property describes itself as a joint venture between freenode and GigaBlast.

Separately, Gigablast claimed in 2010 to be the leading clean-energy search engine, drawing 90% of its power from wind. That is a company claim and was never independently verified.

The open-source release

In July 2013 Wells published the complete search engine — crawler, indexer, query engine and all — as open source under the Apache License 2.0. It reached the front page of Hacker News, and the repository has since been forked and modernised by third parties.

This is genuinely rare. Production, web-scale search engines whose full source has ever been made public can be counted on one hand, and the release is a contribution to the field that outlived the service itself. The caveat is age: the main repository's own description carries a date of November 2017, which suggests open-source maintenance had largely stopped years before the service did. Anyone going to the code today should expect a 2017-era C++ codebase.

It is also worth being clear about what the code is not. Gigablast was a classical information-retrieval engine for its entire life and shut down before generative AI became standard in search interfaces. There was no chat, no summarisation and no generated answers; Giga Bits was statistical related-term extraction, not AI in the current sense. The repository is a crawler and an IR stack, and it is occasionally miscited in AI-search discussions as something more.

Timeline

1999 — Matt Wells leaves Infoseek to build a search engine from scratch in C++.
2000 — Gigablast founded.
21 July 2002 — Public beta launch. Sources give 2000 for the company and 2002 for public availability; both are consistent.
24 September 2003 — Search Engine Watch publishes a long interview with Wells, one of the best contemporary accounts of the engine's design philosophy.
2005 — Gigablast launches blog search.
2010 — Claims status as the leading clean-energy search engine, 90% wind-powered.
January 2012 — Wells launches a separate event-search project, FlurbiT.
July 2013 — Full source code released under Apache 2.0.
2013 — Reports circulate that Yippy, Inc. will acquire Gigablast. Wells later denies any acquisition took place.
1 July 2015 — Deal announced to provide search over the Internet Archive's 400 billion-plus documents; the same release restates Gigablast's independence from Yippy.
14 November 2019 — Private.sh launches.
Early April 2023 — gigablast.com stops serving search, with no warning and no statement.
May 2023 — Mojeek publishes "Farewell Gigablast (for now?)," the most substantial obituary the engine received.

What state the domains are in today

This needs care, because the domains behave confusingly and it is easy to get wrong in either direction.

gigablast.com resolves and returns an HTML page, checked 19 August 2026. It is the old front page — navigation for home, add url, blog, about, privacy, syntax, api and login, the "most advanced B2B search engine ever built" positioning, and promotion of private.sh. Whether it executes web searches could not be confirmed; the search backend could not be exercised. The accurate characterisation is that a residual front end still resolves, that there is no evidence the index is being crawled or served, and that no source reports Gigablast's return.

gigablast.io, fronting for gigablast.org, also resolves. It presents itself as an encrypted proxy-based private search engine without logging and with auditable code, and as a joint venture between freenode and GigaBlast. Its copy is Private.sh-lineage material and repeats the 2015 claims verbatim. Whether it is currently backed by a live Gigablast index or is stale marketing has not been verified.

So: two statements you will see elsewhere are both wrong. "Gigablast is dead and the domain is gone" is factually incorrect — branded pages still resolve. "Gigablast is back" is equally unsupported. The service is what died, not the domain.

What it was good at, what it was bad at, and the bus factor

What it was good at was being independent, and staying independent for twenty-one years. Gigablast is the strongest available proof that a web-scale crawler and index can be built and run outside the duopoly: one person, C++, no venture capital, and it ended up supplying search over the Internet Archive. Its query syntax was excellent for anyone who genuinely knows Boolean operators.

What it was bad at was coverage and freshness. An index in the low single-digit billions of pages against Google's hundreds of billions means popular queries returned less, long-tail queries frequently returned nothing, and news or recent content was often stale. The interface was dated for most of the engine's life.

The structural problem was a bus factor of one. When Wells stopped, everything stopped — no announcement, no handover, no data preservation plan, and an open-source repository that had already gone unmaintained for years. That is less a criticism of Wells than the fact that makes Gigablast a cautionary tale as well as an inspiring one. An independent index run by one person is a real index and a fragile institution at the same time.

