We Checked 200 Startup Domains in One Session. Only 5 Were Free.

What exactly did we test?

We wanted a sample that reflects how founders actually name things, not a list of random gibberish that is available by definition. So we hand-built 50 root words across the three buckets people really pull from: plain dictionary words (harbor, summit, beacon, cobalt, meridian, velocity, fathom), obvious two-word compounds (payflow, cloudnest, datavault, launchpad, swiftpay, openledger), and coined-but-pronounceable brand words (zephyra, novalink, aetherly, mintware, embergrid).

Each of the 50 roots was checked against four extensions — .com, .io, .ai and .co — for 200 total lookups. Every check was a live availability query run this session, not a cached list. When a domain resolved to active nameservers we counted it as taken; only genuinely unregistered names counted as available.

• Sample: 50 brandable root words
• Extensions: .com, .io, .ai, .co
• Total domains checked: 200
• Method: live availability lookups, this session

How many were actually available?

Five. Out of 200 domains, exactly five came back open — an availability rate of 2.5%. Put differently, 195 of the names a founder might realistically want were already gone.

Even more telling: of the 50 root words, only 3 had a single extension free anywhere. The other 47 were fully registered across all four TLDs. If your shortlist looks like a normal founder's shortlist, the base-rate expectation in 2026 is that your favorite name is taken on every major extension simultaneously.

Which extension had the most open names?

Here is the per-extension breakdown from the run. The differences are real but small — and none of them are a rescue.

The headline everyone expects held up: .com was a wall. Zero of our 50 candidates were available on .com. The much-hyped alternatives did marginally better, with .ai leading, but 'marginally better than zero' is still a rounding error.

• .com — 0 of 50 available (0%)
• .io — 1 of 50 available (2%)
• .ai — 3 of 50 available (6%)
• .co — 1 of 50 available (2%)

So is .ai the answer to the .com squeeze?

Not really — and this is the part the data made unavoidable. Switching from .com to .ai took our availability from 0% to 6%. That is a better lottery ticket, but it is still a lottery ticket. The founders who believe 'I'll just grab the .ai' are quietly discovering the same crunch that hit .com a decade ago, because the alt-TLDs are absorbing the exact same demand for the exact same desirable words.

The real lesson is that the extension is not the variable that decides whether you can get a domain. In our run, the same root word was almost always taken on all four extensions or free on some — availability tracked the word, not the TLD. Fighting over .com versus .io versus .ai is optimizing the wrong dimension. The dimension that actually moved the number was how novel the word itself was.

What kind of names actually survived?

Every single dictionary word and obvious compound in our sample was taken everywhere. Harbor, summit, beacon, payflow, cloudnest, datavault — gone on all four extensions, no exceptions. That is the entire 'obvious good name' category, wiped out.

The only three roots with any availability were invented words: mintware, silverstack and embergrid. And even they mostly survived on the alternative extensions rather than .com. The pattern is clean enough to state as a rule: in 2026, the price of a registrable domain is a word nobody else thought of yet. Realness and obviousness are exactly what the market has already exhausted.

Availability can change by the hour as names get registered, so treat those three as a snapshot from this session, not a reservation. The point is not the specific words — it is the shape of what is left.

What should you actually do with this?

If 195 of 200 realistic names are gone, the winning move is not to check harder — it is to name differently. The data points to a few concrete strategies that put you in the 2.5% instead of the 97.5%.

Start from invented words, not dictionary words. Coin a root, add a soft suffix, blend two half-words. The names that survived our run were the ones a thesaurus would never produce. And once you have a coined shortlist, check it in bulk across extensions in one pass rather than one registrar tab at a time — the whole point of our run was that testing 200 at once turns a week of guessing into a single session.

• Prefer coined and blended words over real words — that is where the open space is.
• Assume your favorite real-word name is taken on every major TLD, and check it early before you fall in love.
• Don't treat .ai or .io as an escape hatch from .com scarcity; they face the same demand.
• Check a whole shortlist across extensions in one batch, not one lookup at a time.
• When a coined name is free on .com, take it — that combination is now the exception, not the default.

How honest is this number?

A single 200-domain run is a snapshot, not a census, so here are the caveats we would want a reader to keep. The sample is 50 names we chose to resemble a realistic founder shortlist, which deliberately skews toward names people actually want — that is the population we cared about, but it means the 2.5% figure describes desirable names, not the entire alphabet of possible strings. Nonsense strings would score far higher and mean far less.

Availability here means unregistered, not for sale. Some of the 195 taken domains may be parked and purchasable on the aftermarket for a price — but 'available at standard registration' is the only version of available that matters to a bootstrapped founder, and that is what we measured. Numbers reflect the state of the world during this session and will drift as registrations change.

With those caveats stated plainly, the direction of the finding is hard to argue with: for the kind of name you actually want, the modern namespace is close to fully claimed, and the extension you pick is a footnote next to the word you invent.