How to come up with a domain name that is good and still available.
A domain name generator turns a few keywords, a sentence about your business, or a language into names you can actually register. This guide explains the three ways it works, what makes a name score well, and how to pick one that is easy to own.
Updated September 202610 minute readSources: PNAS, USPTO
- ways to a name: keywords, a sentence, real words
- 3ways to a name: keywords, a sentence, real words
- patterns for building a name
- 10patterns for building a name
- scores on every generated name
- 5scores on every generated name
- real words across 8 languages
- 1,415,427real words across 8 languages
How do you come up with a domain name?
Start from what you have. With a few keywords, the algorithmic generator builds and scores hundreds of names. With a description of the business, AI mode works from the whole idea. And if you want a real word, the word browser searches eight languages for short ones that are still free.
| If | Start with |
|---|---|
| You have a few keywords | Algorithmic |
| You can describe the business | AI |
| You want a real word, not a coinage | Words |
| You want scores you can inspect | Algorithmic |
| You want the widest checked set | Algorithmic |
| You want a large set to scan fast | Words |
All three share one shortlist, so you can move between them without losing work. For more approaches, read how to come up with a domain name when everything is taken.
How does the domain name generator work?
It treats naming as a search problem. Your keywords are expanded into related concepts, hundreds of candidates are built from ten naming patterns, each is scored, the set is spread for variety, and every name is checked live in the extensions you picked.
01 Read
Keywords are separated from filler and assessed for how much material they give the generator.
02 Expand
Each keyword pulls in semantically related concepts, widening the pool well beyond what you typed.
03 Build
Candidates are constructed from that pool using ten distinct naming patterns.
04 Score
Every candidate is measured on five dimensions and ranked by a weighted overall score.
05 Spread
The final set is chosen for variety, so results are not one root word repeated.
06 Check
Every name is checked live in the extensions you picked, and available domains come first.
Ten ways to build a name
One construction method produces monotonous results, because a page of names that all rhyme is not a page of choices. Here is what each pattern does with the same two keywords:
Compound
Brewnook
two words joined whole
Blend
Brook
the start of one, the end of another
Portmanteau
Nookery
a word folded into a familiar one
Prefix
Gobrew
a familiar opening
Suffix
Brewly
a familiar ending
Clip
Nookr
a word cut short
Coinage
Bruvo
an invented word that sounds right
Classical
Cafeum
a Latin or Greek form
Evocative
Hearth
a real word with the right feeling
Phrase
Thebrewnook
a short natural phrase
Why results are spread, not stacked
Ranking purely by score returns near-duplicates, because the best construction tends to win again and again with the same root. The final selection caps how often any root, opening, ending or pattern appears, and rejects names too close to one already chosen.
What makes a domain name score well?
Five things, weighted: fluency, how easily it is read and said; distinctiveness; length; spellability; and sound match. Fluency weighs most, because names that are easy to process are easier to remember and to trust.
Scoring weight
Fluency: How easily the name is read and said
Distinctiveness: How much it stands apart from the input words
Length: Fit against a practical character band
Spellability: Risk of being misheard or mistyped
Sound match: Fit between the sound and the product type
The fluency effect is well documented. In one study, shares with easy-to-pronounce names outperformed hard-to-pronounce ones shortly after listing. The psychology behind it is in what makes a good brand name.
Which names are easiest to trademark?
Invented and unrelated words. Trademark offices rank names on a spectrum from generic to fanciful: generic names can never be registered, purely descriptive ones only after they have acquired distinctiveness, and suggestive, arbitrary and invented names are distinctive from day one.
| Tier | Strength | What it is | Practical read |
|---|---|---|---|
| Generic | the ordinary word for the thing itself | Never registrable | |
| Descriptive | states what the product does | Only once it has acquired distinctiveness | |
| Suggestive | hints at the benefit indirectly | Inherently distinctive, and practical | |
| Arbitrary | a real word unrelated to the product, like Apple for computers | Strong, needs explaining | |
| Fanciful | an invented word, like Kodak | Strongest protection |
Every generated name carries its place on this spectrum, so descriptive names stand out early. Before you commit, Brand Check searches the US and EU trademark registers for the name.
