Brave Search

By Bill Hartzer · Updated 19 August 2026

What Brave Search is, and where its results actually come from

Brave Search is a general-purpose web search engine operated by Brave Software, the company behind the Brave browser. It returns a conventional page of ranked links, with image, video and news verticals and an optional AI-generated summary at the top. It is the default engine in the Brave browser in most regions, and it has been available to anyone, in any browser, since 2021.

The structurally important fact is this: Brave Search answers a query from its own index — the engine's own stored copy of the web, assembled by its own crawler. It does not forward the query to Google or Bing and repackage what comes back. That places it in a very small group. Alongside Google, Bing, Yandex, Baidu, Mojeek and a handful of others, Brave is one of the few organisations anywhere that actually crawls and ranks the open web at scale.

Why that matters is easy to underrate. Most engines marketed as "alternatives" are interfaces over somebody else's results: change the commercial terms of the underlying API and the alternative disappears overnight. Brave would still return results if Microsoft cut it off tomorrow, because Microsoft is not in the loop.

The independence is also recent enough that a lot of published writing about Brave Search is wrong. At launch in June 2021, about 87% of results came from Brave's own index and the remainder from a third-party API — Bing. By mid-2022 that was roughly 93%. The last of the Bing calls came out of the web results page on 27 April 2023, and Brave's own image and video index followed in August 2023. So "Brave is just Bing with a privacy wrapper" was partly true in 2021 and has been false since 2023, which is exactly why the belief has proved so durable.

How the index gets built: Bravebot, Tailcat and the Web Discovery Project

Three components do the work.

The crawler. Brave operates a crawler branded Bravebot, which fetches pages so they can be stored and ranked. It publishes a declared user-agent token, Bravebot/1.0, which means a site operator who wants to allow or disallow it in robots.txt can identify it cleanly. That was a live point of friction earlier in the engine's life: Brave's documentation at times stated that the crawler did not advertise a differentiated user-agent, on the reasoning that sites which permit only Googlebot would otherwise shut out a new entrant. Publishers objected, reasonably, that they could not block what they could not name. Older write-ups describing a deliberately generic Chrome user-agent reflect that earlier position, not the current one.

The index itself. Brave did not build a web-scale index from nothing. In March 2021 it acquired Tailcat, the search stack developed at Cliqz, a subsidiary of the German publisher Hubert Burda Media. Tailcat gave Brave a working independent index to start from; Brave's own crawl has grown and refreshed it since. Without that acquisition Brave would almost certainly have ended up as another reseller.

The Web Discovery Project. This is an opt-in component of the Brave browser that contributes anonymised records of pages users visit, which help Brave work out which URLs are worth crawling and how they should rank. Brave's design claim is that submissions are stripped of identifying material and are not linkable back to the user who sent them. It is an unusual approach — most independent engines rely on the crawl and the link graph alone — and it is a real part of how a company with a startup's crawl budget keeps a reasonably fresh index. It is not a substitute for the crawler. The crawler does the indexing; the Web Discovery Project informs it.

The combination explains both the strength and the weakness. Brave has current coverage of the parts of the web people actually visit, and thinner coverage of the long tail — obscure pages, very fresh content, material outside major Western languages — because an index started in 2021 has had far less time and money spent on it than one started in 1998.

Goggles: the feature that proves the index is really Brave's

Goggles, launched in June 2022, let anyone write a re-ranking filter over Brave's index and publish it as a URL. A Goggle is a small rule set — boost these domains, downrank those, exclude that category entirely — with a limit of 10,000 instructions. Apply one and the same query returns a different ordering: an academic Goggle can push journals and university pages up, a Goggle for a hobbyist community can strip out content farms.

The reason it belongs in a discussion of index ownership rather than a feature list is that it cannot be built on top of somebody else's search API. Re-ranking requires access to the ranked candidate set and the signals behind it, not just the ten results an API hands back. Goggles working at all is practical evidence that the independence claim is structural rather than rhetorical. A related feature, Rerank — per-result upvotes and downvotes — shipped in January 2025.

Who runs it, and how it makes money

Brave Search is operated by Brave Software, Inc., a venture-backed independent company headquartered in San Francisco and co-founded by Brendan Eich and Brian Bondy. No change of ownership has been recorded as of August 2026.