Nobody should be using Gigablast today, because it is not operating. Anyone who lands on a residual Gigablast page in 2026 and types a query should not expect current results. Readers who want a working independent index should look at engines that still crawl, such as Mojeek or Marginalia; developers who want the code should go to the GitHub repository with 2017 expectations.

Common misconceptions

"Gigablast was a metasearch engine, or it used Google's results." Flatly wrong, and the most damaging misconception about it. Gigablast crawled the web itself with Gigabot and ranked its own index. The flow ran the other way: other engines bought results from Gigablast.

"Yippy bought Gigablast in 2013." Widely repeated and untrue. Wells denied any acquisition took place, and Gigablast's own July 2015 press release states that it remained independent with no affiliation to Yippy, Inc.

"It shut down because it was acquired." No. It went offline in April 2023 with no announcement, no acquirer and no explanation. The absence of any statement is itself part of the record.

"Gigablast indexed 200 billion pages." No. That was the architecture's design ceiling. Reported actual index sizes for 2015 range from "over a billion" in Gigablast's own release to "over 12 billion" per Wikipedia — two to three orders of magnitude below the ceiling.

"Private.sh was Gigablast's product." Private Internet Access built it; Gigablast supplied the index. The separation of the two roles is the entire privacy design.

Frequently asked questions

Is Gigablast still working?

No. Gigablast stopped serving search in early April 2023, without warning and without any statement from its creator. Checked 19 August 2026, some Gigablast-branded pages still resolve — gigablast.com returns the old front page and gigablast.io serves Private.sh-era marketing copy — but there is no evidence the index is being crawled or served, and no source reports the engine's return.

Did Gigablast have its own index?

Yes, for its entire twenty-one-year life. Gigablast crawled the open web with its own crawler, Gigabot, and ranked its own index, and it never took results from Google, Microsoft or anyone else. The relationship ran the other way: Ixquick, Clusty, Zuula, Snap, Blingo and the Internet Archive all consumed Gigablast results. That independence is the single most important fact about the engine.

Who created Gigablast?

Matt Wells, who left Infoseek in 1999 specifically to build a search engine from scratch. He founded Gigablast in 2000 and launched the public beta on 21 July 2002. He wrote the crawler, storage layer, indexer, query parser and ranker himself in C and C++ for Linux, and by every credible account ran the operation essentially single-handedly, without venture funding, until it went offline in 2023.

How big was Gigablast's index?

The published figures conflict and this page reports both. For 2015, Wikipedia records over 12 billion web pages, while Gigablast's own press release of 1 July 2015 says "over a billion pages." Those cannot be reconciled from primary sources. The often-quoted figure of 200 billion pages was the architecture's design ceiling, not a measured index, and should never be cited as an achieved size.

Did Yippy buy Gigablast?

No, although the claim is widely repeated. Reports in 2013 said Yippy, Inc. had agreed to acquire Gigablast, but Matt Wells subsequently stated that no acquisition took place, and Gigablast's own press release of 1 July 2015 explicitly reiterated that it remained independent with no affiliation to Yippy. Gigablast was never acquired and remained independent until it went offline in 2023.

Is Gigablast open source?

Yes. In July 2013 Matt Wells released the complete engine — crawler, indexer, query engine and ranker — under the Apache License 2.0, and the repository has been forked and modernised by third parties. Very few production web-scale search engines have ever had their full source published. The main repository's own description carries a November 2017 date, so expect a codebase that stopped being maintained years before the service did.

What was Private.sh?

Private.sh was a private search service launched on 14 November 2019 by Private Internet Access using Gigablast's index. Queries were encrypted in the browser with Gigablast's public key and relayed through a PIA proxy, so the proxy knew the user but not the query, and Gigablast knew the query but not the user. Private Internet Access built it; Gigablast only supplied the index.

Why did Gigablast shut down?

No reason was ever given. The site went offline in early April 2023 with no warning, no announcement and no acquirer, and Matt Wells issued no statement. The most substantial account of the closure is Mojeek's May 2023 post, written by a direct competitor, which called it a sad day for search and noted how little coverage the shutdown received.

Gigablast