Domain Registration & TLD Usage Statistics 2026: How ccTLDs and gTLDs Are Distributed

36.8% of registered domains serve a working website

Share of the 399,851,676 domains registered on the internet

Chart Table

Embed this figure

<a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-funnel"><img src="https://content.stackscan.com/charts/domain-tld-statistics-funnel.webp" alt="From registered to used" width="880" style="max-width:100%"></a> <p>Source: <a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-funnel">StackScan TLD analysis</a></p>

Four populations sit inside one another, and almost every published domain figure is about the first.

Registered. 399,851,676 names across 1,570 extensions. What registries count and what the trade press quotes.

Resolvable. 292,821,718 of those, 73.2%, publish an A or AAAA record, so looking the name up returns an IP address.

Answering. 150,726,446, 37.7% of everything registered, return a page when one is asked for.

Active. 146,990,668, 36.8%, return a page that is not an offer to sell the name. This is the population every other report on this site measures, and the only one of the four that describes something a person can read.

Resolvable means that and nothing more. It does not mean anything is listening on the address, and it does not mean a page comes back. A name carrying only nameserver or mail records and no address is not counted. An extension is also called a zone, after the zone file its registry publishes, and the two words mean the same thing everywhere below.

Each contains the next. A domain cannot serve a page without resolving, so the active web sits inside the answering web, which sits inside the resolvable internet, which sits inside the registered one.

Measured separately they now agree almost exactly: 353 active domains are missing from the DNS scan, out of 150,726,446.

Of the domains that answer, 3,735,778 answer only to offer the name for sale, 2.5% of them. The hosting report finds where those names sit: three of the four busiest IP addresses on the internet are this inventory, and the busiest gets a page out of under one per cent of the names pointing at it. Parking is far larger than that figure and mostly invisible to a page request: 31,975,287 registrations carry a parking signature, and seven in eight of those never serve a page at all. The 2.5% above is of live domains; this one is of registrations.

The remaining 249,125,230 domains return no page at all, 62.3% of everything registered. They are renewed, they sit in a zone file, and nothing answers.

A domain name is not a website, and the distance between the two is about two thirds of everything registered.

28.5% of registered domains have an IP address and no website

Every registration is in exactly one of them, and the first three have an address

399,851,676domains are registered

Each one ends in exactly one of four statesRegistries publish the first number. The other four are measured here.

Of those 399,851,676 registrations

has an IP address: the resolvable internet, 292,821,718, 73.2%

• 146,990,668 serve a working website: they answer with a page that is not a sale notice (36.8%)
• 31,975,287 are parked or listed for sale, on a platform that sells or monetises names (8.0%)
• 113,856,262 have an IP and no website: an A or AAAA record, no usable page behind it (28.5%)
• 107,029,610 never resolve: no address at all, the name exists in a zone file and nowhere else (26.8%)

Embed this figure

<a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-domain-states"><img src="https://content.stackscan.com/charts/domain-tld-statistics-domain-states.webp" alt="The four states a registered domain can end in" width="880" style="max-width:100%"></a> <p>Source: <a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-domain-states">StackScan TLD analysis</a></p>

The 50 largest extensions, each bar its own registrations

Chart Table

Embed this figure

<a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-states"><img src="https://content.stackscan.com/charts/domain-tld-statistics-states.webp" alt="What became of every registration" width="880" style="max-width:100%"></a> <p>Source: <a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-states">StackScan TLD analysis</a></p>

Every registration ends in one of four states, and the largest of them is not a website. 28.5% of registered domains have an A or AAAA record and no usable page behind it: 113,856,262 domains that somebody pointed somewhere and never turned into a site.

26.8% never resolve at all. 8.0% are held for sale. 36.8% serve a working website.

That third state, an address with no website, is the one nobody counts, and it is a residual rather than a measurement: whatever has an IP address, is not parked and did not come back with a usable page lands in it. So it is worth opening.

Shares of the 120,786,964 domains the crawler could classify, a slightly wider set than the state above

What happened	Domains	Share
Not classified	49,008,779	40.6%
Served a page too small or empty to count	32,769,447	27.1%
The fetch did not complete	23,837,217	19.7%
Redirected to another domain	15,171,521	12.6%

Embed this figure

<a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-middle"><img src="https://content.stackscan.com/charts/domain-tld-statistics-middle.webp" alt="What is in the address-but-no-website state" width="880" style="max-width:100%"></a> <p>Source: <a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-middle">StackScan TLD analysis</a></p>

One caution before the shares. They are taken over 120,786,964 domains, not over the 113,856,262 in the state above, because the classification comes from the crawler's own record of every fetch that failed and that record reaches a slightly wider set. Read them as the composition of the state, not as counts to multiply out.

27.1% served a page that was too small or too empty to call a website. 12.6% redirected to a different domain. 19.7% are a fetch that never completed: a timeout, a refused connection, a block page, a proxy that failed. That last group is our limit rather than theirs, and it is the honest reason this state is described as "no usable page came back" rather than "nothing is there".

The remaining 40.6% carry no failure label at all.

A domain with DNS pointing somewhere has had money and attention spent on it beyond the registration fee, and it still is not a site.

In the figure above, sorted by the working-website share, the extensions separate cleanly: country extensions at the top, the extensions sold in bulk at the bottom, and the bottom of that list is mostly the two states that are not websites.

.com holds 49.2% of every domain on the internet

Everything from here on measures extensions, and there are three different ways to do that. Registration share asks how many names exist in a zone. Active share asks how many of them serve a working website. Technology share asks how much is built on those websites. An extension is sold on the first and used on the third, and the sections below take them in that order.

