Data study · June 15, 2026
We probed 9,992,781 of the world’s most popular domains and labelled each one alive, redirect, blocked, or dead. The real dead figure is 14.1% — not the 27.6% a naive crawl reports, because most of “dead” is just blocking you.
14.1%
of the 9,992,781 probed top domains are genuinely dead — gone from DNS or refusing every connection.
8.9%
answer but block bots
1.1%
of responders are parked
Homepage-level reachability from a datacenter IP — a lower bound.
14.1%
of the 9,992,781 probed domains are genuinely dead — no DNS, no connection, nothing answers. That is the real dead-web figure, not the 27.6% a naive crawl reports.
8.9%
answer but block automated clients (403/429/challenge) from a datacenter IP — alive, just not to a bot. Naive scans count these as dead.
10.3%
of all domains no longer resolve in DNS — the dominant cause of true death, 1,027,492 domains gone dark.
33%
.cn is the deadest common TLD — institutional and cheap-registration TLDs rot fastest, well above the .com baseline.
Every probed domain, by outcome
A naive 2024 crawl of the same top-10M list reported 27.6% dead. Probe honestly — separating genuine death from anti-bot blocking and answered errors — and the real figure is 14.1%. Here is where the difference goes.
DNS failure, anti-bot 403s, 404/5xx and timeouts all lumped together
No DNS, connection refused, or nothing accepts a connection
Where the “dead” really goes
The same domains, probed by an honest bot and by a browser-like client (real Chrome TLS/JA3). Where the browser column is lower on dead/blocked, the site is reachable — the bot just wasn't let in.
| Probe arm | Probed | Alive | Blocked | Dead | Dead % |
|---|---|---|---|---|---|
| Polite bot | 9,992,781 | 7,657,422 | 891,517 | 1,412,544 | 14.1% |
| Reachability (browser) | 9,997,315 | 7,743,245 | 819,599 | 1,412,889 | 14.1% |
| China (.cn) | 33% |
| India (.in) | 25.8% |
| United States of America (.us) | 22% |
| Brazil (.br) | 20.9% |
| Spain (.es) | 16.6% |
| Japan (.jp) | 15.6% |
| United Kingdom (.uk) | 15.3% |
| Australia (.au) | 15% |
| Russia (.ru) | 14.8% |
| France (.fr) | 14.5% |
| Canada (.ca) | 14.1% |
| Italy (.it) | 13.5% |
| Poland (.pl) | 13.1% |
| Sweden (.se) | 11.6% |
| Switzerland (.ch) | 9.8% |
| Netherlands (.nl) | 9.7% |
| Austria (.at) | 8.6% |
| Germany (.de) | 7.6% |
| Czechia (.cz) | 7.2% |
The gap between 27.6% and 14.2% is mostly a measurement choice. A crawler that stops at the first response sees only 45.9% return a clean 200; follow the redirects and read the bodies, and 71.9% are alive. Here is where every first response ends up.
| 200 OK → Alive | 4,584,611 (46.3%) |
| 3xx redirect → Alive | 2,677,304 (27%) |
| No response → Dead | 1,413,013 (14.3%) |
| 403 / 429 → Blocked | 410,511 (4.1%) |
| 3xx redirect → Blocked | 365,368 (3.7%) |
| 404 → Alive | 236,685 (2.4%) |
| No response → Blocked | 105,222 (1.1%) |
| 5xx → Alive | 85,728 (0.9%) |
| 3xx redirect → Redirect | 31,267 (0.3%) |
| 3xx redirect → Dead | 1,775 (0%) |
Split the 10 million by popularity and the dead rate climbs more than 20× — from 0.8% in the top 1,000 to 16.1% past rank 5 million — while blocked runs the other way, peaking at the popular head.