There are three structural revenue lines. Described structurally rather than by price, because prices move:

Search advertising. Brave sells advertising on the results page. Its stated design is contextual — ads matched to the words in the query rather than to a behavioural profile built from a user's history.
A subscription that removes the ads. Brave Search Premium is a paid tier whose product is the absence of advertising rather than extra features.
Metered API access. The Brave Search API, launched in May 2023, sells access to the index on tiered, usage-metered terms.

The API has turned out to matter more than a side business usually does. Independent web-scale indexes are scarce, and AI companies and agent frameworks that need live web grounding have very few options that are neither Google nor Microsoft. Brave markets the API on precisely that basis. Search is also cross-subsidised by, and strategically useful to, the rest of Brave's business — the browser, its advertising system, and its Leo assistant.

Privacy: what is structural, and what is a claim

Brave states that Brave Search is anonymous by default: no user profiling, no IP-based tracking of query history, no cross-site identifiers, and no logging that ties a query to a person. Search works without an account.

One part of that is structural and one part is not, and the distinction separates Brave from proxy-based privacy engines.

Structural: because Brave owns the index, a query never has to be forwarded to a third party to be answered. In a proxy engine, the query does reach Google or Bing — stripped of the user's identity, but it arrives. With Brave there is no Google or Bing at the receiving end of a web search at all. That is a property of the architecture and does not depend on trusting a policy document.

A claim: everything about what Brave does with the queries it receives. Brave publishes methodology; a user cannot verify at runtime that queries are not being retained. No independent technical audit specifically of Brave Search's query-logging practice was located in the research for this page. Published methodology is meaningfully better than nothing and meaningfully worse than third-party verification.

Three further caveats belong on the record. The Web Discovery Project is browsing telemetry, and although it is opt-in and designed for unlinkability, that unlinkability is a design argument rather than something a user can check. Brave Software is a United States company, and therefore subject to US legal process — subpoenas, National Security Letters — in a way EU-domiciled engines are not; Brave's counter-argument is that it deliberately holds little that could be compelled, and no public record of such an order against Brave Search was found in this research. And the Brave browser has had privacy and business controversies of its own, described below, which are routinely attributed to Brave Search. They are separate products.

AI answers: Summarizer and Leo

Summarizer launched in March 2023 and puts an AI-generated answer above the ranked results, synthesised from those results and shown with citations back to the pages it drew on. Wikipedia reports coverage of roughly 30–40% of searches as of 2025; that is a secondary-sourced figure rather than a published primary statistic, and should be read as an approximation.

Leo is Brave's in-browser assistant, reachable from the browser sidebar, and able to use Brave Search results as grounding. Brave has used a mix of its own fine-tuned models and third-party open-weight models for Leo; the exact current lineup as of August 2026 could not be confirmed against a single authoritative page, and Brave's published model list changes often.

The detail that distinguishes Brave from most of its peers is mundane and important: the AI summary can be switched off in search settings. A growing number of engines put a generated answer above the results with no way to remove it. If a reader wants an engine that shows links first and only links, this setting is a real reason to look at Brave.

How it got here

The sequence is short and unusually well documented, because Brave published each step.

March 2021 — Brave acquires Tailcat from Cliqz, and with it a working independent index.
22 June 2021 — public beta launches, with about 87% of results from Brave's own index.
October 2021 — Brave Search becomes the default engine in the Brave browser in the US, UK, Canada, Germany, France and other regions, displacing Google.
June 2022 — Goggles launch; the beta tag comes off Brave Search. Own-index share is around 93%.
March 2023 — Summarizer ships.
27 April 2023 — Brave announces the removal of the last Bing API calls from web results, described by the company as 100% independence. TechCrunch and the trade press confirmed it contemporaneously.
May 2023 — the Brave Search API launches.
August 2023 — Brave's own image and video index launches, removing the last third-party dependency in those verticals.
2024 — privacy-preserving search ads and the ad-free subscription tier are formalised.
January 2025 — Rerank ships alongside Goggles.

Separately, and on the browser side rather than the search side, Brave has drawn criticism over its BAT token and creator-tipping model between 2018 and 2020, when creators who had never signed up appeared as tippable, and over a 2020 incident in which the browser auto-inserted Brave affiliate codes into cryptocurrency exchange URLs typed by users, for which Brendan Eich apologised. Those are legitimate context for how much a reader trusts the company. They are not findings about the search engine, and merging the two produces a confused picture of both.

