Dark Web Search Engines: What Can Be Indexed

By Bill Hartzer · Updated 27 October 2026

What a dark web search engine is

A dark web search engine is a search engine that indexes Tor onion services — websites reachable only through the Tor anonymity network — because Google, Bing and other ordinary engines cannot reach them. The best-documented example is Ahmia, an open-source engine associated with the Tor Project that filters abusive content out of its index; a handful of unmoderated alternatives are reported to operate as well.

Two terms are routinely confused.

The deep web is any web content ordinary engines do not index: pages behind logins, paywalls and forms, or marked not to be indexed. Most of it is mundane — webmail, bank accounts, company intranets.
The dark web is content reachable only through anonymity networks that require special software. In practice, and on this page, that means onion services, whose addresses end in .onion. Other overlay networks, such as I2P and Hyphanet, are far smaller.

How onion services work

An onion service (formerly a "hidden service") is a server reachable only from inside the Tor network. Its address is registered nowhere: the Tor Project explains that the address "looks weird and random because it's the identity public key of the Onion Service." Knowing the address is the same as knowing the key that proves the server's identity.

A visitor's Tor client never connects to the server directly. The service publishes a few Tor relays as introduction points, the client picks another relay as a rendezvous point, and the two meet there through chains of relays. Traffic is encrypted end to end, and neither side learns the other's IP address. The design that protects operators and users is the same design that makes onion services hard to catalog.

Reaching an onion service requires a Tor client, normally the Tor Browser. Most Tor use is not onion browsing at all but anonymous visits to the ordinary web.

Why Google cannot index onion services

Four structural reasons keep onion services out of mainstream indexes.

Reachability. Googlebot, Bingbot and other crawlers fetch pages over the ordinary internet. An onion address does not resolve in public DNS; it can be reached only through a Tor client, so a crawler would have to run Tor.
No list to start from. Since version 3 onion services, the descriptor each service uploads to Tor's directory is keyed by a blinded public key, so even the relays holding the directory cannot read which address it belongs to. The older version 2 directory, the Tor Project said, gave relays "too much power to enumerate or even block v2 services." Now onion sites can be discovered only through links or operator submissions.
Sparse links and constant churn. In the Dizzy crawl of 2021–2022, only 52.2% of crawled onion domains were reachable on a typical day, and 75.5% were visually and textually similar to other domains — mirrors, clones and templates. Link-based ranking has little to work with.
Policy and liability. A mainstream engine has little reason to index a small, unstable corpus with a documented share of illegal content — an inference from the evidence below, not a stated Google or Microsoft policy.

The version 2 retirement

The Tor Project announced the deprecation of v2 in June 2020, citing weak RSA-1024 keys, truncated SHA-1 addresses and directory enumeration. v2 was removed from the code base on 15 July 2021 and disabled in all supported stable releases on 15 October 2021; Tor Metrics shows v2 addresses falling from about 171,000 in January 2021 to about 67,000 that December. The retirement closed the last way of harvesting addresses from the network itself.

What Ahmia does, and what it filters

Ahmia is the onion search engine this page can describe from primary sources. It runs its own crawler, built on the Scrapy framework, an Elasticsearch index and a Django website, all open source. Its founder and project leader, Finnish security researcher Juha Nurmi, developed it with Tor Project mentors through Google Summer of Code in 2014; its exact first launch date is not documented. It is reachable on the ordinary web at ahmia.fi as well as through Tor, and was operating as of 27 October 2026.

Ahmia discovers addresses by crawling the ordinary web and onion space for links and by accepting operator submissions. It publishes no current index count; the only dated figure is about 5,000 sites in July 2015.

What distinguishes Ahmia is filtering. It maintains a blocklist of child sexual abuse material (CSAM), publishes a blacklist of banned services and accepts reports of abusive content. A 2024 paper in Scientific Reports, co-authored by Nurmi, analyzed 110,133,715 search sessions on Ahmia and found that 11.1% sought CSAM; Ahmia redirects such searches toward help resources. The paper states that Ahmia then decided, in November 2023, to filter all sexual and suspicious searches (Wikipedia dates the change to October 2023). The figure describes Ahmia's users, not all Tor users.

Ahmia carries no advertising and charges nothing; the Tor Project's 2016 profile called it a "low-budget search engine."

Other onion search engines and a common misconception

Other services are reported to operate, but none could be verified from primary sources.

Torch is described as the oldest onion search engine, running since roughly the early 2010s. An August 2026 security-firm review reported it still operating, unmoderated, with rotating addresses and phishing clones. Its claim of indexing "over a billion pages" is unverified marketing; the same review said it "collapses against Tor Metrics."
Haystak was reported in August 2026 to offer a paid tier with a larger index. Its index-size claims, founding date and operator are unverified.
Commercial threat-intelligence crawlers run by security vendors index onion forums and markets for corporate clients. They are monitoring services, not public search engines.