99.8% of dead domains sit below rank 100,000. The popular top-100K — where most web traffic lives — is only 2.2% dead, so weighted by attention the dead web nearly disappears:
share of the top 10M that are dead
the popular top-100K is only 2.2% dead
Search, filter by outcome, switch the probe arm, and sort. The full dataset is on GitHub. Click a domain to see how each arm fared.
| Rank | Domain | Outcome | Reason | Status | Final URL |
|---|---|---|---|---|---|
| 351 | squareup.com | Alive | ok | 200 | https://squareup.com/us/en |
| 352 | spoti.fi | Alive | ok | 200 | https://open.spotify.com/ |
| 353 | www.jotform.com | Alive | ok | 200 | https://www.jotform.com |
| 354 | www.census.gov | Alive | ok | 200 | https://www.census.gov |
| 355 | de.wikipedia.org | Alive | ok | 200 | https://de.wikipedia.org/wiki/Wikipedia:Hauptseite |
| 356 | theconversation.com | Alive | ok | 200 | https://theconversation.com/us |
| 357 | www.cnet.com | Alive | ok | 200 | https://www.cnet.com |
| 358 | www.samsung.com | Alive | ok | 200 | https://www.samsung.com/us/ |
| 359 | marketingplatform.google.com | Alive | ok | 200 | https://marketingplatform.google.com/about/ |
| 360 | getpocket.com | Alive | ok | 200 | https://getpocket.com/home |
| 361 | www.prnewswire.com | Alive | ok | 200 | https://www.prnewswire.com |
| 362 | httpd.apache.org | Alive | ok | 200 | https://httpd.apache.org |
| 363 | apnews.com | Alive | ok | 200 | https://apnews.com |
| 364 | validator.w3.org | Alive | ok | 200 | https://validator.w3.org |
| 365 | fonts.google.com | Alive | ok | 200 | https://fonts.google.com |
| 366 | jquery.com | Alive | ok | 200 | https://jquery.com |
| 367 | www.starwars.com | Alive | ok | 200 | https://www.starwars.com |
| 368 | youtube-global.blogspot.com | Alive | ok | 200 | https://blog.youtube/ |
| 369 | www.zdnet.com | Alive | ok | 200 | https://www.zdnet.com |
| 370 | www.startnext.com | Alive | ok | 200 | https://www.startnext.com |
| 371 | www.icann.org | Blocked | forbidden | 403 | https://www.icann.org |
| 372 | www.wordpress.org | Alive | ok | 200 | https://wordpress.org/ |
| 373 | gemini.google.com | Alive | ok | 200 | https://gemini.google.com |
| 374 | www.curseforge.com | Blocked | forbidden | 403 | https://www.curseforge.com |
| 375 | www.acm.org | Blocked | forbidden | 403 | https://www.acm.org |
| 376 | academic.oup.com | Blocked | forbidden | 403 | https://academic.oup.com |
| 377 | tv.youtube.com | Alive | ok | 200 | https://tv.youtube.com/welcome/?utm_servlet=prod&rd_rsn=lo |
| 378 | www.irs.gov | Alive | ok | 200 | https://www.irs.gov |
| 379 | nginx.org | Alive | ok | 200 | https://nginx.org |
| 380 | staysafeonline.org | Alive | ok | 200 | https://www.staysafeonline.org/ |
| 381 | www.pwc.com | Alive | ok | 200 | https://www.pwc.com/us/en.html |
| 382 | lexfridman.com | Alive | ok | 200 | https://lexfridman.com |
| 383 | de.m.wikipedia.org | Alive | ok | 200 | https://de.wikipedia.org/wiki/Wikipedia:Hauptseite |
| 384 | blog.hubspot.com | Alive | ok | 200 | https://blog.hubspot.com/ |
| 385 | php.net | Alive | ok | 200 | https://www.php.net/ |
| 386 | codeberg.org | Alive | ok | 200 | https://codeberg.org |
| 387 | www.pnas.org | Blocked | forbidden | 403 | https://www.pnas.org |
| 388 | microsoft.com | Alive | ok | 200 | https://www.microsoft.com/en-us |
| 389 | lh6.googleusercontent.com | Alive | client_error | 400 | https://lh6.googleusercontent.com |
| 390 | www.coursera.org | Alive | ok | 200 | https://www.coursera.org |
| 391 | www.scribd.com | Alive | ok | 200 | https://www.scribd.com |
| 392 | www.udemy.com | Blocked | forbidden | 403 | https://www.udemy.com |
| 393 | www.nvidia.com | Alive | ok | 200 | https://www.nvidia.com/en-us/ |
