We counted 8,741 reported domain sales. One-word names sold for 15× everything else.
Most advice about picking a domain name is written from taste. We wanted numbers, so we counted them ourselves.
How we counted
We read every bi-weekly sales report DNJournal published between February 2025 and September 2026 — 42 reports, 8,741 reported sales. DNJournal is a public record of domain sales that brokers and marketplaces choose to disclose. We aggregated it; we are not republishing it.
Then we ran each name through the same parser our own naming tool uses. It splits a name into pieces and marks which pieces are real dictionary words. That gives us a shape for every name, not a guess:
- one word — the whole name is a single dictionary word (
little) - coined — no dictionary word at all (
zorvanta) - word + fragment — one real word plus an invented tail (
brandbada) - two words — two dictionary words stacked (
brewoak) - many words — three or more (
smartoriginlab) - digits / hyphens — self-explanatory
We looked only at .com sales to keep the comparison clean: 5,126 of them.
What the numbers said
| Shape | Share of .com sales | Median price |
|---|---|---|
| One word | 2% | $55,000 |
| Coined | 21% | $3,900 |
| Word + fragment | 38% | $3,499 |
| Two words | 30% | $3,500 |
| Digits | 4% | $3,500 |
| Hyphenated | 2% | $2,900 |
| Many words | 3% | $2,900 |
Read that first row again. A one-word .com sold for a median of $55,000. Every other
shape landed between $2,900 and $3,900 — they are, for practical purposes, the same number.
There is no gentle slope here. There is one shape, and then there is everything else.
The concentration shows up again at the top of the market. Among sales above $100,000, 37% were single dictionary words — a shape that makes up only 2% of sales overall.
Why this is less useful than it looks
You cannot act on it.
Every single-word .com was registered decades ago. You will not find one available to
register for $11. To own one you buy it on the aftermarket, and the table above is exactly
what you will be asked to pay.
So the honest reading is not "go get a one-word domain." It is this: if you are choosing between shapes you can actually register, the price data says the choice barely matters. Coined, word-plus-fragment, two-word — $3,499, $3,900, $3,500. Pick the one that sounds right for your business and stop agonizing.
The shapes that genuinely cost you are hyphens and digits, at $2,900 and with the smallest share of sales. Those are the two to avoid.
What we could not see
Three limits worth stating plainly, because most articles about this data do not state any.
DNJournal only records what someone reported. Private sales are missing. More importantly, unsold names are missing entirely — so nothing here tells you the odds that a given name sells at all. The median price of sold names is not the expected value of owning one.
Three quarters of the sales have no venue attached. DNJournal does not list where the sale happened for most rows. Of the ones that do, Sedo dominates with 1,084 sales at a median of $11,700. Curated marketplaces barely appear — which means their prices are simply not knowable from public data, whatever anyone tells you.
The shape parser depends on a dictionary. A word it does not know gets marked as a fragment. We checked a sample by hand, but the long tail will have errors.
The extension mattered more than the name
One last table, because it surprised us more than anything above:
| Extension | Median price |
|---|---|
| .ai | $45,410 |
| .com | $3,500 |
| .de | $2,900 |
| .nl | $1,450 |
The gap between .ai and .com is larger than the gap between any two name shapes except
single words. In 2025–2026, what came after the dot moved the price more than what came
before it.
We expect that to compress as the AI naming rush cools. We will re-run this in six months and say whether it did.
Method: 42 DNJournal bi-weekly reports, Feb 2025 – Sep 2026, 8,741 reported sales, 5,126
of them .com. Shape classification by our own parser. Medians, not means — the top of
this market is skewed enough that averages mislead.
Tags