Share of the 399,851,676 domains registered on the internet

Chart Table

Embed this figure

<a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-concentration"><img src="https://content.stackscan.com/charts/domain-tld-statistics-concentration.webp" alt="Where the domains are" width="880" style="max-width:100%"></a> <p>Source: <a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-concentration">StackScan TLD analysis</a></p>

.com alone carries 196,616,562 registrations. The largest ten extensions hold 72.5% of the internet between them, and the largest hundred hold 96.0%.

That leaves 4.0% for everything else, and the tail inside that tail is the striking part: the smallest 870 zones hold 185,110 domains between them, fewer than .ee holds on its own.

Share of the 1,570 extensions in use

Chart Table

Embed this figure

<a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-sizes"><img src="https://content.stackscan.com/charts/domain-tld-statistics-sizes.webp" alt="Most extensions hold very few names" width="880" style="max-width:100%"></a> <p>Source: <a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-sizes">StackScan TLD analysis</a></p>

608 zones hold fewer than a hundred domains each. 40 extensions have more than a million, and those 40 hold 88.6% of the internet.

The number of extensions available has grown for a decade and the distribution has not moved with it. A new extension is a product launch, and almost all of them stay small: 608 hold fewer than a hundred names each.

The working-website rate runs from 6.8% to 61.7%

Share of each extension's own registrations that serve a working website, extensions over 500,000 domains

Chart Table

Embed this figure

<a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-deadest"><img src="https://content.stackscan.com/charts/domain-tld-statistics-deadest.webp" alt="The emptiest large extensions" width="880" style="max-width:100%"></a> <p>Source: <a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-deadest">StackScan TLD analysis</a></p>

Share of each extension's own registrations that serve a working website, extensions over 500,000 domains

Chart Table

Embed this figure

<a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-alive"><img src="https://content.stackscan.com/charts/domain-tld-statistics-alive.webp" alt="The fullest large extensions" width="880" style="max-width:100%"></a> <p>Source: <a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-alive">StackScan TLD analysis</a></p>

Every extension in the first figure is open to anyone, anywhere, in any quantity. The second figure is not the mirror image of that, and it is worth being exact about why, because the obvious explanation is only part of it.

Some of the fullest extensions do restrict who may register. .jp needs a permanent postal address in Japan. .it needs citizenship, residence or a headquarters in the EEA, Switzerland, the UK, San Marino or the Vatican.

Others do not. .za sits at the top of the figure and co.za accepts registrations from South Africans and non-South Africans alike. .sk dropped its Slovak-address requirement in 2017 and now takes any postal address in the EU, EEA, EFTA or the UK. .hu has admitted applicants from any country since March 2025.

So local presence is not the mechanism on its own. What the extensions at the top of this figure have in common is that they are national, and the most likely reading is a home market registering for its own use rather than for the bulk trade. That is an interpretation and not a measurement: this data records what a name does, never who bought it or through which registrar. An extension that is cheap to buy a hundred thousand of will have a hundred thousand of them bought, and the buyer will not build a hundred thousand websites.

39.8% of .xyz is held for sale

Share of each extension's own registrations held for sale

Chart Table

Embed this figure

<a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-speculation"><img src="https://content.stackscan.com/charts/domain-tld-statistics-speculation.webp" alt="Held as inventory, not used as a website" width="880" style="max-width:100%"></a> <p>Source: <a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-speculation">StackScan TLD analysis</a></p>

39.8% of every .xyz domain registered is held for sale: 4,516,409 names on a parking or domain-sale platform, against the websites actually running on .xyz. The same extension serves a working website on 10.9% of its registrations, so .xyz carries more than three names in inventory for every one in use.

.com is the volume. 8.8% of its registrations are held for sale, which is 17,282,948 parked names on one extension, more than the entire .de zone. Behind it .us runs 19.8% of a smaller zone and .de 8.5% of a larger one.

Held for sale is not the same as answering with a sale page. Most parked domains never serve a page at all: 31,975,287 domains sit on a parking platform and only 3,735,778 of them return one, 2.5% of everything live. The rest are pointed at a parking service and render nothing a crawler can keep.

Parking is measured, not guessed, and it takes two passes to see it. One reads the technologies on a page we stored, which only works when the page was worth storing. The other is the crawler's own verdict at fetch time, on domains whose page never survived: a for-sale lander, a parking nameserver, a parking IP.

The two barely overlap, 1,030,110 domains out of 31,975,287, because each sees what the other cannot. A domain on a registrar's default page with no signature we hold is still counted as unparked, so every figure here is a floor.

One caution about the word, and it applies to every parked figure in this report. Parked and for sale are two different things, and the platforms do not divide along that line.

A name on Atom or Dan is a listing, and it is for sale. A name on Sedo usually is. A name on a GoDaddy or Namecheap parking page may be for sale, or it may be earning advertising revenue with no offer attached, or it may be a name its owner is simply holding.

The signal says which platform answered for the name, not what the holder intends, so "held for sale" here means held on a platform that exists to sell or monetise names rather than to publish one. Where the distinction matters, the count of domains that actually serve a sale page is given separately.

