Mojeek

By Bill Hartzer · Updated 19 August 2026

What Mojeek is, and why its verdict is the least ambiguous on this site

Mojeek is a small British search engine that crawls the web with its own crawler and answers every query from its own index — the engine's own stored copy of the web. There is no partner feed. There never has been. At no point in the engine's history has it resold Google's, Bing's or anyone else's web results.

That sentence is worth dwelling on because it can be said of almost nothing else in the privacy-search category. Brave Search became fully independent in 2023 after two years as a hybrid. Qwant was found in 2019 to have no crawler at all while marketing itself as an independent European engine. Ecosia bought its results from Microsoft for well over a decade. DuckDuckGo and Startpage rely on other people's indexes by design. Mojeek's independence claim is the only one in the group that has never had to be revised.

It is deliberately old-fashioned. Keyword search, links as results, no personalisation, no behavioural signals, no filter bubble, and — by explicit editorial decision rather than by lack of ambition — no AI answer engine bolted above the results. Two users typing the same query get the same page.

It is also quietly infrastructure. Mojeek licenses its index through a Web Search API and a Site Search API, and it is one of the very few sources a developer can buy web results from that is not Google, Microsoft or Yandex. That scarcity is worth more now than it was: when Microsoft sharply raised Bing API pricing in 2023, every product built on a resold feed discovered how little it owned.

MojeekBot, and how the index is built

The crawler is MojeekBot, and Mojeek publishes unusually complete documentation for it — the kind of documentation an operator writes when it actually runs a crawler and wants site owners to work with it rather than around it.

What is published: MojeekBot obeys the Robot Exclusion Standard, matching the first MojeekBot record in a site's robots.txt file and falling back to the wildcard record if there is none. It obeys noindex, nocache and nofollow meta tags. It requests no more than one page per second from any given site. And it can be verified two ways — by reverse DNS lookup resolving to mojeek.com with forward confirmation, or against a published list of crawler IP addresses, which Mojeek notes may change at any time.

The index itself has grown steadily and Mojeek publishes milestones. It passed 1 billion pages in 2015, describing itself as the only UK search engine and one of a handful worldwide at that scale. Then 2.3 billion in May 2019, 4 billion in June 2021, 6 billion in October 2022, 8 billion in 2024 and 9 billion in 2025. No 2026 figure has been published, and the older numbers circulate widely enough that any Mojeek index size quoted without its year should be treated as stale.

The comparison that must accompany those numbers: Google's index is generally understood to run to hundreds of billions of pages. Qwant self-reports around 20 billion as of September 2023. Mojeek at roughly 9 billion is a genuine, independently built, web-scale index, and it is one to two orders of magnitude smaller than Google's. Both statements are true. The second is the one that determines whether Mojeek will work for any particular person.

Ranking is keyword and relevance based, with no personalisation and no behavioural inputs. Mojeek presents that as a feature, and for a specific kind of work it genuinely is: results are reproducible. If you need to know what an engine returns for a query — rather than what it returns for you, today, given everything it knows about you — that property is hard to find elsewhere.

Who runs it, and how it makes money

Mojeek is operated by Mojeek Limited, registered in England as company number 06918197, founded and led by Marc Smith, who built the crawler, the billion-scale indexing infrastructure and the ranking algorithms himself. It began in 2004 as a personal project at the Sussex Innovation Centre at the University of Sussex, launched publicly in 2006, and was incorporated in 2009 after taking investment. It is independently held: no acquisition, no Big Tech investor and no change of ownership was found as of August 2026.

Revenue comes from several small independent lines rather than one large one. Structurally, they are:

Mojeek Ads — its own search advertising, sold against the query.
The Web Search API — metered access to the index, sold to developers and to other search products.
The Site Search API — site-scoped search sold to website owners.
Direct support — a newsletter and supporter element.

The API is the strategically important one. There are perhaps a handful of independent web indexes on earth, and after the 2023 Bing API price rises anyone building a search or AI product without wanting to be a Microsoft reseller had a short list to choose from. Mojeek is on it. The engine Kagi has drawn on Mojeek among its blended sources from around mid-2023 — and the direction matters: Mojeek supplies Kagi, not the other way round. Mojeek does not consume anyone else's web index.

