Dark Web

How Big Is the Dark Web, Really?

Ask five sources how big the dark web is and you will get five different answers — some differing by a factor of ten. That is not a sign that everyone is lying. It is a sign that the dark web is designed to resist counting, and that "how big" is the wrong question until you decide what you are counting.

By Abdul Hannan · Published 14 November 2026 · Updated 14 November 2026

The conflicting counts

Here are the two best-known counts, side by side. They look contradictory. They are not — once you see what each one measured.

• A Recorded Future study counted around 55,800 onion domains, and found about 8,400 of them live — responding — at the time of the count. That is a count of websites that actually answered.
• Tor Metrics, the Tor Project's own statistics site, has shown figures in the region of 656,000 v3 onion addresses. That is a count of addresses that exist in the network's directory, not of live websites.

The gap between 8,400 live sites and 656,000 addresses is not an error. An address is not a website. One service can hold many addresses. Many addresses are generated and never used, or used once and abandoned. Many belong to private services that will never answer a stranger's request. And a count of live sites depends entirely on when you knock: onion services appear and vanish constantly, and a site that is live on Tuesday can be gone by Friday.

Last checked: 14 November 2026.

Five reasons the counts disagree

1. Addresses are not websites

An onion address is cheap to create — it is derived from a cryptographic key, and anyone can generate one in seconds. Counting addresses counts potential, not reality. Counting live websites requires actually visiting each address and seeing whether anything answers, which is slower, less complete, and only true for the moment of the visit.

2. Nobody has the master list

The ordinary web can be counted, roughly, because search engines crawl it and keep lists of what they find. The dark web has no central index. Onion addresses are not published in a directory unless their owners choose to list them. Researchers find addresses by crawling links between sites, by monitoring public sources, and by running their own relays — and every method misses something. A count is always a count of what the counter could find, not of what exists.

3. Sites appear and vanish constantly

Onion services are short-lived compared with ordinary websites. Marketplaces are seized or exit-scam their users. Forums move addresses after attacks. Personal sites go offline when their owner stops running them. A count from last year is not a count of today, and a count from this morning is already stale by next week. Any size figure without a date attached is meaningless.

4. Private and invite-only services are invisible

Many onion services are not meant to be found: private file stores, invitation-only forums, personal services shared with a few people. They will never answer a crawler, and they never appear in any public count. Any published number necessarily excludes them — and nobody knows how many there are.

5. Different counters use different definitions

Is a marketplace with fifty sub-pages one site or fifty pages? Is an address that answers with an error page "live"? Does a site that was live once in a month count? Each research team answers these questions differently, and each answer changes the total. Two honest teams can produce wildly different numbers from the same network, on the same day, by defining "site" differently.

The 90 percent myth, again

Every discussion of dark web size eventually meets the claim that the deep web is 90 to 95 percent of the internet, with the dark web as a huge slice of it. The figure traces back to a 2001 estimate by researcher Michael Bergman. It was an estimate of database content in a much smaller internet, not a measurement of anything, and it has been repeated for over two decades without anyone re-measuring it. Security researchers have explicitly debunked it, noting that it has no current basis.

Meanwhile, the most conservative published estimate of the dark web's share comes from Fortinet's documentation, which describes the dark web as likely accounting for less than 0.01 percent of the internet. That figure is itself an estimate, not a measurement — but it points in the direction every serious count points: the dark web is a small fraction of the internet, not a hidden ocean.

What about the deep web?

The deep web is genuinely large — your email, your banking, paywalled articles, private accounts, company intranets — but it is also unmeasurable in any honest sense, because so much of it sits behind logins that no one can survey. The honest statement is that the deep web is big and the dark web is a small part of it. Anyone who gives you a precise percentage for either is guessing, and the percentage they quote usually comes from the 2001 estimate.

What about users?

There is no reliable public count of dark web users, either. The Tor Project publishes statistics for Tor usage as a whole, but those figures cover all Tor use — much of it ordinary private browsing of the regular web, not visits to onion sites — and they do not separate dark web traffic from everything else. User numbers are also hard to define: is a person who opens Tor Browser once a month a "dark web user"? The Tor Project's own usage graphs are the honest source for Tor usage, and they say nothing about how many people visit onion sites. Anyone quoting a dark web user count as a fact is estimating, not measuring.

How to judge any size claim

When a page tells you how big the dark web is, ask these questions before believing it:

• What was counted? Addresses, live sites, pages, users, or bytes? If the page does not say, the number means nothing.
• How was it counted? Crawling, directory data, relay observation, or guessing? A method should be named.
• When? A count without a date is already stale. Onion services turn over constantly.
• Who counted it? A research firm, a vendor with something to sell, or an anonymous blog? Vendors have a long history of inflating the dark web to sell security products.
• Does it cite the 2001 estimate? If a page quotes 90 to 95 percent for the deep web, it is repeating folklore, and nothing else on the page should be trusted without checking.

The honest summary

The dark web is small. The best counts of live onion services run to the thousands, not the millions. The number of addresses is far larger, but most addresses are not live websites. The deep web is much larger — but it is mostly your email and your bank, not hidden crime. And the precise size of either will never be known, because both are built from things that cannot all be counted.

That uncertainty is not a failure of research. It is a feature of the technology: a network designed so that no one can see the whole of it is, by the same token, a network no one can measure the whole of. The honest answer is a range with a method and a date — and this page will keep giving you that instead of a single confident number.

Sources

• Tor Metrics — Servers: relays and bridges (accessed 14 November 2026)
• Tor Project — A short introduction to Tor (Tor Specifications, accessed 14 November 2026)
• Onion services — Wikipedia (accessed 14 November 2026)

The Recorded Future domain count and the Tor Metrics address figure are reported in the research reviewed for this site, which also documents the CloudSEK correction of the 2001 Bergman percentages. Last checked: 14 November 2026.

Frequently asked questions

How big is the dark web?

Nobody can say with a single reliable number. Published counts disagree wildly because they measure different things: active websites, registered addresses, live domains, or content volume. The honest picture is that the dark web is small — thousands of onion services, not millions — and far smaller than the deep web.

How many onion sites are there?

It depends on what you count. One published study counted around 55,800 onion domains, of which about 8,400 were live at the time. Tor Metrics, which counts addresses rather than live sites, has shown figures in the hundreds of thousands. Those numbers are not contradictory — they measure different things.

Is the dark web 90 percent of the internet?

No. That figure traces back to a 2001 estimate that was never a measurement, and researchers have been correcting it for years. The 90 to 95 percent figure was about the deep web, not the dark web — and even for the deep web, it has no current basis.

Why do dark web size estimates disagree so much?

Because the dark web is designed to resist counting. Addresses change, sites appear and vanish, many are private or invite-only, and nobody has a master list. Different counters use different methods, different definitions of a site, and different dates — and each of those choices changes the answer.

How many people use the dark web?

There is no reliable public headcount. The Tor Project publishes usage statistics for the Tor network as a whole, but those figures cover all Tor use — including ordinary private browsing — not dark web visits specifically. Anyone quoting a dark web user count is estimating, not measuring.