One platform per domain, and what share of every parked name on the internet it holds

Parking or sale platform	Domains	Share of all parked
GoDaddy	16,418,562	51.2%
Afternic	5,868,344	18.3%
SEDO	3,801,782	11.9%
Above.com	1,182,890	3.7%
ParkLogic DNS	1,023,951	3.2%
ParkingCrew DNS	793,413	2.5%
DNS-Expired (Parking)	591,486	1.8%
Godaddy CashParking DNS	561,446	1.8%
Ztomy (Domain Parking) DNS	319,930	1.0%
Not attributed	314,191	1.0%
SNPARKING DNS	290,855	0.9%
Aftermarket.pl DNS	210,635	0.7%
ParkLogic	148,165	0.5%
Bodis	135,235	0.4%

Embed this figure

<a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-providers"><img src="https://content.stackscan.com/charts/domain-tld-statistics-providers.webp" alt="Who is holding the parked domains" width="880" style="max-width:100%"></a> <p>Source: <a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-providers">StackScan TLD analysis</a></p>

The parking market has one owner and a long tail. GoDaddy holds 16,418,562 on its own, 51.2% of every parked name on the internet.

GoDaddy also owns Afternic and CashParking, which sit in the table as their own brands. Counted together the group holds 22,848,352 parked domains, 71.2% of every parked name on the internet. Sedo, Above.com and ParkLogic follow, and the remaining two dozen platforms in the table carry what is left.

Almost none of GoDaddy's answer with a page of their own: the name bounces a visitor to a sale page on another host and leaves nothing behind on the domain itself. Above.com and Afternic serve theirs, which is why a page-reading measurement sees those two and not the platform holding half the market.

Those names are not invisible everywhere. The mail report found GoDaddy publishing an enforcing DMARC policy across millions of domains with no MX record at all, which is very likely much of this population seen from the DNS side, though the two sets were counted separately and never joined.

That is the whole reason the parked figure was half of what it should have been. A measurement that reads technologies off a stored page cannot see a platform whose entire design is to store nothing.

The 32,077,545 domains carrying a parking or domain-sale signature

Identified by	Domains	Of which serve a page
A sale page or header	24,269,144	2,372,192
A parking nameserver or mail record	6,924,099	1,128,110
A sale page and a DNS record	498,391	293,224
A parking IP address	385,911	0

Embed this figure

<a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-parking"><img src="https://content.stackscan.com/charts/domain-tld-statistics-parking.webp" alt="How a domain was found to be parked" width="880" style="max-width:100%"></a> <p>Source: <a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-parking">StackScan TLD analysis</a></p>

Most are found by what they serve rather than by where they point. A sale page or header accounts for 24,269,144 of them, a parking nameserver or mail record for 6,924,099, 498,391 carry both, and 385,911 were caught by the IP they resolve to and nothing else.

The sale-page group is the one that shows how little of parking is a website. Of its 24,269,144 domains only 2,372,192 count as serving a page at all: the rest answer with a redirect to a sale lander and nothing else, which is why parked is its own state here rather than a slice of the live web.

Counting registrations measures a market in names and counting websites measures a market in publishing, and the extensions at the top of this figure are trading in the first.

The largest extensions are mostly not websites either

Share of each extension's own registrations that serve a working website

Chart Table

Embed this figure

<a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-biggest"><img src="https://content.stackscan.com/charts/domain-tld-statistics-biggest.webp" alt="The twenty largest extensions" width="880" style="max-width:100%"></a> <p>Source: <a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-biggest">StackScan TLD analysis</a></p>

.com serves a working website on 39.2% of its registrations, which is better than the internet's 36.8% and not by much. .xyz has 11,337,202 registrations and is a website on 10.9% of them.

.cn and .top and .info sit in the same band. They are large and lightly used at the same time. Neither of the two obvious explanations survives the rest of this report: the price bands below find the renewal price predicts almost nothing, and restricted extensions turn out to be built on no more often than open ones.

Size and use are different measurements and an extension can lead on one while sitting near the bottom of the other.

A working website on .es runs 3.7 times the technology of one on .top

Technologies detected per working website, extensions with at least 300,000 working websites

Chart Table

Embed this figure

<a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-density"><img src="https://content.stackscan.com/charts/domain-tld-statistics-density.webp" alt="How much is built on each extension" width="880" style="max-width:100%"></a> <p>Source: <a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-density">StackScan TLD analysis</a></p>

Counting only websites that work, an extension still predicts how much is built on one. A working .es site runs 19.9 detected technologies. A working .com site runs 13.1. A working .top site runs 5.4.

This is a different measurement from the active rate and it is not a restatement of it. Every site counted here is live and serving a page. The cheap extensions are not only emptier, the sites that do exist on them are thinner.

The exception is instructive. .dev carries 9 technologies a site, below .com, and it is not a weak extension: it is developers publishing one page about themselves with a modern framework and nothing else on it.

Size, liveness and complexity are three different measurements, and an extension can lead on one while sitting near the bottom of another.

.dev websites are 7.1 times more likely to run Next.js than .com

Share of each extension's own working websites

Extension	Working websites	WordPress	Shopify	Wix	Squarespace	Webflow	Next Js
.dev	355,595	4.4%	0.1%	0.2%	3.4%	0.3%	17.3%
.ai	331,548	8.7%	0.4%	2.3%	2.6%	2.7%	15.5%
.app	807,745	4.2%	0.2%	0.4%	2.4%	0.4%	14.0%
.io	382,145	13.9%	0.7%	2.2%	4.7%	2.7%	11.1%
.in	771,857	28.4%	3.7%	2.9%	0.6%	0.2%	3.8%
.co	541,195	25.2%	6.8%	5.5%	7.4%	1.7%	3.8%
.us	548,232	15.1%	4.5%	2.3%	4.2%	0.4%	3.6%
.xyz	1,312,936	11.6%	0.5%	0.4%	0.7%	0.2%	3.5%
.pro	551,325	11.2%	0.6%	0.6%	0.7%	0.2%	3.0%
.site	933,961	17.2%	1.5%	0.4%	0.4%	0.1%	2.9%
.online	1,558,075	13.0%	2.7%	1.9%	0.7%	0.1%	2.8%
.top	1,277,783	5.3%	0.5%	0.0%	0.0%	0.0%	2.7%
.store	929,227	9.1%	36.2%	1.3%	1.9%	0.0%	2.5%
.com	78,562,496	21.4%	3.7%	6.6%	8.2%	0.4%	2.4%
.br	1,496,023	36.4%	1.0%	8.1%	0.2%	0.3%	2.4%
.org	5,743,144	26.3%	0.8%	6.6%	10.5%	0.4%	1.8%