Mojeek does not publish financials. Any future claim about its revenue or headcount should come from its Companies House filings rather than from estimates.

Privacy: the structural argument and its limits

Mojeek states that it does not track, profile or personalise, sets no tracking cookies, and keeps no logs that build a user profile. It says it adopted an explicit no-tracking privacy policy in 2006 and describes itself as the first search engine to have done so — a claim that would predate DuckDuckGo's founding in 2008. The 2006 date is consistently reported; the primacy is Mojeek's own claim and has not been independently adjudicated.

The strongest privacy point about Mojeek is not in its policy at all. It is architectural: because Mojeek answers from its own index, no query is ever forwarded to another company. Startpage, Ecosia, Swisscows and DuckDuckGo all hand at least some queries to Microsoft or Google, anonymised to varying degrees. Mojeek has nobody to hand them to. Whatever weight a reader gives Mojeek's privacy policy, there is no second party on the other end of the request to have a policy about.

A shallow but reproducible independent check exists: a 2025 examination with uBlock Origin found no trackers on Mojeek's pages. That is more verification than most engines can show, and it is not a substitute for the thing that does not exist — no formal independent privacy audit of Mojeek was located. It is a small operator with no published third-party attestation.

The jurisdiction deserves an honest note rather than a flattering one. Mojeek is a United Kingdom company, with servers reported at a data centre in Maidstone, Kent as of 2017. Post-Brexit the UK sits outside the EU's GDPR enforcement structure, though UK law mirrors it, and the UK has the Investigatory Powers Act 2016, a more aggressive surveillance regime than most EU states operate. Mojeek's answer is the same as Brave's: it does not retain data linking a query to a person, so there is little to compel. For a threat model that includes state actors, that answer is data minimisation rather than legal muscle — and Mojeek is a small company without a large firm's capacity to fight compulsion.

The refusal to become an answer engine

In September 2025 Mojeek published a post titled "Mojeek is Not an Answer Engine," arguing that Mojeek is about search and that AI is not the answer. That is an explicit strategic position, not an unfinished roadmap item, and it is increasingly what distinguishes the engine.

The nuance is that Mojeek is not absolutist about the technology. It offers result summarisation built on an open-weight Mistral model — a utility that summarises results rather than a generative system that replaces them, and one that runs on a European open-weight model rather than a commercial American API. Whether that summarisation is on by default, and whether it has a persistent off switch, could not be established from a primary source for this page. Further back, in 2017, Mojeek ran an unusual experiment in emotional or sentiment search, built with EMRAYS Technologies, letting users filter results by emotional tone.

What does not happen is a generated answer being placed above the links. As of August 2026 Mojeek's default surface is results. For a reader who has grown tired of scrolling past a paragraph of synthesis to reach the sources, that is the entire pitch.

Reddit, and why independence keeps getting harder

In 2024 Reddit blocked non-Google crawlers, cutting Mojeek and every other independent engine off from Reddit's content while Google retained access under a paid arrangement.

This is the clearest single illustration of how search consolidation now sustains itself. The barrier to entry in search used to be technical and financial: you needed to build a crawler, store billions of pages and rank them, and that was hard enough. Mojeek proves a small company can still do it. What a small company cannot do is write the cheque that buys access to a large platform's content, and increasingly the material people actually search for — forum discussion, reviews, first-hand accounts — sits behind exactly those arrangements.

The result is that an independent engine can crawl the open web perfectly well and still be missing the answer, not because its technology failed but because the content was licensed to somebody else. When Mojeek returns nothing useful for a question whose best answer is a Reddit thread, that is not a defect in Mojeek. It is the shape of the market. It is also a problem Mojeek cannot solve on its own, and there is no obvious reason to expect it to get better.

Who it suits, and who it does not

It suits anyone who wants to see the web without Google's hand on it. Because results are unpersonalised and identical for everyone, Mojeek is unusually useful wherever reproducibility matters — comparing what different engines return, checking whether a page is indexed anywhere outside the duopoly, or simply looking at a web that has not been optimised for engagement. It surfaces small, old and obscure sites that the large engines have quietly stopped showing; the long tail of the pre-platform web is disproportionately visible here. Its crawler documentation is exemplary. And it never sends a query to anyone else.

