Brave Search

Brave Search is *one of the only Western search engines running on its own independent index* — not a relabelled Bing or Google feed — offered without an account, without profiling ads, and with a Tor onion for good measure.

Background. Launched in 2021, Brave Search was bootstrapped from the Tailcat index that Brave acquired (the technology team behind Cliqz's privacy search). That origin matters: most "privacy search engines" are thin front-ends that proxy Bing or Google results, which means their independence ends the moment the upstream changes terms. Brave instead crawls and ranks from its *own index*, publishing an "independence" metric for how many results come from its own crawl versus fallbacks. That genuine index, plus a no-account/no-tracking posture, is why it earns an A in /search.

What you trust. The core trust property is *self-sufficiency*: because Brave runs its own index, it isn't merely re-skinning a surveillance-advertising company's results, and it can make and keep its own privacy commitments. Queries work *without an account* — there's nothing to log in to, so there's no profile to build. The free tier carries *no personal-data ad targeting*. There's a published API tier that downstream tools and other privacy front-ends consume, which both funds the index and demonstrates the results are a real product rather than a loss-leader scrape. And the *Tor onion* lets you query without revealing your network origin at all.

Operational specs. Free, instant, no signup. Distinctive features include *Goggles* (open, shareable re-ranking rule-sets that let you bias or filter results by your own criteria — a rare grant of ranking control to the user), an optional AI answer/summariser, and image/video/news verticals. The independence percentage is shown so you can see when a query fell back to a partner index. A paid API tier exists for developers and for the downstream apps that resell privacy search. Access via clearnet or the `search.brave…onion` address; no JavaScript-gated paywall for basic querying.

Philosophy. Search is the most sensitive query stream most people generate — it's a log of intentions, fears, and plans — and the dominant engines monetise exactly that by profiling. Brave Search's premise is that you can have a mainstream-quality engine that *doesn't* run on behavioural advertising, funded instead by an API business and an optional ad model that isn't built on personal-data targeting. Owning the index is the structural prerequisite: you can't credibly promise privacy on results you're renting from a profiling giant. Goggles extends the philosophy to ranking autonomy — letting users, not just the engine, decide what "relevant" means.

Grade rationale. A in /search. The grade reflects a genuinely independent index (the rarest and hardest property in this category), no-account querying, absence of personal-data ad targeting, user-controllable ranking via Goggles, a funding model that isn't surveillance, and a Tor onion. It sits at the top of the category with the privacy-proxy options (Leta, SearXNG), differentiated by being a primary index rather than a proxy.

Useful when. Use Brave Search as a daily default when you want mainstream result quality without feeding a profiling engine — it's set-and-forget as a browser default. Reach for Goggles when you want to filter SEO spam, surface only independent/non-commercial sources, or apply a community ranking lens. Use the onion when you're searching over Tor and don't want to expose your origin. It's also a sensible upstream for self-hosted front-ends that want a non-Google/Bing source.

Caveats. "Independent" is a spectrum, not an absolute — for some long-tail or regional queries Brave still falls back to partner results, and the published independence metric is exactly how you should check that per session rather than assuming 100%. Brave the company also operates a separate ads/rewards ecosystem and a crypto token; none of that touches the no-account search path, but users who dislike that broader model should know it exists. The optional AI summariser carries the usual generative-model caveats (it can be confidently wrong — verify from the underlying links). And result quality, while strong, is still maturing against two decades of incumbent tuning on rare queries. These are scope notes, not trust problems, and the A stands on the strength of owning a real index with a non-surveillance funding model.