Embed this figure

<a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-platforms"><img src="https://content.stackscan.com/charts/domain-tld-statistics-platforms.webp" alt="What those websites are built on" width="880" style="max-width:100%"></a> <p>Source: <a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-platforms">StackScan TLD analysis</a></p>

17.3% of working .dev websites run Next.js, against 2.4% of .com's. 2.7% of working .io websites run Webflow, 6.9 times .com's rate.

WordPress runs the other direction. It is 38.5% of working .it websites and 21.5% of .com's, against 4.4% of .dev's.

Every share here is of that extension's own working websites. Taken over registrations instead the shares break, because a technology detected in DNS is detected on parked domains too and those are not websites.

The readiest explanation is that an extension is chosen by a kind of builder who brings a stack with them, which would be why an extension tracks a technology at all. The association is measured and the explanation is not, and it is the part of this report a registry's own statistics cannot produce. The CMS report measures those platforms across the whole web, where the same split shows up by country rather than by ending.

Cheap extensions are 2 of the top thousand websites

Each bar is that band's own domains, Tranco ranking

Chart Table

Embed this figure

<a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-bands"><img src="https://content.stackscan.com/charts/domain-tld-statistics-bands.webp" alt="The mix changes with the ranking" width="880" style="max-width:100%"></a> <p>Source: <a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-bands">StackScan TLD analysis</a></p>

Each bar is one slice of the ranking, split by what kind of extension the domains in it use. .com is 59.8% of the top thousand and 49.2% of every registration on the internet, which is the least interesting line on the chart. Everything else moves.

The cheap new extensions are almost absent from the top and ordinary at the bottom: 2 of the top thousand websites, 12.9% of all registrations, and 11,337,202 on .xyz alone.

.net and .org run the same way in reverse. They are 17.5% of the top thousand and 7.3% of the whole, which is what a legacy extension looks like: bought early by the sites that mattered early, and not bought much since.

Each band is cumulative: the top million row contains the top thousand. Country extensions peak in the middle rather than at either end. They are 17.5% of the top thousand, 36.4% of the Top 1 million band, and 23.3% of everything registered. The top of the web is international and buys .com, the bottom is bulk registration, and the working web in between is national.

Only 11 extensions are bigger than co.uk

Share of each suffix's own registrations that answer, suffixes over 100,000

Chart Table

Embed this figure

<a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-secondlevel"><img src="https://content.stackscan.com/charts/domain-tld-statistics-secondlevel.webp" alt="Where a country actually registers" width="880" style="max-width:100%"></a> <p>Source: <a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-secondlevel">StackScan TLD analysis</a></p>

5,663,846 domains are registered under co.uk, part of the .uk extension. That is 85.8% of .uk itself, and more registrations than all but 11 extensions hold. com.br and com.au follow.

Whether a country registers at the second level at all is a registry policy, and it splits the list cleanly. .za puts 99.9% of its registrations under a second level label; .pl puts 19.5% and sells the rest directly.

The public suffix list is what says which labels are suffixes rather than subdomains, so this is measured rather than guessed at from the shape of a name.

A domain is counted once, under its extension, so co.uk sits inside .uk everywhere else in this report.

A country extension is not the same as a country

Population from the World Bank, 2024

Extension	People	Registered per 1,000	Websites per 1,000
.tv	9,646	29,569	8,367
.me	623,525	1,092	343
.ch	9,005,582	300	152
.nl	17,993,485	200	101
.dk	5,976,992	163	88
.se	10,569,709	146	84
.ee	1,372,341	146	82
.de	83,516,593	196	70
.cz	10,905,028	101	59
.at	9,177,982	116	59
.sk	5,422,069	96	58
.be	11,858,610	95	48
.no	5,572,279	99	47
.nz	5,290,000	96	46
.au	27,194,286	82	42
.fi	5,619,911	70	42

Embed this figure

<a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-percapita"><img src="https://content.stackscan.com/charts/domain-tld-statistics-percapita.webp" alt="A country and its extension" width="880" style="max-width:100%"></a> <p>Source: <a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-percapita">StackScan TLD analysis</a></p>

Set registrations against the population of the country whose extension it is and two different things fall out.

Most country extensions are used by their country, and the rate varies the way you would expect from internet economics rather than from policy. .ch leads ordinary national use at 152.2 working websites per thousand people, while 95 of the 214 countries we can match to an extension sit under one per thousand.

Ordinary means the country actually uses its own extension. .tv, .me, .cc, .io, .co, .ai and nine other two letter endings are set aside for that sentence, because they are sold for what their letters say rather than to the country they belong to; .me at 388 working websites per thousand Montenegrins is the clearest of them.

Then there is .tv. Tuvalu has 9,646 people and 285,219 registered .tv domains, 29.6 for every person in the country. Nobody thinks .tv is about Tuvalu, and the number says so plainly.