Who it suits, and who it does not

It suits a reader who wants an everyday engine that is not downstream of Google or Microsoft, who wants the option of no advertising, and who wants control over whether an AI answer appears at all. It suits developers who need web results for grounding and cannot or will not build on Google's or Microsoft's APIs — a narrow requirement, but one with few other answers. And it suits anyone who wants to shape their own ranking, which Goggles makes possible in a way no reseller can match.

It is a poor choice for work that depends on exhaustive recall of the long tail. An index built since 2021 is smaller than Google's, and the gap shows on obscure queries, on very fresh pages, and outside major Western languages. Local and map results are weaker. Researchers, journalists chasing an unindexed document, and anyone working primarily in a language outside Brave's coverage will still find themselves falling back to Google, and should expect to.

A note on measuring it: there is no reliable market-share figure for Brave Search. StatCounter, the source most commonly cited for search-engine share, does not publish a separate Brave line — it falls inside "Other" — so any percentage attributed to StatCounter for Brave should be treated as unsourced. What exists are Brave's own self-reported query counts, which are not independently audited: around 2.5 billion queries cumulatively across the beta period, reported in June 2022, and around 1.6 billion queries per month, reported around September 2025. For scale, Google is routinely described as handling hundreds of billions of queries a month. Brave is a real independent index and a small one.

Frequently asked questions

Does Brave Search use Bing?

No, not since 27 April 2023, when Brave announced it had removed the last third-party API calls from its web results. It does use its own crawler and index. The confusion is well founded historically: at launch in June 2021 roughly 13% of results came from Bing, falling to about 7% by mid-2022 before being removed entirely. Brave's own image and video index followed in August 2023.

Does Brave Search have its own index?

Yes. Brave crawls the web with its own crawler, Bravebot, and ranks results from an index it owns and operates. The index descends from Tailcat, the search stack Brave acquired from Cliqz in March 2021, and has been extended by Brave's own crawling since. Only a handful of organisations worldwide run a general web index; Brave is one of them, alongside Google, Bing, Yandex, Baidu and Mojeek.

Who owns Brave Search?

Brave Software, Inc., an independent venture-backed company headquartered in San Francisco and co-founded by Brendan Eich and Brian Bondy. It is not owned by Google, Microsoft or any other search company. The index technology originated at Cliqz, a subsidiary of the German publisher Hubert Burda Media, and was acquired outright by Brave in March 2021. No ownership change has been recorded as of August 2026.

Is Brave Search actually private?

Its strongest privacy property is architectural: because Brave owns the index, your query is never forwarded to Google or Bing to be answered. Brave also states it does not profile users, log queries against identities, or require an account. Those statements rest on published methodology rather than an independent audit, and Brave is a US company subject to US legal process, so the claim is credible but not externally verified.

Can I block Bravebot from crawling my site?

Yes. Brave publishes a declared user-agent token, Bravebot/1.0, which can be named in robots.txt like any other crawler, and it documents the crawler publicly. Earlier in the engine's life Brave's documentation stated that the crawler did not advertise a differentiated user-agent, on the argument that sites permitting only Googlebot would otherwise shut it out. Older write-ups describing that behaviour are out of date.

What are Brave Search Goggles?

Goggles are user-written re-ranking filters over Brave's index. A Goggle is a rule set of up to 10,000 instructions that boosts, downranks or excludes sets of sites, published as a URL anyone can apply. They launched in June 2022. They matter beyond their usefulness because re-ranking requires access to the underlying ranked candidates, which an engine reselling somebody else's API does not have.

Can I turn off Brave's AI answers?

Yes. The Summarizer, which places an AI-generated answer with citations above the results, can be disabled in Brave's search settings. It launched in March 2023, and Wikipedia reports it covering roughly 30–40% of searches as of 2025 — a secondary-sourced approximation. The ability to switch it off is a genuine difference from engines that place a generated answer above results with no opt-out.

What is Brave Search's market share?

There is no reliable published figure. StatCounter, the usual source for search-engine share, does not break Brave out as a separate line; it is folded into "Other", so any percentage attributed to StatCounter for Brave Search is unsourced. Brave's own self-reported numbers are around 2.5 billion queries cumulatively during the beta, reported June 2022, and around 1.6 billion queries per month, reported around September 2025.