How are AI domain name ideas different?
They start from a sentence rather than keywords. “Coffee” and “remote” do not tell a generator you are building a subscription for people who work from home; a sentence does. AI mode reads the whole idea, spreads each run across ten naming directions, and filters out names you cannot register before you see them.
As keywords
Three independent tokens. The generator sees what the words mean, but not how they relate, so it treats a coffee shop and a coffee subscription identically.
As a sentence
“a coffee subscription for remote workers”
- It is a subscription, not a shop
- The audience works from home
- Delivery is recurring
- The tone is everyday, not luxury
Ten directions, one run
Availability is checked first
Availability filter
How do you find a real word that is still available as a domain?
Search a dictionary by length and position instead of scrolling one. The word browser holds 1,415,427 words in eight languages, filters by exact length and by where your letters sit in the word, and sends a matching set straight to a bulk availability check.
Words indexed
1,415,427 total
Made safe for domains, without losing the original
Domain names accept a narrow character set. Rather than dropping words that do not fit, each is converted with its language’s standard convention, so German ä becomes ae and Danish ø becomes oe, and the original spelling is kept beside it. That is why 122,139 French words stay browsable. Which characters a domain may hold is explained in what characters are allowed in a domain name.
Are the generated names actually available?
They are checked, not assumed. Algorithmic names are checked in every extension you pick as soon as they exist, with available ones first; AI names are filtered before they are shown; words are checked on demand. Anything a registry did not answer is marked for a retry, never shown as free.
Each name gets one card with every extension on it, so you pick the free one you want. Regenerate for a fresh set that skips names you have seen, or mark names “More like this” to steer the next set. To check a list of your own, use the bulk checker.
Domain name generator questions
Short answers to what people ask most about generating names.
What is the difference between Algorithmic and AI domain ideas?
Algorithmic works from keywords. It expands each word into related concepts, builds candidates using ten construction patterns, and scores every result on five measured dimensions. Every name carries a score you can inspect, and each one is checked live in the extensions you selected, with available domains shown first. AI works from a full sentence describing your business, so it can respond to context that individual keywords lose, and it checks each suggestion against the extensions you selected before returning it.
Why do some generated names score higher than others?
Every candidate is scored on fluency, distinctiveness, spellability, sound match, and length. Fluency carries the most weight because names that are easy to process are easier to remember and trust. The overall score is a weighted combination, and the individual dimensions are shown on each result so you can see why a name ranked where it did.
What does the naming type on each name mean?
Each name is labelled with the kind of name it is, using the categories of the distinctiveness spectrum from trademark law: generic, descriptive, suggestive, unexpected word, and coined word. The label comes from a word-list heuristic, not a trademark search, so it says nothing about whether a mark is already taken. Use it to spot names that only describe the product, and search the relevant trademark register before committing. This is not legal advice.
Which languages does the word browser support?
English, German, French, Spanish, Italian, Dutch, Danish, and Latin, covering 1,415,427 indexed words in total. Words containing characters that are not valid in a domain name are converted to their standard domain equivalents, so German ä becomes ae and Danish ø becomes oe. The original spelling is kept alongside so you can always see what changed.
Are generated names checked for availability?
Yes, in both name modes. AI mode filters results against the extensions you choose before showing them. Algorithmic mode checks every generated name in every extension you choose as soon as the names are ready, streaming available domains to the top while the rest are verified. Word-browser results are checked on demand, so you only spend checks on the words you like.
Further reading
Sources
- Alter and Oppenheimer (2006), PNAS: processing fluency and share prices
- USPTO TMEP 1209.01: the distinctiveness continuum
Found one you like? Brand Check tests it as a website, an app and a social profile, with handles and trademarks.