How to choose a startup name and domain
How to choose a startup name and domain
Pick a short, brandable, phonetically clear name (5-12 characters), confirm it's free in USPTO TESS and on the social handles you actually need, then buy the cheapest TLD that doesn't actively confuse customers. If the .com is taken and the squatter wants $40,000, run on a subdomain or a .ai/.io for now and revisit after Series A. Almost every founder over-spends here, and almost none of it correlates with whether the company works.
The setup: where founders get stuck at 11pm
You've validated the idea. You have 30 customer calls in a Notion doc. You're three browser tabs deep into Namecheap, and every name you actually like is either parked, listed at $18,500, or owned by a dental practice in Wichita that hasn't updated its WordPress since 2014.
So you start considering compromises. Add an "ly". Drop a vowel. Buy the .io even though you secretly hate it. Or pay the squatter, because what's $9,000 against the rest of your life.
This is the wrong frame. The name matters. The domain matters less than your gut tells you it does. And the actual decision is a sequence of small, cheap, reversible choices, not one large irreversible one.
What every top result gets half-right
The standard advice is fine: brandable beats descriptive, .com beats everything else if you can get it, no hyphens, no numbers, check the trademark. All true. None of it answers the actual question, which is what to do when the brandable name you love costs $35,000 in the aftermarket.
The honest 2026 answer is that ~30% of recent Y Combinator batches use .ai, roughly half use something other than .com, and Google ranks them identically. The "you must have the .com" gospel is a holdover from 2008. It still applies to consumer brands targeting non-technical buyers, and it stops applying for almost everyone else.
The full decision tree below assumes you're a founder, pre-seed or seed, with a runway you'd rather spend on engineers than on a four-letter URL.
The 5-step process, in order
1. Generate 30 names, kill 25
Brainstorm in batches. Don't aim for the perfect name on the first pass; aim for 30 candidates you don't actively dislike. Use Namelix, ChatGPT, or Claude to expand patterns: invented words (Stripe, Notion, Asana), compound words (Dropbox, Mailchimp), suffix plays (Spotify, Bitly, Calendly), or tortured Latin roots (Anthropic, Vercel).
Now apply four kill filters in order:
- Pronounceable on a podcast in one try
- Spellable after hearing it once
- Not within one Levenshtein edit of an existing tech brand
- Doesn't make you cringe to say "Hi, I'm from [name]"
You'll be down to 5 candidates. Take them to the next step.
2. Trademark check (free, 10 minutes)
Open USPTO TESS at tmsearch.uspto.gov and run a basic word-mark search for each candidate. Use the "live" filter. You're looking for two things: identical marks in the same class, and confusingly similar marks in adjacent classes.
For software, that's primarily Class 9 (downloadable software), Class 42 (SaaS), and Class 35 (advertising/business services). If a registered mark exists in your class, kill the name. UDRP disputes run $1,500 to $5,000; full federal trademark litigation routinely passes $100,000. The math is bad, and the founder's time cost is worse.
For international, run the same name through WIPO's Global Brand Database and the EUIPO database. Markify and Trademarkia offer paid scans starting around $200 if you'd rather outsource. At pre-seed, a USPTO + WIPO check is enough.
3. Domain availability sweep
Once you have 3 trademark-safe candidates, check domain availability across your real options. Use Namecheap or TLD-List for fast bulk lookups; use Cloudflare Registrar or Porkbun when it's time to actually buy (both sell at near-wholesale; Cloudflare Registrar in particular charges renewal-only with no markup).
Standard 2026 prices for a fresh registration:
| TLD | First-year cost | Typical renewal | Best for |
|---|---|---|---|
| .com | $10-15 | $15-25 | Consumer brands, anything your mom will type |
| .io | $25-50 | $35-60 | Dev tools, B2B SaaS, infra |
| .ai | $50-90 | $70-100 | Anything explicitly AI-shaped |
| .co | $25-35 | $30-40 | Brandable .com alternative (warning: people type .com by reflex) |
| .dev / .app | $15-30 | $15-30 | Niche dev tools, mobile apps (forced HTTPS) |
| .xyz | $1-12 | $10-15 | Cheap, low-trust signal, fine for side projects |
If your candidate's .com is available at registrar pricing, buy it immediately. The whole conversation ends. You spent $12.
If the .com is parked or held by a squatter, you've now hit the .com tax.