| 394 | pk.linkedin.com | Alive | ok | 200 | https://pk.linkedin.com |
| 395 | coub.com | Alive | ok | 200 | https://coub.com/featured/creators |
| 396 | google.qualtrics.com | Blocked | forbidden | 403 | https://google.qualtrics.com |
| 397 | www.docker.com | Alive | ok | 200 | https://www.docker.com |
| 398 | www.unicef.org | Blocked | forbidden | 403 | https://www.unicef.org |
| 399 | zen.yandex.ru | Alive | ok | 200 | https://sso.passport.yandex.ru/push?uuid=086c6b44-20a2-4ca0-8ed6-56f5b5632be3&retpath=https%3A%2F%2Fdzen.ru%2F%3Futm_referrer%3Dzen.yandex.ru%26is_autologin_ya%3Dtrue |
| 400 | www.wikihow.com | Alive | ok | 200 | https://www.wikihow.com/Main-Page |
We probe a top-popularity domain list HTTPS-first from a datacenter IP, following redirects, and label each domain alive, redirect, blocked, or dead by the evidence the probe captures — a final HTTP status, or a transport error plus whether a raw TCP connect still succeeds. A served 404, a 5xx, or a Cloudflare 52x is alive (the host answered); a 403/429 or anti-bot challenge is blocked; only no DNS, a refused/reset connection, or nothing accepting a connection is dead. Every domain is probed twice — as a polite bot and as a browser-like client (real Chrome TLS/JA3) — and the full per-domain dataset is open.
This measures whether the domain itself still resolves and answers — a different question from Pew Research’s 2024 link-rot study (25% of pages from 2013–2023 are gone; 38% of 2013 pages) and Ahrefs’ link-rot study (66.5% of links have rotted), which measure broken links insideliving pages. It is also not the “dead internet theory” — that is a claim about AI-generated content, not domain reachability.
Cite this
Crawlora (2026). Dead-Web Index 2026. 14.1% of 9,992,781 top domains are genuinely dead; 8.9% answer but block automated clients. https://crawlora.net/dead-web-index.
14.1% of the top 9,992,781 domains are genuinely dead — about 1,412,544 sites that no longer resolve in DNS or refuse every connection. That is far below the often-quoted "27.6% of the web is dead," which counted anti-bot blocks and answered errors as death.
A dead site never answers — no DNS record, or nothing accepts a TCP connection. A blocked site is alive and answering, it just refuses an automated client (a 403, 429, or anti-bot challenge). 8.9% of the top web (891,517 sites) is blocked, not dead — a distinction naive crawlers miss.
No. The dead internet theory is a claim that AI-generated content and bots have replaced human activity on the living web. This index measures the opposite and concrete thing: how many domains have gone completely dark and unreachable — DNS gone, connection refused, server gone.
Earlier top-10M crawls counted three non-dead things as dead: anti-bot 403/429 blocks, 404/5xx pages served by a live server, and domains a single flaky DNS resolver failed to look up. Classifying honestly — dead means genuinely unreachable — brings the real figure to 14.1%.
.cn has the highest death rate among common TLDs at 33%. Institutional TLDs like .gov and .edu also rank high — matching Pew Research's finding that government and reference pages suffer the worst link rot.
Anti-bot systems (Cloudflare, DataDome, and others) serve a 403 or a challenge to a datacenter IP while letting a real browser through. A matched browser TLS/JA3 fingerprint reaches the site where a naive bot is blocked — which is exactly why this index probes every domain twice, as a polite bot and as a browser-like client.
8.9% of the top web answers but blocks a naive bot. Crawlora escalates from a plain request to a real browser fingerprint only as far as a site demands, and bills on success — so you reach the live web that the 14.1% genuine dead doesn’t include.