Working Onion Links: Why Onion Services Go Down and What a Failure Actually Means
Why working onion links are so much rarer than working web links
Ordinary websites are up almost all the time because they run on professionally hosted infrastructure with redundancy, monitoring and people paid to keep them running. Most onion services are not like that, and the reasons working onion links are scarce follow directly from how the network is used. Many onion services run on a single machine. Hobbyist services, personal pages and small forums frequently live on a home computer, a Raspberry Pi or a single rented server, and when that machine reboots, loses power or is taken offline by its owner, the service disappears until it returns. Onion services have no obligation to persist. An operator who loses interest simply stops running the software, and the address dies with no notice, no parked page and no redirect. The network itself adds latency and occasional failures. Reaching an onion service involves a rendezvous through introduction points and a six-hop path, and a busy or poorly connected service may time out even when it is technically running. And a large fraction of circulating addresses were never good. Lists copy from lists, v2 addresses persist years after the protocol was retired, and typos propagate. A list of a hundred addresses assembled from directories will contain dead protocol entries, long-defunct services, mistyped strings and clones alongside whatever genuine, currently running services it happens to include. A scenario: a reader tries twenty addresses from a list and eight open. The interesting question is not why twelve failed but how many of the eight are what they claim to be.
Reading Tor Browser's errors when onion links are not working
Tor Browser gives specific error pages for onion service failures, and learning to read them turns onion links not working from a vague complaint into a diagnosis. - Onionsite not found. The browser could not find the service's descriptor in the network's distributed directory. This usually means the service is offline, since running services publish descriptors regularly, or the address is wrong but well-formed. - Onionsite cannot be reached. The descriptor was found but the browser could not complete the connection through the introduction points, typically because the service went offline recently or is overloaded. - Onionsite has disconnected. A connection was established and then dropped, usually indicating a service that is up but unstable. - Unable to connect to onionsite. A general failure to build the circuit, sometimes on the client's side. - Invalid onionsite address. The string is not a valid onion address: wrong length, wrong characters or a failed checksum. - Onionsite requires authentication. The service uses client authorisation and will only talk to clients holding a key. The practical value is that the errors sort failures into categories. Invalid address means the string is wrong and no retry will help. Not found means offline or wrong, and a retry later is reasonable. Cannot be reached or disconnected means the service exists and is struggling. A scenario: a reader gets invalid address on one link and not found on another. The first is a lost cause; the second gets a retry tomorrow.
A diagnostic sequence for any working tor links that suddenly fail
When working tor links you have used before stop responding, a short sequence separates the likely causes without guesswork. 1. Confirm your own connection; if that fails, the problem is local. Open a clearnet site in Tor Browser, or the Tor Project's connection check page. 2. Confirm the address. Compare the string you are using against your verified bookmark, character by character. 3. Read the error. Invalid address means the string is wrong. Not found means offline or a wrong but well-formed address. Cannot be reached or disconnected means the service is present but struggling. 4. Try a fresh circuit from the circuit display. 5. Wait and retry. Many onion services are single machines and come back within hours. 6. Check the operator's channel. If the service has a clearnet presence or a known announcement channel, look there for maintenance notices, address changes or a shutdown announcement. 7. Only then consider that the service is gone. A verified address that fails for weeks with no word from the operator is probably defunct. What the sequence never includes: searching for the service's name to find a working address. That replaces a verified address that may be temporarily down with an unverified one that may be a clone. A scenario: a reader's bookmarked forum returns not found. Their connection works, the address matches, a new circuit does not help. They check the operator's account, find a note about a hardware failure, and wait. Two days later the same address works.
Why working is not the same as real: the problem with active onion sites
The most important thing to understand about active onion sites is that a page loading proves only that some machine is answering at that address. It does not prove anything about who runs the machine, and a large fraction of the working addresses on any list are working precisely because someone is operating a copy. Clones are more reliably online than the services they imitate. An impersonator running a phishing copy of a popular service has a strong incentive to keep it up, since downtime is lost credentials,. Lists that filter for working addresses therefore systematically favour clones, because clones pass the filter more consistently than originals do. Automated directories make this worse, and their online indicator comes to mean answering rather than genuine. Many directory sites check their entries by attempting a connection and mark those that respond as online or verified. This is why the two questions, does it work and is it real, must be kept apart. The first is answered by trying it. The second is answered only by provenance: tracing the address to the operator's own channel. A working address without provenance is an address you can visit read-only and nothing more. A scenario: a reader finds a directory listing five addresses for the same service, three marked online. They check the operator's clearnet site. It publishes one address, which is one of the three. The other two are copies, kept online by people who wanted the reader to log in.