A two letter extension is assigned to a territory from the ISO 3166-1 alpha-2 list, not sold to one, with a handful of exceptions such as .uk, .eu and .su, and several have been leased to the meaning of their letters instead.

.army is not the army, and 4.2% of it is held for sale

Registrable names only, so a university's hundreds of hosts count as the one registration they are

Extension	Who may register	Registered	A working website	Held for sale
.academy	Open	95,913	29.4%	6.7%
.travel	Open	33,326	38.0%	6.7%
.gov	Restricted	29,757	16.8%	0.5%
.school	Open	27,660	37.6%	8.6%
.law	Open	26,951	38.6%	9.0%
.institute	Open	24,755	41.8%	7.9%
.edu	Restricted	17,664	32.6%	0.0%
.aero	Restricted	14,772	45.3%	3.3%
.college	Open	12,778	16.1%	6.4%
.army	Open	10,521	24.5%	4.2%
.jobs	Open	10,517	22.4%	4.8%
.coop	Restricted	10,136	47.0%	0.4%
.university	Open	9,213	28.1%	9.2%
.cpa	Restricted	6,818	52.1%	0.0%
.ngo	Open	5,544	48.2%	2.1%
.bank	Restricted	2,610	60.4%	0.0%
.museum	Restricted	1,416	33.1%	2.0%
.navy	Open	1,362	25.6%	5.4%

Embed this figure

<a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-institutional"><img src="https://content.stackscan.com/charts/domain-tld-statistics-institutional.webp" alt="The institutional extensions, and the ones that look institutional" width="880" style="max-width:100%"></a> <p>Source: <a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-institutional">StackScan TLD analysis</a></p>

Some extensions are restricted to the institution they name. .gov and .mil are for United States government and military bodies, .edu for accredited institutions, .bank and .pharmacy and .cpa for licensed firms. Others read as institutional and are sold to anyone: .army, .navy, .airforce, .academy, .university.

These figures count registrable names only. An institution runs hundreds of hosts under one registration, so counting names as the crawl finds them made .edu read six times its real size, and the error fell almost entirely on the restricted side.

Corrected, the two groups are almost identical on the measure this report leads with. The 12 restricted extensions serve a working website on 33.1% of their registrations and the 12 open ones on 32.7%.

They are not identical on speculation. The restricted extensions hold 0.8% of their names for sale and the open ones 7.1%, a factor of 8.9. So vetting who may register does not raise the share of a zone that gets built on. What it changes is how much of the zone is held for sale.

.army carries 10,521 registrations, of which 24.5% serve a working website and 4.2% are held for sale, and .airforce holds 10.9% for sale. .mil carries 744 names and not one of them is for sale, because there is nobody to sell one to.

Governments register at the second level, and that is where the working sites are

Restricted second levels only, government, academic, school, military and health, over five thousand names

Suffix	Registered	A working website
sch.id	32,642	62.6%
gov.cn	28,164	11.7%
ac.cn	25,479	6.7%
edu.vn	18,934	41.7%
edu.pl	16,268	59.0%
ac.in	11,116	62.1%
ac.id	10,533	54.2%
gov.uk	10,114	49.1%
edu.mx	9,774	64.8%
go.th	9,732	65.8%
edu.it	8,993	85.3%
gov.in	8,593	22.0%
edu.in	8,535	65.7%
nhs.uk	8,458	46.6%
edu.co	8,202	65.1%
ac.th	7,778	52.2%
gob.mx	7,267	44.4%
ac.uk	6,801	38.1%
gov.tr	6,630	13.9%
gov.co	6,472	56.3%

Embed this figure

<a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-govsld"><img src="https://content.stackscan.com/charts/domain-tld-statistics-govsld.webp" alt="Where governments and schools actually register" width="880" style="max-width:100%"></a> <p>Source: <a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-govsld">StackScan TLD analysis</a></p>

.gov is the weakest entry in the table above: a working website on 16.8% of its 29,757 registrations, against 32.6% for .edu's 17,664. A United States government domain is a published website about one time in six.

That is not a crawler being turned away. It is what a government registration is for: an agency registers the programme names, the campaign names and the misspellings, points them at one site, and publishes on that. The registration is a reservation, and reserving is most of what these zones are used for.

Outside the United States the pattern is different, because most countries put government and education at the second level of their own country extension rather than in a zone of their own. Most of those suffixes are used far more than the internet manages, and the education ones most of all.

Only restricted suffixes are counted here: gov, gc, gob, gouv, govt, go and gv for government, edu, ac and sch for education, mil for the armed forces, nhs for the health service. Nothing anyone can buy.

Education is the largest of the three families and the best used. 183 education suffixes carry 257,017 names, 52.4% of them a working website. 168 government suffixes carry 186,180 at 33.8%.

edu.it tops the figure at 85.3% and ac.cn is the floor at 6.7%, which is the same split .cn shows everywhere else in this report. The largest is sch.id at 62.6%, Indonesian schools.

Two cases are worth naming because they run against the pattern. gc.ca, which is the Canadian federal government, holds 2,115 names and serves a working website on 15.1% of them, which is what a federal government looks like after it consolidates onto one site and keeps the old departmental names reserved.

The military suffixes are the smallest population in this report. 44 countries run one and they hold 2,697 names between them, 14.5% a working website. Armed forces do not publish much on the open web, and the top level .mil reads the same way at 20.8%.

A name that has to be earned gets used, with the exception of the names nobody was ever meant to visit. It is the same shape as the country extensions at the top of the working-website figure, arrived at from a different direction.

