New gTLD Ghost Towns: Where a Third of Domains Are Dead
Published
Figures as of 2026-08-16 · methodology v9. This is a recurring report; each edition re-measures the same population so the numbers can be tracked over time. All figures are aggregate — we never publish an individual business’s grade or name an individual registrant’s domain.
The headline: 36% of new-gTLD domains are dead or unreachable
Of the 63,782,045 domains we scanned across 985 new generic TLDs in the August 2026 census round, 22,935,111 — 36.0% — were dead or unreachable. That is 2.6 times the rate in the legacy gTLDs (.com, .net, .org and their pre-2013 peers, at 14.0%) and nearly 4 times the ccTLD rate (9.2%). In the worst case, .bond, 87.3% of the namespace answers to nobody.
This is not a story about bad security. A dead domain has no security to grade. This is a story about aliveness: how much of what the post-2013 TLD expansion sold is actually running anything at all. The answer, for a third of it, is nothing.
Key numbers
- 36.0% of new-gTLD domains are dead or unreachable — 22,935,111 of 63,782,045 scanned in the August 2026 census round.
- The legacy gTLDs sit at 14.0% dead or unreachable (207.5M domains scanned); ccTLDs at 9.2% (88.2M scanned). New gTLDs run 2.6× and 3.9× those rates.
- .bond is the emptiest sizeable new gTLD: 87.3% ghost — 1,020,165 of its 1,168,294 domains dead or unreachable.
- .top contributes 38.3% of the entire new-gTLD ghost mass on its own: 8,780,287 dead or unreachable domains out of 12,548,818 scanned (70.0%).
- 29 of the 258 new gTLDs with at least 10,000 domains are majority-dead — more of the namespace gone than present.
- Only 58.4% of new-gTLD domains were alive enough to receive a security grade, versus 80.9% in the legacy gTLDs.
What counts as a ghost domain?
Our census scans every domain in its inventory and assigns a disposition before it assigns a grade. Two dispositions make up the ghost count in this article:
Dead means the domain shows no signs of life at the DNS layer — the delegation exists in name only. Unreachable means DNS gave us something but nothing answered when we knocked: no usable web or service response at capture time.
We deliberately exclude a third disposition, indeterminate (partial or ambiguous evidence — 3,587,410 new-gTLD domains, 5.6%), from the ghost count. Some of those are alive and awkward, some are dying. Counting them as ghosts would inflate the story, so we don’t.
Everything else — 37,259,524 domains, 58.4% — was alive enough to grade. For comparison, 80.9% of legacy-gTLD domains and 79.5% of ccTLD domains graded. The new namespace is not just deader at the bottom; it is thinner at the top.
One distinction matters here. Our sibling article on grade decay by TLD covers domains that are alive but let their security rot. This article is about domains that are simply not there. Different failure, different economics.
Which new gTLDs are the emptiest?
Among the 258 new gTLDs with at least 10,000 scanned domains, these are the ten with the highest share of dead or unreachable names:
| TLD | Domains scanned | Dead + unreachable | Ghost share |
|---|---|---|---|
| .bond | 1,168,294 | 1,020,165 | 87.3% |
| .skin | 170,118 | 141,276 | 83.0% |
| .garden | 248,726 | 196,675 | 79.1% |
| .yachts | 53,799 | 41,788 | 77.7% |
| .ren | 24,253 | 17,373 | 71.6% |
| .autos | 214,720 | 152,030 | 70.8% |
| .wang | 87,911 | 61,687 | 70.2% |
| .top | 12,548,818 | 8,780,287 | 70.0% |
| .luxe | 11,565 | 7,724 | 66.8% |
| .lol | 1,055,345 | 704,275 | 66.7% |
Two rows carry the story. .bond is the highest percentage: nearly nine of every ten names in a million-domain TLD lead nowhere. .top is the mass: at 12.5 million scanned domains it is by far the largest new gTLD in our census, and 70% of it is ghost. Those 8.78 million dead .top names are 38.3% of the entire new-gTLD ghost population by themselves. Strip .top out and the remaining new gTLDs still run 27.6% dead or unreachable — better, but still double the legacy rate.
The pattern across the table is registration volume without operational intent. Domains get registered in bulk — speculation, promotions, campaigns that end — and then nothing is ever built, or what was built is switched off and the name keeps renewing, or lapses into a zombie delegation. As industry context (not census data): several of the TLDs on this list have run aggressive first-year promotional pricing, which pulls in exactly the kind of registration that never intends to host anything. Our census measures the aftermath, not the pricing; our sibling article on cheap TLDs takes up the price side directly.