What lists of active onion links and active dark web sites are actually measuring
Pages offering active onion links and active dark web sites. Most such lists are produced by a crawler. The crawler holds a pool of addresses gathered from other lists, attempts to connect to each on a schedule, and records which responded. The result is then published with a timestamp and labels such as online, active or last seen. What has been measured is that a machine at that address returned something at that time. What has not been measured is whether the machine is run by the claimed operator, whether the content matches the label, whether the service was ever legitimate, or whether it will be up in an hour. Several consequences follow. Labels are inherited from wherever the address was collected, so a clone labelled with the original's name keeps that name. Manually maintained lists, which are rarer, still tend to check responsiveness rather than provenance, because provenance takes a human an hour per entry and responsiveness takes a script a second. A useful test of any such list: does it state where each address came from, and does that source lead back to the operator. If the only claim is that the address is active, the list has told you something you can check yourself in ten seconds and nothing you actually need to know. A scenario: a reader finds a list with fresh timestamps beside every entry. Not one has a source. The list has confirmed that forty machines are on.
Building your own set of working dark web sites that stay working
The reliable way to have working dark web sites at hand is not to find a better list but to build a small, verified set and maintain it,. Start with services that have a reason to stay up. Institutional onion services, run by news organisations, privacy providers, nonprofits and software projects, run on professional infrastructure and are announced when they change. They are the backbone of any durable set, and their addresses come from the institutions' own clearnet sites. Record provenance and date for every address. A bookmark that says where it came from can be re-verified in a minute when it fails. Expect and tolerate downtime for hobbyist services. If a personal site or small forum matters to you, note that it runs on a single machine and that not found on a given evening is normal. Give it days, not minutes, before concluding anything. When a service does change address, the operator's channel is the only source that counts. Institutional services announce moves on their clearnet sites; hobbyist services announce on whatever channel they have always used. Retire rather than delete. An address that has failed for weeks moves to a retired folder with the date. A scenario: a reader maintains a Verified folder of fourteen addresses, twelve institutional and two hobbyist. Over a year, the twelve never fail for long, one hobbyist site vanishes and is retired, and the other has three multi-day outages and returns each time.
Working is the easy question
Working onion links are scarce because most onion services run on single machines with no redundancy, because operators stop without notice, because the network adds its own failures and because circulating lists are full of dead protocol entries, typos and defunct services. Tor Browser's error pages sort failures usefully: invalid address means the string is wrong, not found means offline or mistaken, cannot be reached means present but struggling. A short sequence, checking your connection, the address, the error, a fresh circuit, a later retry and the operator's channel, diagnoses nearly every failure without resorting to a search for a replacement. But working is the easy question, and lists that answer it are measuring responsiveness while saying nothing about identity. Clones are reliably online because impersonators need them to be, and directories that filter for activity drift toward the clones. A durable set of working addresses is built from operators' own channels, records provenance, tolerates hobbyist downtime and retires rather than replaces. One step to take today: pick one onion address you use that failed recently. Run the diagnostic sequence on it, ending at the operator's channel, and record what you learn beside the bookmark. Whether the answer is offline, moved or gone, you will have found it from the one source that could actually tell you, and the next failure will take a minute to understand.
Frequently asked questions
why do so many onion links not work?
Most onion services run on single machines without redundancy and disappear when the machine is off or the operator stops. Circulating lists also contain retired v2 addresses, typos and long-defunct services. A substantial failure rate is normal and does not indicate a problem with your browser.
what does onionsite not found mean in tor browser?
The browser could not find the service's descriptor, which usually means the service is offline or the address is wrong but well-formed. Retrying later is reasonable. Invalid onionsite address, by contrast, means the string itself is malformed and no retry will help.
how do i check whether an onion site is down or the problem is on my end?
Open a clearnet page or the Tor Project's connection check in Tor Browser first. If that works, compare the address to your verified bookmark, read the error page, request a new circuit for the site, and if it still fails, wait and check the operator's own channel for notices.
if an onion link works does that mean it is the real site?
No. A page loading proves only that a machine is answering at that address. Clones are often more reliably online than the originals they imitate, and directories that filter for working addresses drift toward clones. Only tracing the address to the operator's own channel establishes that it is genuine.
where can i find onion links that stay working?
Build a small set from institutional operators such as news organisations, privacy providers and software projects, whose onion services run on professional infrastructure and are announced when they change. Take each address from the operator's own clearnet site and record its source and date beside the bookmark.