42 cities have their own extension and 29.9% of the names on them are websites

Every city extension in use, by how many names are registered on it

Extension	Registered	A working website	Held for sale
.tokyo	137,200	31.0%	3.2%
.nyc	66,194	32.4%	9.9%
.berlin	45,677	28.1%	3.2%
.london	39,945	30.1%	7.2%
.paris	21,550	29.4%	1.8%
.amsterdam	19,305	24.4%	31.4%
.koeln	18,043	25.0%	1.7%
.miami	17,538	17.8%	12.5%
.hamburg	15,864	23.5%	6.7%
.wien	15,655	33.2%	1.2%
.vegas	13,167	20.3%	14.5%
.brussels	7,837	35.1%	2.1%
.quebec	7,187	27.0%	4.2%
.sydney	6,516	29.8%	4.0%
.boston	5,433	19.1%	11.6%
.melbourne	5,273	31.6%	3.5%
.istanbul	4,946	21.8%	6.9%
.barcelona	4,869	36.5%	2.4%
.okinawa	4,861	53.4%	1.0%
.moscow	4,854	57.8%	0.4%

Embed this figure

<a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-cities"><img src="https://content.stackscan.com/charts/domain-tld-statistics-cities.webp" alt="What a city does with its own extension" width="880" style="max-width:100%"></a> <p>Source: <a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-cities">StackScan TLD analysis</a></p>

A city can buy its own extension, and 42 of them have. Together they hold 502,227 registrations, which is fewer than .lol.

.tokyo is the largest at 137,200, then .nyc, .berlin and .london. Across all of them 29.9% serve a working website, below the 36.8% the internet manages, and 8 of the 42 have fewer than a thousand names on them at all.

The rate does not track the size of the city. .brussels runs 35.1% and .miami 17.8%.

The smaller Japanese extensions are the fullest of the lot, .okinawa at 53.9% of 5,089 names and .kyoto at 59.1% of 1,161, though the last of those is a small enough zone that the rate is a handful of sites either way. At the bottom .miami, .boston and .vegas sit under 21% on populations five to fifteen times larger. What separates them is whether local businesses were sold the extension or investors were.

.amsterdam makes the point on its own: 31.4% of its registrations are held for sale, the highest share of any city extension, and of every extension over ten thousand names only .ph and .xyz are higher. A city name is a brand, and a brand nobody is obliged to use is inventory.

Extensions that renew at 5x their first year serve a working website on 24.3% of their names

Share of each band's own registrations that serve a working website

Chart Table

Embed this figure

<a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-renewal"><img src="https://content.stackscan.com/charts/domain-tld-statistics-renewal.webp" alt="What the renewal costs against the first year" width="880" style="max-width:100%"></a> <p>Source: <a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-renewal">StackScan TLD analysis</a></p>

Two different price questions are being asked here. Does the renewal price itself track how much an extension gets used, and does the gap between the first year and the renewal track it? The answers are not the same.

Registration prices are promotions. The renewal is what a holder pays every year after, so it is the price used here, and the gap between the two turns out to be the thing that predicts what gets built.

Extensions that renew at five times their first year or more serve a working website on 24.3% of their registrations and held for sale on 12.9%. Extensions that renew at what they cost to buy run 39.0% and 7.9%. In between, 33.2%.

A first year at $1.99 against a renewal at $14.45, which is .xyz's own pricing, is not a discount, it is a different product. It prices a year of holding a name rather than a place to publish, and the second year is what sorts the two apart. 48 of the extensions measured here price that way.

Working websites in each band, and what share of that band's own registrations they are

Renewal price	Working websites	Share of its registrations	Held for sale
Under $10	22,548,361	34.5%	5.5%
$10 to $15	92,638,024	37.6%	9.7%
$15 to $20	3,152,723	29.7%	6.9%
$20 to $30	7,710,244	30.6%	5.7%
$30 to $50	4,517,431	30.9%	6.6%
$50 and up	1,005,055	39.1%	4.7%

Embed this figure

<a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-price"><img src="https://content.stackscan.com/charts/domain-tld-statistics-price.webp" alt="The renewal price on its own" width="880" style="max-width:100%"></a> <p>Source: <a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-price">StackScan TLD analysis</a></p>

The renewal price on its own predicts almost nothing, which is the part worth saying out loud. Across the bands the working-website rate moves between 29.7% and 39.1% with no direction to it. A dear extension is not a better-used one.

The counts say why a rate alone would mislead. The $10 to $15 band holds 92,638,024 working websites and the $50 and up band 1,005,055, so the two ends of the table are not comparable populations even where their rates look close.

Holding the discount still separates the two. Take only the extensions with no first year discount at all and compare the cheap ones with the dear ones: 38.4% against 39.1%. Half a point. Where the price is flat, the level of the renewal barely moves the rate, and the gap that does track with use is the one between the first year and every year after.

