We Checked the Same 150 Names on .com, .app, and .ai. Here Is Where Availability Actually Lives

Updated 2026-07-21

The short version

Registrars love to say "the .com is taken, but these extensions are available!" Nobody publishes the actual base rates. So we measured them: one fixed list of 150 brandable candidate names, 50 single real words, 50 two-word compounds, 50 coined near-words, checked against the authoritative registry for .com, .app, and .ai. No sampling tricks, the identical names on every extension, every verdict straight from the registry operator via RDAP. The results say something more interesting than "alternatives are more available," and one thing we did not expect to find at all.

The headline rates

Of 150 names checked on 2026-07-21: .com: 5.3 percent available (8 of 150, zero indeterminate) .app: 44.6 percent available (58 of 130 determinate; 20 lookups stayed indeterminate after retries and are excluded) .ai: 47.3 percent available (69 of 146 determinate; 4 excluded) For the same names, the newer extensions are roughly eight to nine times more available than .com. That is the honest version of the registrar upsell, and as far as it goes, it is true. The interesting part is where the availability is not.

Real words are gone everywhere, not just on .com

The most striking row in our data: of the 50 single real English words on the list (spark, river, harbor, ember, willow, and 45 more of the same quality), not one was available on any extension. Zero of 50 on .com. Zero of 46 determinate on .app. Zero of 50 on .ai. The common belief is that newer extensions are where you can still get the clean one-word name. Our data says the opposite: speculators and early registrants swept the dictionary on the new extensions too, because a finite list of good English words is the first thing anyone registers on any TLD. If your naming plan is "a real word, but on .ai," the registry has bad news that no amount of extension-shopping fixes. This matches what we see in the .com data at larger scale: in our 6,369-name registry dataset, brandable names of 6 or fewer characters, where most single words live, are between 0 and 3.5 percent available.

Constructed names are where the alternatives actually deliver

The availability gap between extensions lives almost entirely in constructed names: Two-word compounds (willowmend, emberline, harborlane): 10 percent available on .com, 88 percent on .app, 82 percent on .ai. Coined near-words (harbry, willovia, peachlio): 6 percent available on .com, 50 percent on .app, 61 percent on .ai. Read those .com numbers again, though, because they are the practical takeaway for most people: even on .com, one in ten well-made compounds was available, and our 150-name run surfaced eight registrable .coms, including willowmend.com and vervefield.com. The compound and coined styles are not just where .app and .ai open up, they are where .com itself is still alive. The extension question and the naming-style question turn out to be the same question, and style matters more.

The finding we did not expect: .co and .io cannot be verified at all

We planned to include .co and .io, the two most recommended .com alternatives. We could not, and the reason is worth knowing before you buy either. Every major TLD registry answers availability through RDAP, the ICANN-mandated successor to WHOIS: ask the registry about a domain, get a definitive registered-or-not answer. Verisign answers for .com. Google answers for .app. Identity Digital answers for .ai. But .co and .io are country-code domains (Colombia and the British Indian Ocean Territory), and ccTLDs are exempt from the RDAP mandate. Neither registry publishes a public RDAP service, and neither appears in IANA's official RDAP bootstrap registry. We confirmed the gap directly: RDAP aggregators return "not found" even for t.co, github.io, and stripe.io, domains that are very much registered. The practical meaning for buyers: any tool that claims registry-verified availability for .co or .io is overclaiming, because there is no registry interface to verify against; checkers fall back to registrar APIs or DNS heuristics, which is where false "available" results come from. It also means we had to apply our own standard to ourselves: our generator no longer offers .co or .io results, because we could not stand behind the word "verified" on them.

Methodology

The name list is fixed and published in the study script in our repository: 50 single real words, 50 two-word compounds, and 50 coined near-words, all 4 to 11 characters, pronounceable, hyphen- and number-free, chosen to represent the brandable space founders actually search. Each name was checked against the authoritative registry RDAP endpoint for each extension (Verisign for .com, Google Registry for .app, Identity Digital for .ai) on 2026-07-21, with retries and backoff; lookups that stayed indeterminate after three attempts are reported as unknown and excluded from the rates rather than guessed. Caveats, honestly stated: 150 names is a deliberate, style-stratified sample, not a census; the per-style splits (roughly 42 to 50 determinate names per cell) are directional rather than precise. Availability changes continuously, so these are snapshot numbers, and the run is reproducible against the live registries at any time.

Frequently asked questions

What is the most available domain extension in 2026?

Among registry-verifiable extensions in our test, .ai (47.3 percent of our 150 brandable names available) and .app (44.6 percent) were roughly eight to nine times more available than .com (5.3 percent) for identical names. But the availability concentrates in compound and coined names; single real words were 0 percent available on all three.

Are one-word domains still available on new extensions like .ai?

Effectively no. In our test, 0 of 50 quality single English words were available on .com, and 0 were available on .app or .ai either. Real dictionary words were registered early on every desirable extension. Available names on every TLD are overwhelmingly compounds and coined words.

Why were .co and .io not included in this study?

Because they cannot be measured. Both are country-code TLDs whose registries publish no public RDAP service, so there is no authoritative interface to verify availability against; aggregators return "not found" even for registered domains like t.co and github.io. Any "verified available" claim for .co or .io is built on registrar APIs or DNS guesses, not the registry.

Does this mean I should buy a .ai or .app instead of a .com?

Not by default. Availability is only one axis: .com still wins type-in traffic, trust, and resale value, which is why our verdict pieces recommend it first. The better use of this data is style: the compound and coined patterns that are 44 to 88 percent available on the alternatives are also the patterns where .com itself is still 6 to 10 percent available, which a wider search can find.

How were these availability numbers measured?

Each of the 150 names was checked directly against the authoritative registry RDAP endpoint for each extension (Verisign, Google Registry, Identity Digital) on 2026-07-21. RDAP 404 means unregistered, 200 means registered; anything indeterminate after retries was excluded rather than counted. The fixed name list and script are in our repository, so the run is reproducible.

How much of the .com space is still available overall?

Across our broader registry-verified dataset of thousands of brandable name ideas, only a small minority of .coms come back available, and they skew longer: available names average about 9 characters versus 7 for taken ones. The full numbers are in our live availability study and our length study.

By the DomainGenius team. Every verdict in this study came directly from the authoritative registry (Verisign, Google Registry, Identity Digital) via RDAP on 2026-07-21; the fixed name list and measurement script ship in our repository.

See available .com domains for your idea

Describe your idea, get brandable .com names verified available in real time.

Try DomainGenius free