Search engine comparison: index, ranking, privacy and money

Comparing engines on the four things that actually differ

Most search engine comparisons rank engines by result quality, which is subjective and changes weekly. This one compares four structural facts instead — stable, checkable, and between them the explanation for almost every difference a user notices.

• Index source — does the engine crawl and rank its own copy of the web, blend its own index with a partner's feed, or have no index at all? This is the least-known and most useful fact about any engine, because it decides whether an "alternative" is genuinely an alternative.
• Ranking philosophy — what the engine optimises for. Engagement authority, unpersonalised keyword matching, a bias toward non-commercial text and generated answers over links are four different products.
• Privacy stance — what is collected, in what jurisdiction, and crucially whether the query is forwarded to a third party. An engine with no index of its own necessarily sends your query somewhere else.
• Monetisation — advertising, subscription, API metering, donation or a mix. Whoever pays is who the ranking must satisfy.

Engines that run their own index

A working web-scale index is rare — perhaps a dozen organisations operate one at meaningful scale. Owning one is the only thing that makes an engine independent of the two companies supplying almost everyone else.

Google

• Index source: Own. Googlebot crawls the open web and Google ranks its own index; the crawler is publicly documented.
• Ranking philosophy: Link and engagement authority at enormous scale, personalised by account, location and device, with AI summaries above the ranked links.
• Privacy stance: The most comprehensive behavioural profile of any search engine, by design — signed-in activity, location, device identifiers and cross-product signals from Android, Chrome, YouTube and the ad network. Controls are real; the default is collection.
• Monetisation: Advertising, overwhelmingly — auction-priced search ads. Google has also paid very large sums for default placement, central to the US antitrust case. Free to users.

Microsoft Bing

• Index source: Own. Bing crawls and ranks its own index and syndicates it to Yahoo!, DuckDuckGo and others.
• Ranking philosophy: Conventional relevance-and-authority ranking with heavy Copilot integration; Bing shipped an LLM into mainstream search before Google did.
• Privacy stance: Ad-funded engine inside a very large consumer and enterprise data estate; signed-in search is tied to a Microsoft account. Not a privacy engine, and it operates a filtered version in mainland China.
• Monetisation: Microsoft Advertising, plus syndication revenue share and Azure metering for its grounding product. Free to users.

Brave Search

• Index source: Own. Bravebot crawls and Brave ranks its own index; the last Bing API calls left web results on 27 April 2023, with own image and video results that August.
• Ranking philosophy: Ranked links with an optional AI summary, plus Goggles — user-written, shareable re-ranking rules, which are only possible on an owned index.
• Privacy stance: Anonymous by default, no account, no profiling claimed. Because it owns the index, the query is not forwarded to Google or Microsoft. US company.
• Monetisation: Contextual search ads, a paid ad-free tier, and the Brave Search API sold to developers and AI companies.

Mojeek

• Index source: Own. A British independent crawling the web itself since 2004; its about page put the index at 9 billion pages in 2025.
• Ranking philosophy: Deliberately old-fashioned keyword search: no personalisation, no filter bubble, no generative answer engine — an editorial choice, not a missing feature.
• Privacy stance: No tracking, no profiling, and no third party in the pipeline to receive the query. UK jurisdiction.
• Monetisation: Its own search ads, a metered Web Search API, a Site Search API and supporter revenue. No Big Tech backer.

Marginalia Search

• Index source: Own. A single-operator Swedish index, AGPL open source, running on modest hardware.
• Ranking philosophy: The most distinctive philosophy here: it ranks toward text-heavy, non-commercial, hand-made pages and away from optimised commercial content. It surfaces what mainstream ranking buries, and is explicitly a complement, not a replacement.
• Privacy stance: No accounts, ads, ad-tech or tracking scripts. The position is structural rather than promised: no ad business means no reason to profile, and the query never leaves the service.
• Monetisation: Not a business — donations, grants and commercial API deals, against stated running costs of roughly a couple of hundred dollars a month.

Perplexity

• Index source: Own, now. Its engineering write-up of 25 September 2025 describes an exabyte-scale index tracking over 200 billion unique URLs. It was substantially a Bing API reseller in 2022–2024 and has never dated the change.
• Ranking philosophy: An answer engine, not a results page: prose with numbered inline citations, where links are evidence rather than the product. Generation routes to several vendors' frontier models; retrieval stays Perplexity's.
• Privacy stance: Consumer AI product with accounts and history. Publishers have repeatedly disputed its crawling and citation behaviour, including in litigation.
• Monetisation: Subscription-first: it introduced AI-integrated advertising in late 2024 and discontinued it in February 2026, a notable reversal. Also metered APIs and commerce features.

Hybrid engines: own crawler plus somebody else's feed

The hybrids are where the confusion lives: marketing emphasises the owned part; the ranked links usually come from the partner part.

DuckDuckGo

• Index source: Hybrid, heavily weighted to a partner. It runs DuckDuckBot and its own indexes, but in its own words the traditional links and images are "largely sourced from Bing".
• Ranking philosophy: Bing's relevance ordering with DuckDuckGo's Instant Answers and !bang shortcuts layered on. No personalisation, no result history.
• Privacy stance: The strongest promise in mainstream search: no stored history, no IP or unique identifier saved alongside queries, settings in cookies not accounts, referrers stripped on click-through. The standing criticism is that it is a US company whose results come from another US company.
• Monetisation: Query-matched ads served mostly through Microsoft Advertising, affiliate commissions, and a paid privacy-tools subscription. No behavioural targeting.

