Naming a Startup — the Only Checklist That Matters

The strongest opinion in startup naming: do not pay $8,000 to a naming agency. They will deliver a shortlist of eleven names, three of which have unregistered domains, two of which are already trademarked in your category, and six of which sound like they were generated by someone who has never worked in your industry. You can do better in a weekend with a spreadsheet and the right generators.

Here is the actual checklist. Thirteen items. If a name fails any of them, cut it. Do not fall in love before item 13.

1. The ".com is available" check (do this first, cut ruthlessly)

The .com for your exact brand name is table stakes. "You can use .io / .ai / .co" is what people say when they can't get the .com. It costs you brand recall and credibility, especially outside tech.

Hard filter: if the .com is parked by a domainer asking $12,000+, cut the name. You will not pay that bill. Pretending you might is sunk-cost thinking.

Acceptable: the .com is parked for $500–$3,000 and you can budget it. Common for 2–3 syllable invented words. If it's $200 on registry, that's a gift.

Not acceptable: the .com belongs to a working business, even a tiny one. You will spend 2 years explaining why you aren't them.

For first-pass bulk generation that produces .com-plausible names, the Startup Name Generator biases toward invented-word patterns with better domain availability than real-word names.

2. Say it on a podcast

Say the name three times, in a sentence that sounds like a podcast guest introduction: "Today we're talking with the founder of [NAME], a company that..."

If the host would have to spell it, cut it. If "the" has to come in front ("The Scrape") to make it work in a sentence, cut it. If it rhymes with something embarrassing in common English, cut it.

The podcast-test matters because every piece of word-of-mouth growth you'll ever get runs on someone saying your name out loud to someone else. If the receiver can't type it into Google from hearing it, your growth stops at the retelling.

3. The spelling test

Can an English-speaking adult type your domain correctly after hearing it one time? Not two times — once.

"Flickr" fails this. "Lyft" fails this. Both companies spent 7–8 figures teaching the market the spelling anyway. You probably don't have that budget.

"Figma" passes. "Linear" passes. "Notion" passes. All invented or repurposed words with a single obvious spelling.

The trade-off: unique-spelling names are easier to trademark and own in search; obvious-spelling names are easier to grow. For small bootstrap startups, obvious wins. For VC-funded consumer brands, unique can be worth the teaching cost.

4. The trademark search

Three places you have to check before committing:

1. USPTO TESS (free, uspto.gov/trademarks). Search exact match and phonetic equivalents. Especially check your category (TM categories 9, 35, 42 for software). 2. The equivalent registry for your top three target markets. EUIPO for Europe, JPO for Japan, CNIPR for China if relevant. 3. Google. Search "[name] [your category]." If a competitor is using it even without a registration, you will fight them.

A conflict in your exact TM class is a hard stop. A conflict in a completely unrelated class (your "Comet" is software; their "Comet" is a cleaning brand) is usually fine, but talk to a trademark attorney before launch.

5. The international check

Google Translate your proposed name into the top 10 world languages. If it means "diarrhea" in Portuguese or something regionally obscene in Mandarin, you will hear about it from early international users forever.

Classic examples: Chevy Nova (nova = "doesn't go" in some Spanish idioms, probably overstated but the story stuck), Mitsubishi Pajero (regional slang in Spanish), the Chevy LUV (fine until you say it with any Asian-market accent).

You don't need to perform exhaustive due diligence. Google Translate + a quick native-speaker sanity check from a Twitter DM is enough.

6. The one-sentence pitch test

Fill in: "We're [NAME], and we help [AUDIENCE] [DO JOB]."

If the name makes the sentence grammatically awkward ("We're An.Co, and we..."), rework. If the name is too generic ("We're Software, and we..."), rework. If the name forces you into a backstory before you can state the value ("Well, 'Rhymenoceros' is actually a reference to..."), rework.

The best startup names make the sentence trivial to say and remember. The founder of Linear can say "We're Linear, and we help software teams move faster" without the audience needing a glossary.

For first-pass name candidates that read as drop-in-sentence-ready, the AI Startup Name Generator tends to produce short, invented-word options that fit this pattern more cleanly than literal-word combinations.

7. The one-sentence pitch test — again, in six months

Will the name still fit when you pivot?

"MattressBuddy" is a fine name for a mattress company. It's a terrible name for the sleep-tracking app you'll pivot to in 18 months.

The pattern: name the company around the problem space, not the current product. "Figma" is design. "Notion" is thought. "Linear" is velocity. None of them encode a specific feature or category you'll outgrow.

Products get names. Companies get brand names. A startup is both. Decide which you're naming before you commit — if you're pre-launch, pre-category, name the company; the product can take a descriptive name.

For a worked exercise in product-vs-company naming, the Product Name Generator gives you product-ish options and the Business Name Generator gives you company-ish options. Running both side by side makes the distinction visible.

8. The TLD-split test

Try the name as different TLDs. If .com is unavailable but .co, .ai, .app, .io are, think hard about why.