The misconception worth correcting concerns DuckDuckGo, the default search engine in the Tor Browser. DuckDuckGo's own onion address is often described as a dark web search engine. It is not: it is a Tor-accessible front end to ordinary DuckDuckGo search and does not index onion sites.

How big onion space actually is

Claims about the size of the dark web vary by orders of magnitude, and the measured figures are far smaller than the marketing ones.

Measurements of onion space
Source	What it measured	Finding
Tor Metrics (Tor Project)	Unique v3 onion addresses per day, 26 Sep–25 Oct 2026	About 656,000 on average
Hoeller, Roland, Mayrhofer (FOCI '21)	v3 onion services, March–April 2021	Between 600,000 and 700,000; about 86% of blinded keys never requested from the authors' relays
Boshmaf et al., Dizzy (ARES 2023)	Crawlable onion websites, to 31 Jan 2022	39,536 domains and 63,267,542 pages; 52.2% up on a typical day; 75.5% near-duplicates
Tor Metrics (Tor Project)	Tor users per day, 26 Sep–25 Oct 2026	About 3.28 million direct, plus about 222,000 via bridges

The gap between the first and third rows is the point: Tor counts hundreds of thousands of onion addresses a day, but crawlers find only tens of thousands of onion websites, half offline on a given day and three-quarters duplicates. The Hoeller study's finding that most services are never requested suggests many addresses belong to machine endpoints, such as messaging or botnet services, rather than websites. No measurement found supports claims of "billions of pages" on the dark web.

Each figure carries a caution. The address count is an estimate extrapolated from directory-relay statistics; its 30-day average and annual means (about 749,000 in 2022, about 882,000 in 2025) were computed for this page from the daily Tor Metrics v3 address series. The user estimate swings with bot and client anomalies, from annual means of about 2.1 million in 2020 to 4.7 million in 2025. The Hoeller estimate rests on about 0.8% of the directory ring, and Dizzy saw only sites its seed links could reach; half the domains it classified were marketplaces.

Legal and security risks

Search engines in this space cannot vouch for what they list.

Illegal content. Accessing, possessing or distributing CSAM is a serious crime nearly everywhere. The 2024 Scientific Reports study found such material prevalent on onion services — 22.1% of 1,000 domains sampled in January 2023 — which is why responsible engines filter it.
Fraud. Scams are endemic. In July 2015 Ahmia published a list of hundreds of cloned onion sites used to steal payments, for example by swapping Bitcoin addresses, and its live site still warns of a fake clone of Ahmia itself. Phishing copies of Torch have also been reported.
Malware. An unmoderated index can point to any page, including pages serving malicious downloads. No engine in this space screens results for security, and anonymity hides who runs a site.
Legality of Tor itself. Tor is used by journalists, activists and ordinary users seeking privacy. Whether using it is lawful depends on the country; this page assesses no jurisdiction.

In short: the dark web is far smaller than its reputation, ordinary engines cannot see it by design, and the one documented engine that indexes it works hard at deciding what not to show.

Frequently asked questions

Is there a dark web search engine like Google?

No. No engine indexes onion space the way Google indexes the web, because onion addresses cannot be enumerated. The best-documented dark web search engine is Ahmia, an open-source engine associated with the Tor Project, which filters child sexual abuse material and, since late 2023, sexual and suspicious searches. Torch and Haystak are reported to operate unmoderated, with unverified index claims.

Why can't Google search the dark web?

Onion services are reachable only through the Tor network, and their addresses do not resolve in public DNS, so Google's crawler cannot fetch them over the ordinary internet. Since Tor's version 3 design, addresses cannot be read from Tor's directory either, leaving links and submissions as the only discovery method. A 2021–2022 crawl found only 52.2% of onion sites reachable on a typical day.

What is Ahmia?

Ahmia is a search engine for Tor onion services, founded and led by Finnish security researcher Juha Nurmi and developed with Tor Project mentors in 2014. It runs an open-source crawler, Elasticsearch index and website, is reachable at ahmia.fi as well as through Tor, and maintains a blocklist of child sexual abuse material. Its search logs underpin a 2024 Scientific Reports study of 110,133,715 search sessions.

Is DuckDuckGo a dark web search engine?

No. DuckDuckGo is the default search engine in the Tor Browser and runs its own onion address, but that address is a Tor-accessible front end to DuckDuckGo's ordinary web search. It does not index onion sites. Searching DuckDuckGo inside Tor returns ordinary web results while hiding the searcher's IP address from DuckDuckGo.

How big is the dark web?

Much smaller than commonly claimed. Tor Metrics estimated about 656,000 unique onion addresses per day in late October 2026, but most host no public website. The Dizzy academic crawl found 39,536 crawlable onion domains to January 2022, only 52.2% online on a typical day and 75.5% near-duplicates. No measurement supports claims of billions of dark web pages.

Dark web search engines and what they can index