Kagi

• Index source: Hybrid. Kagibot feeds Kagi's own Teclis web and TinyGem news indexes, but most of a results page comes from anonymised API calls into third-party commercial indexes. Kagi's docs named Google, Yandex, Mojeek and Brave until May 2024, when the wording was generalised; only Brave is still named.
• Ranking philosophy: User control as the product — per-domain raise, lower, pin and block rules that persist across searches, plus Lenses. Ranking is edited rather than received.
• Privacy stance: No advertising means no commercial reason to profile — the cleanest alignment argument here. The caveat: upstream API calls still leave, and their anonymisation is Kagi's own description, with no published independent audit.
• Monetisation: Subscription only, plus metered developer APIs. There is no free ad-supported tier, which is the point.

Ecosia

• Index source: Hybrid. Ecosia's help centre states results come from Bing, Google and EUSP — the joint venture it co-owns 50/50 with the French engine Qwant, whose index Staan has served a growing share of both engines' queries since August 2025. The owned share is still a minority.
• Ranking philosophy: Partner relevance ordering, with Ecosia's presentation layer and green-energy markers on results.
• Privacy stance: EU jurisdiction, EU-hosted servers, no profile-based advertising. But queries served by the Bing or Google feed reach a US company, so the position is only as strong as the anonymisation before hand-off.
• Monetisation: Advertising, with profits directed into tree planting. Unusually, since October 2018 the founder's shares sit in a steward-ownership structure that makes the company unsellable and bars dividends.

Engines with no index of their own

"Resold results" is a neutral description, not a criticism — Startpage resells Google deliberately and says so; that is the product. What matters is knowing, because a resold engine cannot rank differently from its supplier or outlive it.

Yahoo! Search

• Index source: None. Yahoo!'s web results come from Microsoft Bing under a long-running syndication deal; Yahoo! ran its own crawler until the late 2000s and stopped.
• Ranking philosophy: Bing's ranking inside Yahoo!'s portal. Yahoo! Japan is a separate case — it licenses Google's results.
• Privacy stance: A mainstream ad-funded portal with account-linked services; not a privacy product.
• Monetisation: Advertising, including a share of Microsoft search ad revenue for traffic it sends.

Startpage

• Index source: None, by design. Startpage is a privacy proxy: it submits the query anonymously on the user's behalf and returns the results. Its own documentation names those third parties as Google and Bing — not Google alone, as is widely assumed.
• Ranking philosophy: Whatever Google and Bing rank, stripped of personalisation. Anonymous View opens a result without the destination seeing you.
• Privacy stance: Netherlands operation, EU jurisdiction, GDPR — the core pitch. The complication is ownership: majority-owned since late 2018 by a subsidiary of System1, a US-listed advertising company, a fact undisclosed until October 2019.
• Monetisation: Advertising on the results page.

SearXNG

• Index source: None, and no company either. SearXNG is free metasearch software that forwards a query to other engines — its docs cite up to 269 services — and merges the results. The upstream mix is set by whoever runs the instance.
• Ranking philosophy: Merged upstream rankings, configurable per instance. Nothing is ranked by SearXNG itself.
• Privacy stance: The most glossed-over fact about it: jurisdiction, logging and trustworthiness depend entirely on the instance — a public one is a server run by an unidentified volunteer under some country's law. Self-hosting is the only configuration you control.
• Monetisation: None. AGPL community software with no owner and no revenue. The predecessor searx was archived in September 2023 — anyone still running plain searx is running abandonware.

What the comparison actually changes

Three consequences follow from reading engines this way.

1. Switching engines does not necessarily change your results. Moving between Bing, Yahoo! and DuckDuckGo's web links means changing front ends over one index. Genuinely different results require an engine with its own index, and that shortlist is short.
2. Resold engines carry supplier risk. Microsoft retired its public Bing Search APIs outright on 11 August 2025, and every product reselling Bing that way lost the route overnight.
3. Privacy claims are stronger when structural. There is a real difference between an engine that promises not to log your query and one that never sends it anywhere. Mojeek, Marginalia and Brave Search are in the second group because they own their indexes; every proxy and hybrid is in the first.

None of this makes owned-index engines better at answering questions — Google's results beat Mojeek's on most queries, and it is not close. It makes them different in kind, which is what matters if the point of an alternative is not getting everyone else's answers.

When this comparison was last checked

Every index-source claim above was checked against the engine's own documentation or a primary source on 19 August 2026. These positions move — Brave was a hybrid until April 2023, Perplexity was substantially a Bing reseller until roughly 2025, Ecosia's own-index share has grown since August 2025 — so this is a snapshot.

Pricing is described structurally — subscription, advertising, API metering, donation — with no current prices stated, because they change faster than anything else here.

Where market share is mentioned, the figures are StatCounter Global Stats for July 2026.

Frequently asked questions

Sources

• duckduckgo.com/duckduckgo-help-pages/results/sources
• brave.com/blog/search-independence/
• help.kagi.com/kagi/search-details/search-sources.html
• support.ecosia.org/article/579-search-results-providers
• support.startpage.com/hc/en-us/articles/4522435533844
• mojeek.com/about/
• about.marginalia-search.com/
• gs.statcounter.com/search-engine-market-share