Prices are Dynadot's published list for 625 extensions, matched to the 108 measured here with at least a hundred thousand registrations, 91.2% of everything registered. Registrars discount, so read the bands rather than the boundaries.

nic is registered in 1,575 zones

How many different zones hold the same second level name, and how many of those serve a working website

Name	Zones it is registered in	A working website in
nic	1,575	756
com	927	146
amazon	917	202
google	872	309
1	863	182
info	814	209
data	800	251
news	794	352
bitcoin	782	295
a	779	193
api	777	265
2	734	84
ai	734	276
mail	734	244
cloud	728	259
e	723	251
home	723	267
seo	716	315
app	715	217
yahoo	712	164
3	704	99
online	700	282
abc	699	243
crypto	693	237
smart	693	266
btc	669	216
web	669	263
blockchain	668	219
prime	668	140
blog	666	175

Embed this figure

<a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-words"><img src="https://content.stackscan.com/charts/domain-tld-statistics-words.webp" alt="The names registered in the most extensions" width="880" style="max-width:100%"></a> <p>Source: <a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-words">StackScan TLD analysis</a></p>

Take the name off the front of every registration and count how many different zones hold it. A zone here means an extension or a second level that sells names, so co.uk counts separately from .uk and the total runs a little above the extension count. The answer at the top is not a brand.

nic leads with 1,575, which is registry policy rather than a market: ICANN's base registry agreement allocates the nic label to the registry and requires it live in the DNS for every extension from the 2012 round, and the older registries follow the same habit. The most widely registered name on the internet belongs to the people selling the names. It is also the best used name in the table and still not a majority: a working website in 756 of those zones, under half.

Below it the list is brand names and tradeable words side by side: amazon, google and yahoo next to bitcoin, crypto, btc, blockchain, seo, cloud and ai. Defensive registration and speculation are the obvious readings of those two groups, though the registrations themselves carry no motive.

Single letters and single digits appear near the top on their own, because a one character name is worth buying in any zone that will sell one.

Registering one name across a hundred extensions is not a publishing plan, and the figures say as much: across the 23,248 registrations in the table above, only 31.7% serve a working website, against 36.8% for the internet as a whole.

Every second level name in the zone data, counted by how many different extensions hold it

How many zones hold the name	Names	Registrations
In 500 zones or more	242	141,263
In 200 to 499	8,283	2,286,195
In 100 to 199	27,934	3,733,755
In 50 to 99	95,820	6,405,179
In 25 to 49	305,514	10,199,119
In 10 to 24	1,518,821	21,574,508
In 2 to 9	44,714,803	124,878,884
In one zone only	230,632,773	230,632,773

Embed this figure

<a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-wordbands"><img src="https://content.stackscan.com/charts/domain-tld-statistics-wordbands.webp" alt="How many names are held across how many zones" width="880" style="max-width:100%"></a> <p>Source: <a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-wordbands">StackScan TLD analysis</a></p>

The whole distribution says how rare that is. 277,304,190 different names hold the 399,851,676 registrations counted here, and 230,632,773 of them, 83.2%, exist in one zone and nowhere else. Someone wanted a name, bought it once, and stopped.

This one figure runs on its own denominator. It reads the name off a domain, so it can only count the two and three label registrations the apex rule recognises, which is 399,851,676 of the 399,851,676 in the rest of the report.

Above ten zones the behaviour changes. 1,956,614 names sit in ten extensions or more, 0.7% of all names, and they hold 44,340,019 registrations between them, 11.1% of the total.

At the top it is vanishingly thin: 242 names in five hundred zones or more, out of two hundred and seventy seven million.

A name in a hundred zones is a name with ninety nine registrations that carry no website, whoever holds them, and it is the reason the extension count is a market size rather than a usage figure.

One extension, one language

Share of each extension's own language declarations

Extension	Sites declaring	Top language	Share of declarations
.au	979,438	en	99.7%
.nz	228,856	en	99.6%
.za	411,563	en	99.4%
.uk	2,297,003	en	99.0%
.in	692,114	en	98.0%
.us	505,441	en	96.4%
.jp	795,724	ja	94.3%
.ai	288,902	en	92.8%
.ca	789,888	en	91.5%
.dev	332,999	en	90.7%
.io	317,463	en	89.7%
.рф	291,682	ru	84.8%
.live	286,359	en	84.2%
.org	4,635,346	en	83.3%
.xyz	1,303,609	en	82.8%
.br	1,324,697	pt	82.4%

Embed this figure

<a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-language"><img src="https://content.stackscan.com/charts/domain-tld-statistics-language.webp" alt="One extension, one language" width="880" style="max-width:100%"></a> <p>Source: <a href="https://www.stackscan.com/blog/domain-tld-statistics#fig-language">StackScan TLD analysis</a></p>

.au declares English on 99.7% of the sites that declare a language at all. .nz and .za and .uk are all at or above ninety nine per cent.

The share is of declarations, not of domains. An extension where most sites declare nothing would otherwise read as having no language, and 25.6% of the web declares nothing.

At the other end the multilingual countries show as multilingual: .no is split across the two written forms of Norwegian, .ch across German and French and Italian, .be across Dutch and French.

Country extensions concentrate by declared language about as sharply as they do by jurisdiction, and the ones that do not are the countries that were never monolingual to begin with. The language report takes the same field across every website rather than across extensions.

What TLD statistics do not tell you

A registration count is a poor measure of an extension's size. An extension with ten million registrations can hold fewer working websites than one with three million, and two of the largest extensions on the internet do exactly that: .xyz has 11,337,202 registrations against .it's 2,463,213, and .it has more working websites.

Three separate questions hide inside the single figure registries publish.

Registration share asks how many names exist. It is the number everyone quotes and the one an extension can buy with a discount.

Active share asks how many of them are a website. That is 36.8% of the internet, and it moves from 6.8% to 61.7% depending on whether a registry checks who is registering.

Technology share asks how much of the working web actually runs on them. A working .it site carries 19 technologies and a working .top site carries 5, so two extensions with the same number of live sites are not the same amount of internet.

An extension is sold by the first number and used by the third. Anyone choosing a domain, valuing a portfolio or sizing a market from registration counts alone is reading the one measure that the seller controls.