It is a poor choice for everything that requires scale. An index of roughly 9 billion pages against Google's hundreds of billions means that on a large share of specific queries Mojeek simply will not hold the page you need. Freshness lags, because a small crawl budget cannot recrawl news-velocity sites at Google's cadence. Non-English coverage is thin. There is no local pack, no meaningful maps, no shopping, and no rich verticals. Relevance on ambiguous commercial queries is well behind Google's. And content behind crawler blocks it cannot negotiate past is invisible.

So: it is a poor fit for anyone who needs one engine for mainstream consumer searching — local businesses, shopping, breaking news, anything that lives on the platform web. It is a poor fit for anyone working mainly in a language other than English. And it is the wrong engine for anyone who wants AI answers, since Mojeek has declined to build them on purpose.

The fair verdict: Mojeek is the most genuinely independent search engine a Western user can reach, and it works best as a second engine for anyone who cares about search diversity rather than as a Google replacement. Presenting it as a full substitute is the mistake most enthusiast write-ups make, and it sets users up to be disappointed by an engine that is doing exactly what it says it does. Presenting it as the thing you check when you want the web without the duopoly's hand on it is precisely right.

One last honest note on scale: there is no published market-share figure for Mojeek and no audited query volume. StatCounter does not break it out; it falls well inside "Other", and any percentage attributed to it should be treated as invented. Mojeek's significance is structural rather than commercial. It matters because it exists and is independent, not because many people use it.

Frequently asked questions

Does Mojeek have its own index?

Yes, entirely. Mojeek crawls the open web with its own crawler, MojeekBot, and answers every query from an index it built and operates. It has no partner feed for web results and has never had one at any point in its history. Its published index size was around 9 billion pages in 2025, up from 6 billion in October 2022 and 1 billion in 2015.

Is Mojeek owned by Google or Microsoft?

No. Mojeek is operated by Mojeek Limited, an independently held English company registered as number 06918197, founded and led by Marc Smith. No acquisition, Big Tech investment or change of ownership was found as of August 2026. It does not use Google's or Bing's results either — it is one of very few Western engines that resells nobody's index and buys nobody's feed.

How big is Mojeek's index compared to Google's?

Mojeek published a figure of about 9 billion pages in 2025. Google's index is generally understood to run to hundreds of billions. That is a gap of one to two orders of magnitude, and it is the single fact that determines whether Mojeek will work for a given person: it is a genuine web-scale index, and it will still not hold the page you need on a large share of specific queries.

Why does Mojeek not show Reddit results?

Because Reddit blocked non-Google crawlers in 2024, while Google retained access under a paid arrangement. This is not a fault in Mojeek's technology. It is the clearest example of how search consolidation is now maintained through content licensing rather than engineering: a small company can build a crawler, but it cannot outbid Google for access to a large platform's archive.

Does Mojeek have AI answers?

Not in the usual sense, and deliberately so. In September 2025 Mojeek published a post titled "Mojeek is Not an Answer Engine", stating an explicit position rather than describing an unbuilt feature. It does offer result summarisation using an open-weight Mistral model, which summarises results instead of replacing them, and it ran a sentiment-search experiment back in 2017. No generated answer is placed above the links.

Was Mojeek the first search engine not to track users?

Mojeek says so, and the date it gives — a no-tracking privacy policy adopted in 2006 — is consistently reported and does predate DuckDuckGo's 2008 founding. The claim to being first has not been independently adjudicated, so it is best treated as the company's own. What is beyond dispute is that Mojeek, unlike DuckDuckGo, does not rely on Bing for any part of its results.

Is Mojeek private?

Its strongest privacy property is structural: answering from its own index means no query is ever handed to a third party, which is not true of Startpage, Ecosia, Swisscows or DuckDuckGo. Mojeek states it does not track, profile or personalise, and a 2025 check with uBlock Origin found no trackers on its pages. No formal independent privacy audit exists, and the UK's Investigatory Powers Act is a legitimate consideration.

Can Mojeek replace Google?

For most people, not as a sole engine. It is excellent for seeing the web without personalisation, finding small and old sites, and checking what exists outside the duopoly. It is weak on freshness, non-English content, local businesses, shopping and anything on the platform web. The realistic use is as a second engine for people who care about search diversity, not as a straight substitute.