Why are so many new-gTLD domains dead?
The honest answer from the data is: because they were never alive. A ghost share of 70-87% in a TLD is not decay from a healthy peak. Healthy namespaces don’t lose seven-eighths of themselves. These numbers describe namespaces where most registrations were disposable from the day they were created — bulk speculation, throwaway campaign names, defensive registrations nobody wired up.
Industry context, labelled as such: the new-gTLD programme (ICANN’s 2012 application round, first delegations in late 2013) created hundreds of registries that live or die on registration volume. Volume is easy to manufacture with low prices; usage is not. Twelve years on, the census reads out the difference between the two.
A structural note in the ghost mass itself: of the 22.9 million ghosts, 11,513,834 were fully dead and 11,421,277 unreachable — an almost even split between names that are gone at the DNS layer and names that still resolve to something that no longer answers. The second group is arguably worse hygiene: infrastructure that was stood up and abandoned rather than never built.
Two caveats before anyone quotes the table. First, “unreachable” is measured from our EU vantage at capture time in one census round; a domain that blocks scanners, or was mid-outage that day, lands in the same bucket as one abandoned in 2019. At the scale of millions of domains per TLD those cases are noise, but they exist. Second, this is aliveness, not judgement of the TLDs’ operators — a registry sells names; what buyers do with them is the buyers’ business.
Are all new gTLDs ghost towns?
No, and the counterexamples are the most useful part of the dataset. Same programme, same era, wildly different outcomes:
| TLD | Domains scanned | Ghost share | Graded (alive) share |
|---|---|---|---|
| .realtor | 23,847 | 4.5% | 95.0% |
| .srl | 12,107 | 4.3% | 92.8% |
| .realestate | 13,676 | 5.6% | 92.7% |
| .berlin | 43,309 | 5.6% | 92.6% |
| .hamburg | 15,562 | 5.8% | 92.4% |
| .digital | 415,462 | 7.1% | 84.2% |
| .xyz | 9,078,617 | 9.7% | 85.1% |
| .com (legacy, for scale) | 172,325,990 | 13.4% | 81.7% |
The healthy new gTLDs cluster into two types: restricted or community TLDs where registration implies a real entity (.realtor, .srl, city TLDs like .berlin and .hamburg), and general TLDs that found genuine users (.digital). When the buyer is a real business registering its own name, the domain gets used.
The row that breaks the lazy narrative is .xyz: 9.1 million domains, famously cheap, and a ghost share of 9.7% — lower than .com’s 13.4%. Cheap does not automatically mean dead. .xyz and .top sit at similar price points (industry context, not census data) and similar scale, yet one is 70% ghost and the other beats the flagship legacy TLD on aliveness. Whatever separates them, it is not the sticker price alone.
The spread among the 258 new gTLDs with ≥10,000 domains: 24 sit under 10% ghost, 164 between 10% and 25%, 41 between 25% and 50%, and 29 are majority-dead. The median TLD runs 18.8% — the catastrophic aggregate is driven by a heavy tail of very large, very empty namespaces, .top above all.
One oddity worth flagging for anyone reusing this data: .feedback shows almost no ghosts (0.9% dead or unreachable) but only 3.7% graded — 95.3% of its 14,691 names landed in the indeterminate bucket. Every name responds, almost nothing behaves like an operated domain. We report it as measured and draw no conclusion about the registry’s configuration here, beyond noting that it is exactly the kind of TLD-wide pattern our disposition logic exists to keep out of both the dead and the alive columns.
Does this change how big the “domain industry” really is?
For the new gTLDs, yes, materially. Headline registration counts for this part of the namespace describe names sold, not names used. On our figures, an aggregate registration statistic for the new gTLDs overstates the operationally live namespace by roughly 70%: 63.8 million scanned, 37.3 million alive enough to grade.
Unlike the phantom-domain problem in some ccTLDs (see our report on wildcarded TLDs), these ghosts are real registrations. New gTLDs publish zone files, and their inventories are auditable in a way passive-DNS-derived ccTLD counts are not — as industry context, most new-gTLD zones are available through ICANN’s CZDS programme. Someone paid for these names. They exist in the registry database. They just don’t do anything. That makes new-gTLD ghost towns the opposite failure to phantom TLDs: there, the counts were inflated by names nobody registered; here, the counts are accurate and the usage is the mirage.