Acceptable .com-alternative scenarios:

• Dev-tool brands using .dev (platform norm).
• AI-first brands on .ai (platform norm, costs real money but reads as intentional).
• Specific-vertical brands on .app or .io (same).

Not-acceptable .com-alternative scenarios:

• Consumer brand on .co because .com is taken by a dormant parked page.
• Any startup with "The" in front of its name to make .com work (thefoobar.com). "The" prefix is almost always a mistake.
• Hyphenated .com (foo-bar.com). Reads as spam, kills word-of-mouth.

If you can't get the .com and you're not in a vertical where the alt-TLD is a norm, keep brainstorming. You will be happier.

9. The social handle check

Same name available on Twitter/X, Instagram, GitHub (if dev), LinkedIn Company Page, and TikTok (if consumer).

Must-haves: Twitter/X, LinkedIn. These are where your acquisition email signatures and your investor conversations live. Nice-to-haves: Instagram and TikTok for consumer, GitHub for dev-tool. Workaround acceptable: @getfoo or @foohq as a Twitter handle when @foo is taken by a dormant 2008 account.

If your main-brand handle requires a @try / @get / @use prefix on more than one platform, your name is too common. Cut it.

10. The two-year-test

Say the name out loud and imagine saying it 400 times in investor meetings over the next two years. If you cringe, cut it.

This is a soft test and it matters more than any other item on this list. You will say the name constantly. If you don't like saying it in month one, you will loathe it by month six.

Edge case: founders who pick a name they find clever enough to say but their partner / co-founder hates. The co-founder will stop saying it. The company will drift to the generic descriptor nobody chose on purpose. Co-founder veto is real; use it early.

11. The autocomplete test

Type the first three letters of your name into Google. If the autocomplete suggests a huge brand in a different space, you'll fight for SEO ranking forever.

"Linea" autocompletes to "Linear" (software) and "Lineage" (anime). Linear owns their autocomplete through brand strength; a startup starting from zero can't.

"Postwell" has clean autocomplete — no major conflict. Easier to rank on brand search.

This matters less in year one. It matters a lot in year three when "google [brandname]" is your main acquisition channel.

For brainstorming names with uncrowded autocomplete, pull from the SaaS Product Name Generator which biases toward invented terms rather than real-word mashups that collide with existing brands.

12. The pronunciation drift test

Does the name have one obvious pronunciation, or multiple plausible ones?

"Hugging Face" is one pronunciation. "Nike" has two (Nike-ee vs Nike, resolved by the brand). "Hyundai" has three. "Deebo" and "Dee-bo" are different.

Multiple-pronunciation names mean every podcast interview starts with "and how do you say it," every meeting has someone sheepishly asking, every TV ad has to teach you. It's a tax on every piece of brand communication for the lifetime of the company.

Pick the one that reads one way.

13. The six-month regret check

Imagine you launched today with the name. In six months, you've closed a $2M seed and are hiring. A new engineer joins and asks "where did the name come from?"

Can you tell a one-sentence story that doesn't make you wince?

Good: "It's an old word for the thing we're building." Good: "It's the first non-boring thing available with a .com." Good: "It was the co-founder's grandma's cat." Bad: "I don't know, we just picked it." Bad: "Oh, you wouldn't believe the Reddit thread that led us here."

The story doesn't have to be impressive. It has to be told without apology.

The checklist, in a usable form

For every candidate name:

• [ ] .com is available or affordable
• [ ] Passes podcast-say-it-out-loud test
• [ ] One obvious spelling
• [ ] No TM conflict in your category
• [ ] No awful meaning in top 10 languages
• [ ] Fits a one-sentence pitch
• [ ] Will fit the company in 2 years, not just the product
• [ ] Not forced into weird TLD
• [ ] Social handles available
• [ ] You can bear to say it 400 times
• [ ] Clean search autocomplete
• [ ] One pronunciation
• [ ] You can tell its origin story unembarrassed

A name that passes all 13 is a keeper. A name that fails on items 1, 4, or 13 is a hard cut regardless of everything else. The rest are negotiable depending on your budget and runway.

The generation workflow

For a weekend of naming:

1. Generate 300 candidates from 4–5 generators with different biases. 2. First-pass cut on items 1–3 (.com, podcast, spelling). Should drop you to about 30. 3. TM search (item 4) on all 30. Drops to about 15. 4. Live-test the remaining 15 on items 5–13 with 2–3 people whose opinion you trust. 5. Pick 3. Buy the domains. Sit with them for a week. 6. Commit to one, register the trademark, grab the handles.

The founders who agonise about this for 3 months end up with a slightly worse name than the founders who ran this process in 72 hours. Perfect is the enemy of shipping.

For the "generate 300 candidates" step, mix outputs from the AI Startup Name Generator, the Startup Name Generator, the Tech Startup Username Generator (for the invented-word slot), and the Business Name Generator. Different generators have different biases — invented words, blend words, metaphor words, domain-suffix words — and the mix gives you the breadth to find the one that passes all 13.

The last mental model: your startup name is a password you're assigning yourself for the next decade. Make it one you can live with typing a thousand times.