Naming a startup when every dictionary word is taken.
Every good name is already taken. Every available name sounds invented. The naming problem in 2026 isn't creative — it's structural. You're picking from a shrinking pool of options that have to pass four near-incompatible filters at once.
This is the framework we use for naming work. It is not a magic generator. It is a way to think about the problem that produces fewer dead ends.
Why every good name is already taken
When the internet was small, you could name a company "Apple" or "Amazon" because dictionary words still had availability. That window closed roughly 2008–2012. After that, the gold rush on common words ended.
Today, naming a startup means picking from one of three strategies:
- Invent a new word that sounds like it could be a real word
- Modify an existing word in a way that's still pronounceable
- Compound two real words into a new combination
Each strategy has different costs. The framework here helps you pick the one that fits your situation.
Strategy 1: Invent
Companies like Lyft, Tesla, Slack, Stripe, and (yes) Egzecute are all invented or near-invented words. The benefit: trademark and domain availability are almost guaranteed. The cost: nobody knows how to spell it the first time they hear it.
The test for an invented name:
- Is it pronounceable on first read? "Tesla" passes, "Mxqz" fails.
- Does it follow the phonotactics of the target language? English allows certain consonant clusters and not others. "Stripe" is allowed, "Stkripe" is not.
- Does it sound like a real word in some recognizable register? "Lyft" sounds like "lift." That association is the work.
Invented names trade short-term recognition for long-term ownability. If you can afford a 6–12 month branding investment to make the name familiar, this is the strongest strategy. If you can't, it's the riskiest.
Strategy 2: Modify
Replace a letter, drop a vowel, swap a consonant. Kobo (from "book"), Reddit (from "read"), Tumblr (from "tumble"), Egzecute (from "execute").
The benefit: the root word stays recognizable, so the meaning carries. The cost: you're now in the territory of Google's spell-correction filter. Every search engine will assume your brand is a typo for the original word until you've built enough entity signal to override that assumption. (See Why your brand name isn't showing up on Google — this is exactly the problem we solved for ourselves.)
The test for a modified name:
- Is the modification a recognizable alteration? "Kobo" sounds like "book" backward, intentionally. "Krbq" doesn't sound like anything.
- Does it pass the spell-check test? Will Google ask "did you mean…" forever, or only at first? The answer depends on how aggressive the modification is.
- Is the original word semantically aligned? "Reddit" makes sense (read it). "Tweezer" wouldn't, for a content company.
Modified names are the most common pattern in modern startup naming because they balance recognizability with availability. They also produce the most search-engine pain in year one. Plan for that pain.
Strategy 3: Compound
Combine two real words. Airbnb (air + bnb), Facebook (face + book), Linktree (link + tree), Workspace, Salesforce, Quickbooks.
The benefit: each word carries meaning, the combination feels coined-not-invented, and search engines parse the components correctly.
The cost: most good compounds are also taken. You're competing with twenty other founders for the same recognizable pair.
The test for a compound:
- Does the compound describe something the company actually does? "Airbnb" is literal (air mattress + B&B). "Salesforce" is literal. If the compound is poetic-but-vague, it ages worse.
- Is the compound pronounceable as a single word? "Workday" yes. "Quickworkflowapp" no.
- Does either component carry trademark conflicts? Two-word compounds are more often blocked than invented names.
The naming problem in 2026 isn't creative. It's structural. — Egzecute naming brief
The trademark and domain gauntlet
Before you fall in love with a name, run the four-filter test:
1. USPTO trademark search
Search the United States Patent and Trademark Office database. If a registered mark exists in your trademark class (computer software is class 9 and 42), the name is dead. Period. Same applies for EU, UK, and any market you're entering.
2. Domain availability
The .com is the only one that matters for default brand recall, regardless of what anyone tells you about .io, .co, .ai. The .com redirects. Buying the .com of a name that's already taken on .com-redirect costs $5k–$50k typically. Budget for it or move on.
3. Social handle availability
Check @yourbrand on at least Instagram, LinkedIn, X, and YouTube. If three of the four are taken, the name fragments your brand presence — every channel becomes "yourbrand_official" or "the.real.yourbrand" which dilutes recognition.
4. Google spell-correction test
Open an incognito browser. Search the name. If Google asks "did you mean…" and suggests a dictionary word, you have an entity-recognition problem baked in from day one. Modified names usually fail this. Invented names sometimes pass. Compounds almost always pass.
Reject any name that fails two of these four. Names that pass all four are gold — buy them immediately.
Founders lose months to names that failed filter one on day one. Search the register before you run the workshop, not after. — Egzecute naming brief
A worksheet for testing names
For every serious candidate, fill in:
NAME: [the candidate]
STRATEGY: [invent / modify / compound]
USPTO check: [pass / conflict in class ___]
.com domain: [available / taken / available for purchase at $___]
@handles: [Instagram: ] [LinkedIn: ] [X: ] [YouTube: ]
Google spell-correction: [passes / fails — suggested word: ___]
Pronounceability (1–10): [score]
Memorability (1–10): [score]
Spelling consistency (does everyone spell it right after hearing it?): [yes/no]
Connotation (positive/neutral/negative/none): [note]
Run this against ten candidates and the top three usually emerge cleanly. The candidates that founders are emotionally attached to but score poorly on the worksheet are exactly the ones to walk away from.
Reject any name whose .com costs more than $20k unless you're a Series A or later. Below that stage, the money is better spent on product or distribution. Domain-squatting tax is real but it's optional — pick a different name.
What to do after the name
Once you've picked a name that passes the gauntlet:
- File a trademark application immediately, in your primary class. This is the most important $300 you'll spend.
- Register the .com — and the
.co,.io, and.aivariants if your category is competitive. - Claim all four social handles even if you don't post to them yet. Squatters are real.
- Submit a Wikidata entry for the company. Wikidata is the structured backbone Google uses for entity validation; an early entry accelerates entity recognition.
- Start writing the brand name into every piece of content you publish, in natural prose. Repetition is how Google's NLP learns that this character string is an entity.
Then live with it. Names don't reveal their quality in week one. They reveal it in year three, when the brand has either become a verb or remained a typo.