For anyone citing domain statistics: a registration count for a new gTLD is a sales figure. If the claim is about the internet people actually use, the honest denominator for this namespace is roughly 58% of the headline number, and for individual TLDs it can be as low as 11%.
How we measured this
- Population: the August 2026 census round — figures as of 2026-08-16, methodology v9. The full round scanned 432 million domains; this article uses the 63,782,045 scanned domains in TLDs we classify as new gTLDs.
- Classification: a TLD counted as a “new gTLD” when its ASCII label is three or more characters and it is not one of the 23 pre-2013 legacy gTLDs (.com, .net, .org, .info, .biz, .mobi, .name, .pro, .asia, .cat, .jobs, .tel, .travel, .xxx, .aero, .coop, .museum, .post, .edu, .gov, .mil, .int, .arpa). That yields 985 TLDs. Internationalised (xn—) TLDs were excluded from the new-gTLD cohort to avoid mixing IDN ccTLDs into it; a small number of IDN new gTLDs are therefore not counted, which makes our cohort slightly conservative.
- Ghost definition: dead + unreachable dispositions at capture time. Dead = no meaningful DNS-layer life; unreachable = resolution without a usable service response. Indeterminate (5.6% of the cohort) is excluded from the ghost count.
- Denominators: all percentages in this article use domains scanned per cohort as the denominator, stated inline. Cohort totals: new gTLDs 63,782,045; legacy gTLDs 207,510,432; ccTLDs 88,160,050.
- Vantage and timing: single EU vantage, single census round. A domain blocking our scanner or suffering an outage at capture time lands in unreachable; at TLD scale this is noise, but per-domain it is not a verdict.
- Aliveness, not decay: this report measures whether domains respond at all. It says nothing about the security posture of the live ones — that is the grade data, covered in the sibling articles.
- Inventory caveat: gTLD inventories derive from our census source inventory. Because new gTLDs publish zone files, registration-count inflation of the kind we documented for wildcarded ccTLDs is unlikely here, but we have not reconciled every TLD against its registry zone file for this edition.
- Registries not named: we identify namespaces by TLD string only and make no claims about the conduct of any registry operator. High ghost share is a property of what buyers did with the names, not an accusation against whoever sold them.
- Industry context (pricing history, ICANN programme dates, CZDS availability) is labelled as such throughout and is not census data.
- Aggregate only. We never name, grade or publish data about an individual registrant’s domain. Data is stored and processed within the EU.
FAQ
What percentage of new gTLD domains are actually in use? In our August 2026 census round, 58.4% of the 63.8 million new-gTLD domains scanned were alive enough to receive a security grade. 36.0% were dead or unreachable, and 5.6% returned indeterminate evidence. By comparison, 80.9% of legacy-gTLD domains graded.
Which new gTLD has the most dead domains? By share, .bond: 87.3% of its 1,168,294 scanned domains were dead or unreachable. By raw mass, .top: 8,780,287 ghost domains — 70.0% of its namespace and 38.3% of all new-gTLD ghosts in the census on its own.
Are cheap TLDs always full of dead domains? No. .xyz, one of the most aggressively discounted TLDs (industry context), showed a 9.7% ghost share across 9.1 million domains — lower than .com’s 13.4%. Meanwhile .top, at a similar price point, ran 70.0%. Price correlates with disposable registrations in some TLDs but does not determine the outcome.
Does a dead domain matter for security? Not directly — nothing is running, so there is nothing to attack in the usual sense. It matters for measurement: any statistic that treats registrations as active internet presence overstates the new-gTLD web by a wide margin, and lapsed-but-delegated names are the raw material for takeover and re-registration abuse. Our census grades only the live population.
Is a “ghost town” TLD a bad choice for my own domain? The ghost share describes a TLD’s population, not your domain. A well-configured site on a 70%-ghost TLD is exactly as secure as the same site anywhere else. The practical considerations are reputation and filtering: some security tooling treats high-abuse or high-ghost namespaces with more suspicion (industry context, not a census finding).
What This Means
For IT managers and business owners choosing a domain, the ghost-town data provides a concrete way to think about TLD selection beyond price and availability. A TLD where 70-87% of the namespace is dead or empty is, by definition, dominated by non-operational registrations. Security tooling, email filtering, and browser reputation systems increasingly weight the TLD as a signal when they have no history for a specific domain. A new domain on a heavily ghost TLD starts its life in a neighbourhood that automated systems treat with more suspicion — not because of anything you have done, but because of the company you are statistically keeping.
