Are all the good .com domains taken?

Short answer: no, but that is not the useful part.

We built 520 two-word .com domains out of ten word lists and checked every one against the registries. Availability did not come out anywhere near evenly. Some words produced nothing at all, every combination registered. Other words produced a free domain every single time. And which words landed in which group is not random enough to be interesting as luck.

The words everyone reaches for are gone where you would put them

Take new. Paired in front of fifteen different nouns, newledger.com and the fourteen like it, every single one is registered. Not most. All fifteen.

Now flip it. ledgernew.com and its fourteen siblings: twelve of them are free.

The flip is the finding. new is not a used-up word in some general sense. It is used up in front, because in front is where everyone puts it, and ledgernew.com survives only because nobody wants to be called that.

So the honest version of "all the good ones are taken" is narrower and more useful: the good ones are taken in the shape you would build them. Change the shape and the same word comes back.

Some words are gone in both directions

A handful do not come back. digital returned nothing in either position, 0 out of 30. That is a word that has been the obvious modifier for twenty-five years in every industry at once, and there is no angle left on it.

At the other end, daemon, glacier, waterproofing and refined are open almost everywhere we put them, leading or trailing, better than 90%.

What we are actually comparing

Worth being plain about the method, because it decides how much of this you should believe.

Every domain here is an arbitrary two-word combination. newledger.com is not a name anyone set out to own, and neither is daemonledger.com. That is the point. Both are equally unwanted in the abstract, they are the same shape and roughly the same length, and one of them is registered and the other is not. The only variable left is how many people have already worked through combinations containing that word.

So read every number here as relative demand for a word, not as a shortlist. daemonledger.com being free does not mean it is a good name. It means nobody has been down that row yet, which is a different and more useful thing to know before you start.

That also means the comparison breaks when the two sides are not the same shape. We tested northdakotaactive.com, northdakotaelite.com and three more like them, and all five are unregistered, against four of five taken for Georgia. Georgia is plainly more contested. But North Dakota did not come out "open" in any sense you can use: nobody wants a 21-character state adjective domain, which is most of why they are sitting there. Where the two words are the same length and the same kind, the comparison holds up better:

Every word below was crossed with the same fifteen nouns: instinct, remix, anchor, cornerstone, vault, canvas, pivot, totem, dominion, chart, anvil, lens, registry, tome, relic. So security means securityinstinct.com, securityremix.com and thirteen more like them.

Read one row across. securityinstinct.com and securityvault.com are registered; waterproofinginstinct.com and waterproofingvault.com are not. Same second word, same construction, same day. pixelinstinct.com is gone and daemoninstinct.com is free.

The exceptions are worth noticing too. daemon misses one of its fifteen, and the one it misses is daemonvault.com — which is also the one vault takes from every other word we tried. Contested partners exist as well as contested words.

That is the whole argument in one line: the lucrative trade is worked through and the unglamorous one is not, and nobody has been down the daemon row at all. None of these are names anyone was hunting for. The difference is only how many people got as far as that word.

The one number that actually predicts availability

It is not the extension, and it is not the length. Our biggest pattern generates 4,200 combinations and our smallest 2,430, and they came out at 70% and 13% available, which is the wrong way round if size were driving anything.

What predicts it is how ordinary the word is, and how ordinary the arrangement is. Ordinary has been ordinary for a long time, and somebody registered it in 2003.

What this is and isn't

An earlier version of this piece said seven words produced no available domain at all. That was measured against five partners each, and three of the seven did not survive a wider test: vision went from 0/5 to 5/15 on the same test, which makes the zero a draw rather than a property. Everything above is fifteen partners in each position, and the words that held up held up completely.

Fifteen is still a sample. Read a single row as a strong hint about that word and the table as a whole as the actual finding.

This is .com only. Other extensions are less picked over and behave differently, and we have not measured that.

So what do you actually do

Assume your first instinct is the exhausted one, and that the arrangement you put it in is exhausted too. The words that occur to you immediately occurred to everyone else immediately, in the same order.

So there are two levers, not one. Change the word: swap digital for something with a texture to it, and drop my entirely. Or keep the word and change where it sits, which costs you nothing and, for new and your and pixel, moves you from nothing available to most of them.

Then run the combinations rather than guessing which survived, because 4,200 of them take about as long to check as one, and nothing above would have been predictable from the outside.