Dark Web Search Engines Compared: What Actually Works

Why searching onion sites is hard

A search engine on the regular web starts from known pages, follows links, and reads DNS and certificate data to know who it is talking to. On the Tor network none of that infrastructure exists. Onion addresses are not registered anywhere, there is no zone file to walk, and many services are deliberately unlinked.

Crawlers for onion sites work by collecting addresses from directories, forums, other engines and pages they have already indexed, then fetching each through Tor. This is slow, since every request traverses six relays, and it produces a corpus full of duplicates, offline servers and abandoned pages. Studies of onion services consistently find that a large share of addresses are unreachable at any given moment.

The result is that even a good onion search engine is more like a partial phone book than a live index. Expect stale entries and check dates when the engine shows them.

Ahmia: the filtered option

Ahmia is a Finnish open-source project that indexes onion services and publishes the code behind it. Its defining policy is a blocklist: it removes child abuse material and accepts reports of it, which makes it the closest thing to a search engine you can use without worrying about what appears on the results page.

Ahmia has both a surface-web front end and an onion address, which means you can search from a normal browser and then open results in Tor Browser. It also publishes statistics about the services it sees, useful for understanding the network.

Trade-offs: the index is smaller than the "index everything" engines, and like all of them it cannot fully prevent phishing clones from appearing. For most readers, Ahmia is the sensible default.

Ahmia also publishes a list of onion services it has blocklisted for abuse, which shows that the filtering is active rather than nominal.

Torch, Haystak and the large indexes

Torch is one of the oldest onion search engines and advertises a very large index. It does not filter content the way Ahmia does, which means results can include material that is illegal to view in most countries. It is ad-supported, and the ads have historically promoted markets and services you should not touch.

Haystak markets itself on index size and offers a paid tier with advanced features. Like Torch, it takes a broad approach to what it indexes.

These engines are useful for researchers who need coverage and are prepared for what they may see. For a general reader, the unfiltered approach means that a typo or a curious click can put content on screen that you did not want. If you use them, keep Tor Browser's security level high and search for specific lawful services by name rather than browsing.

DuckDuckGo and the mainstream onion mirrors

DuckDuckGo runs an onion mirror of its search engine and is the default search provider in Tor Browser. It is important to understand what it does and does not do: it searches the regular web anonymously. It does not index onion sites. Searching it for an onion service will return surface-web articles about that service, not the service itself.

That still makes it useful. Many organizations publish their onion addresses on their surface-web sites, so a DuckDuckGo search through Tor for "organization name onion" often leads you to the official page where the real address is listed. It is a safer discovery path than any onion index because the address comes from the organization itself.

The same applies to the onion mirrors of other mainstream services: they exist to protect users of the regular web, not to explore the hidden one.

Not Evil, OnionLand, Tordex and the churn problem

Lists of "best dark web search engines" published a few years ago included names like Not Evil, Candle, Grams and OnionLand. Some of those have gone offline, changed addresses or been replaced by clones using the old name. This churn is normal on the network and it creates a specific hazard: an old article's onion address for a search engine may now point at a phishing page that shows manipulated results.

Before trusting any search engine address, confirm it through a current, maintained source. Ahmia's surface-web site lists its own onion address. The Tor Project's documentation and a small number of maintained resource pages track others. If you cannot find a current confirmation, assume the address is stale.

For your purposes this means the "best" engine is partly the one whose address you can verify today, not the one with the most impressive index.

A method for checking results before you click

Results from any onion search engine deserve a short verification routine.

1. Compare the full onion address against a trusted source, not just the first few characters; phishing clones often match the readable prefix.
2. Prefer results that link to a service's own surface-web site or to a maintained directory that dates its checks.
3. Treat any result that mentions accounts, payments, cards, "cheap" or "guaranteed" as a scam or worse, and skip it.
4. Open unfamiliar results at the Safest security level so that scripts do not run.
5. If the page you reach asks you to log in, pay or download before showing content, leave.

This routine takes seconds and removes most of the risk that comes from the low quality of onion indexes.

If a result fails any step, do not investigate further; close the tab and move on. Curiosity is the mechanism that phishing and abuse listings rely on.

Context from the projects and researchers

• Ahmia's published policy describes its filtering of abuse material and its open-source indexing code, which is why it can be recommended to general readers where unfiltered engines cannot.
• Tor Project documentation confirms that DuckDuckGo's onion mirror is the default search in Tor Browser and searches the regular web, which clarifies the common confusion about what it indexes.
• Academic measurements of onion services report that a large proportion of addresses are offline at any given time and that many active sites are duplicates or clones. For you this explains the dead links and the need to verify.
• Security-vendor reports have documented phishing operations that cloned popular onion services and seeded the clones into directories and search results, which is the concrete reason to compare full addresses.

The engines themselves rarely publish accuracy figures, so independent measurement is the only check on their claims.

Where to start

Use Ahmia as your default onion search engine, opened from its surface-web site so you know the address is current. Use DuckDuckGo inside Tor Browser to find an organization's official page and take onion addresses from there. Use Torch or Haystak only if you are a researcher who needs coverage and can handle unfiltered results.

Then apply the five-step check above to every result. The engines are imperfect; your verification habit is what makes them usable.

Expect to spend more time verifying than searching at first. That ratio improves as you build a short personal list of confirmed addresses, at which point the search engines become a tool for occasional discovery rather than a daily dependency. That is the healthiest relationship to have with them.

One last point about expectations: an onion search engine returning a few hundred results for a query is normal, and most of them will be duplicates or offline. Quality, not quantity, is the measure here.

Frequently asked questions