For security teams tracking phishing and brand impersonation, the new-gTLD ghost-town pattern has a direct operational implication. Ghost TLDs with high dead rates are also TLDs where domain registration is cheap and buyer intent is frequently non-operational. That combination — cheap, disposable registrations, reputation damage already priced in by the namespace — is the exact profile of infrastructure used for short-lived phishing campaigns. Monitoring for your brand name across high-ghost new gTLDs is a proportionally more valuable use of threat intelligence budget than monitoring in European ccTLDs where nearly every registration has operational intent behind it.
For anyone writing about the “size” of the domain industry, the new-gTLD data is a direct correction to registration-count narratives. A headline count for a new gTLD is a sales figure, not a measure of internet presence. The honest denominator for “domains people actually use” is roughly 58% of the new-gTLD registration total — and for individual TLDs like .bond and .top, it is far lower than that.
Data to Cite
- “36.0% of 63,782,045 new-gTLD domains scanned in August 2026 were dead or unreachable — 2.6 times the legacy-gTLD rate of 14.0% and 3.9 times the ccTLD rate of 9.2%.” — defaults.exposed August 2026 Domain Security Census (432M domains)
- “Only 58.4% of new-gTLD domains were alive enough to receive a security grade in the August 2026 census, versus 80.9% of legacy-gTLD domains.” — defaults.exposed August 2026 Domain Security Census (432M domains)
- “.bond is the emptiest sizeable new gTLD: 87.3% of its 1,168,294 scanned domains were dead or unreachable.” — defaults.exposed August 2026 Domain Security Census (432M domains)
- “.top alone contributes 38.3% of the entire new-gTLD ghost mass: 8,780,287 dead or unreachable domains from 12,548,818 scanned — 70.0% of the namespace.” — defaults.exposed August 2026 Domain Security Census (432M domains)
- “.xyz, one of the most aggressively discounted TLDs, showed a 9.7% ghost share — lower than .com’s 13.4% — demonstrating that cheap price does not automatically produce dead domains.” — defaults.exposed August 2026 Domain Security Census (432M domains)
- “A registration count for a new gTLD is a sales figure: the live new-gTLD namespace is roughly 58% of the headline number, and for individual TLDs like .bond it is as low as 11%.” — defaults.exposed August 2026 Domain Security Census (432M domains)
See where your own domain stands
Ghost domains can’t be secured, but yours can. Our census grades the live internet — more than 376 million domains — across externally observable security checks, and most of what a failing domain is missing is free and quick to fix. The barrier is almost never cost; it’s that nobody told the owner it mattered. You can check your domain privately and free, and see exactly which checks you pass.
Read the flagship census report: The State of Domain Security 2026 →
Related: New gTLDs vs legacy domains: who does security better? · Domain decay by TLD: where security rots fastest
Aggregate data only. Data stored and processed in the EU.
Check your domain free at defaults.exposed — see your domain’s full security grade regardless of which TLD it is in, and find out how it compares to the live domains across the namespace. Takes 30 seconds. No account needed.
How to cite this report
Press / blog: defaults.exposed (2026). New gTLD Ghost Towns: Where a Third of Domains Are Dead. defaults.exposed August 2026 Domain Security Census (432,127,908 domains scanned, asOf 2026-08-16). Retrieved from https://defaults.exposed/en/articles/new-gtld-ghost-towns
Academic: defaults.exposed. (2026, August 18). New gTLD Ghost Towns: Where a Third of Domains Are Dead. In defaults.exposed Domain Security Census: August 2026. https://defaults.exposed/en/articles/new-gtld-ghost-towns
In-line citation: (defaults.exposed, August 2026 Domain Security Census, n=63,782,045)
About the defaults.exposed August 2026 Census
The defaults.exposed Domain Security Census is a recurring independent measurement of the public domain namespace. The August 2026 edition scanned 432,127,908 domains between 1–16 August 2026 and graded 376,928,781 of them using methodology v9. Scans are conducted from EU infrastructure. No individual domain, registrant, or business is named in any report. All figures are aggregate distributions. Data is stored and processed within the EU.
Methodology: defaults.exposed/en/articles/domain-security-scoring-methodology-v9 Full census report: defaults.exposed/en/articles/the-state-of-domain-security-2026