Frequently asked questions

How many domains are registered on the internet? 399,851,676 across 1,570 extensions, on the day this report was built. 150,726,446 of them answer a request and 146,990,668 serve a working website. The figure moves every day as names are registered and dropped. The extension count is ours rather than the root's: IANA delegates about 1,438 strings, and this report counts every distinct ending it finds names under, which includes retired brand extensions whose names are still in zone data and a handful of malformed strings that reached a zone file. They hold a few thousand names between them, so the totals barely move, but the count of extensions is not a count of the root.

Does a cheap domain extension mean fewer real websites? Not the price itself, the discount. Extensions that renew at five times their first year serve a working website on 24.3% of their registrations, against 39.0% for extensions that renew at what they cost to buy. Compare cheap and dear extensions that have no first year discount and the difference is half a point, 38.4% against 39.1%.

What is a domain zone, and is it the same as a TLD? Yes. A zone is what a registry publishes for its extension: the file listing every name registered under it and the nameservers each one points at. .com the extension and the .com zone are the same set of names. This report counts 399,851,676 domains across 1,570 zones, and the registered figure is a zone count rather than a website count, which is the distinction the rest of the report is about.

What does it mean for a domain to resolve? That the name has an A or AAAA record, so a DNS lookup for it returns an IP address. Resolving is not serving. Of the 292,821,718 domains that resolve, 150,726,446 return a page when one is asked for; the rest have an address with nothing answering on it.

What percentage of domains are actually used? 36.8%. 62.3% return no page at all, and a further 2.5% of the ones that do answer only to offer the name for sale.

What is the most popular domain extension? .com, with 196,616,562 registrations, 49.2% of every domain on the internet. .de is second with 16,373,955.

How many TLDs are there? IANA delegates about 1,438. This report finds names under 1,570 distinct endings, the difference being retired extensions whose names linger in zone data and malformed strings, and 40 endings hold more than a million names.

Which domain extension is the most used? .za, a working website on 61.7% of its registrations, of the extensions with at least half a million. The extensions at the top all restrict who may register.

Which domain extension is the emptiest? .bond, a working website on 6.8% of its registrations, of the extensions with more than half a million domains.

How many domains are parked? 31,975,287 registrations carry a parking or domain-sale signature, 8.0% of everything registered. Only 3,735,778 of them return a page when asked; the rest point at a parking service and render nothing. This is a floor, because it counts only the platforms and signatures we detect by name.

Does a cheaper extension get used less? Not measurably. The extensions that get used are national ones bought by their own market, and some of those restrict registration while others do not; the first year discount tracks use more closely than either the price or the registration policy does.

Methodology and sources

The population is registered domains, not websites, and this is the only report on this site that uses it. 399,851,676 domains across 1,570 extensions, taken from the same zone data the website lists are built from. Every share in this report is a share of that number unless the figure's subtitle says otherwise.

A registration is a name at the apex: two labels, or three where the middle one is a second level from the Public Suffix List. A host under a registration is not a registration, which is why .edu counts 17,664 names rather than the six times as many its universities run.

A domain resolves when it has an A or AAAA record. Nameserver and mail records do not make a name resolvable on their own, because an address is what a browser needs.

The DNS scan reads the registered name itself, not its subdomains: two labels, or three where the middle one is a second level from the Public Suffix List, so www.example.com and mail.example.com count as example.com and nothing is counted twice.

A domain is live when a request to it returns a page. 150,726,446 domains are live, and that set is the population every other StackScan report calls the web. Resolving and living are separate measurements, and the first does not imply the second: 113,856,262 domains have an address and return no page at all.

A domain is parked when it carries one of the parking or domain-sale signatures detected by name, among them Afternic, Sedo, Atom, Dan, Bodis, ParkingCrew, GoDaddy's NameFind and the registrar parking services.

Where the provider figure names GoDaddy it is the IP the domain resolves to, verified by hand: those addresses answer with a redirect to a for-sale lander on forsale.godaddy.com and carry GoDaddy nameservers. Domains on a parking IP we could not attribute are counted as parked and listed as not attributed rather than assigned to a guess.

Parked is not a synonym for listed. Some of those platforms only sell, some only monetise with advertising, and the largest do both from the same page, so the signal identifies the platform answering for the name and not the holder's intention.

The report uses parked and held for sale interchangeably for that combined population, and gives the count that actually serves a sale page wherever the difference changes the reading.

Two passes produce it. The first reads technologies off a page we stored. The second is the crawler's verdict at fetch time on domains whose page was never stored: a redirect to a for-sale lander, a parking nameserver, a parking IP.

They overlap on 1,030,110 domains only, because a domain that bounces a crawler to a lander leaves no page for the first pass to read.

Together: 31,975,287 registrations are parked and 3,735,778 of them return a page. A domain on a registrar's default page with no detectable signature counts as unparked, so the parked figure is a floor and the working-website figure is generous to the registrars.

Working website means live and not parked. It does not mean the site is maintained, and it does not mean anyone visits it.

Language shares are shares of the sites on that extension that declare a language, never of its domains. 25.6% of the web declares nothing, and folding those back in would treat silence as an answer.

Second-level structure such as co.uk is read from the ICANN section of the Public Suffix List, and a domain is counted under its extension, so bbc.co.uk counts once under .uk.

Prices come from Dynadot's published price list in US dollars, 625 extensions with a first year and a renewal price, taken as a single snapshot.

The renewal is the price used, because a first year is a promotion on most extensions and is not what a holder pays. Registrars discount and promotions move, so the price sections read in bands and never quote a price as the price. The bands cover 108 of the extensions measured here, 91.2% of everything registered.