4. Decide whether to pay the .com tax
In 2026, "the .com tax" for any tech-adjacent four-to-eight-letter word starts around $5,000 and routinely lands at $20,000-50,000. Six-letter pronounceable .coms like Plaid-style brands trade in the low six figures. Squatters know what YC checks look like, and they price accordingly.
Whether to pay is a runway question, not a branding question:
| Situation | What to do |
|---|---|
| Pre-seed, no revenue, $50-500k raised | Don't pay. Use .io / .ai / .co or a subdomain. Revisit after Series A. |
| Seed with traction, $1-3M raised | Maybe, if it's $5-15k and the alt feels wrong. Negotiate hard. |
| Series A+, real customers, brand confusion already biting | Pay. Budget $20-50k. Treat as marketing capex. |
| Consumer brand targeting non-technical buyers | Pay earlier; the .com is operationally load-bearing. |
If you decide to negotiate with a squatter, here's the script that works:
- Don't email from your funded-startup domain. Use a Gmail.
- Don't introduce yourself as a startup. You're a hobbyist or a consultant.
- Open at 20% of the listed price. List price is fiction; squatters expect haggling.
- Use a broker (Sedo, Escrow.com, GoDaddy Domain Broker) once you're past $5k. The escrow is worth the 10%.
- Walk away easily. There are 10 other names that work.
A founder I know got Stripe-tier .com listed at $45,000 down to $11,000 by sending a single email from a personal address claiming the name was for a hiking blog. The squatter took the cash. It happens.
5. Buy defensively, but don't get cute
Whatever you land on, also buy the obvious typos and the .com if you ended up on a different TLD. Total spend: under $100. This protects against future competitor cybersquatting and makes redirects trivial.
Skip the long tail. You don't need yourcompany.pizza. You need the .com (or your primary), the most likely typo, and the matching social handles on Twitter/X, GitHub, LinkedIn, and whatever else your buyer actually uses. Namechk.com or Namecheckr both do bulk handle searches in one shot.
The subdomain option almost no one talks about
Here's the unsexy answer: you don't always need a top-level domain. A subdomain on a parent brand is a perfectly legitimate operating address for the first 12-24 months of a company.
Cadence runs on cadence.withremote.ai. We made that choice deliberately. The .com was a five-figure number, the .ai was available but $90/year and we were already on a parent brand, and the engineering team's instinct was that the right time to spend on a domain is when you have customers complaining about the URL, not before. We've had zero complaints. We have, instead, the engineers we would have spent the money to hire.
The subdomain pattern works when:
- The parent brand is established and trustworthy
- Your buyer arrives via SEO, ads, or referral, not by typing the URL
- You're early enough that the brand is fluid
It breaks when:
- You're building a consumer brand (mom needs to type it)
- You're raising a Series A and a brand-conscious lead investor flinches
- The parent brand and the product brand confuse customers
If you're a B2B SaaS sold to engineers, this is a perfectly defensible move. If you're a DTC mattress company, buy the .com. The framework matches our build vs buy framework for founders: the right answer is the one that frees your money for the thing that actually moves the needle.
What it actually costs to ship
Most founders read all this and then spend a Saturday on it. That's the real cost, the Saturday plus the next two evenings of name doubt. The dollar figure is usually noise.
Here's a realistic budget at each stage:
| Stage | Domain spend | Trademark spend | Total |
|---|---|---|---|
| Pre-product, exploring | $12 (.com) or $50 (.ai) | $0 (USPTO TESS only) | $12-50 |
| Pre-seed, building MVP | $30 / yr across primary + 2 defensives | $0-225 (DIY USPTO filing) | $30-255 |
| Seed, paying customers | $5-15k (premium .com) + $50/yr maintenance | $1,500-3,000 (lawyer-filed marks in 2 classes) | $7-18k |
| Series A | $20-50k (premium .com) + international filings | $5-15k (international + opposition support) | $25-65k |
For comparison, that pre-seed total is about half a week of a senior engineer at Cadence's $1,500 weekly rate. A premium .com at seed is the same as 10 weeks of a mid engineer at $1,000. Founders who buy the $30k .com before they ship the product are usually trading product velocity for vanity. Don't.
Common founder mistakes
- Spending three weeks naming. It's a one-Saturday decision. Pick one and ship.
- Optimizing for the perfect SEO match. Brandable wins in the long run. "buy-cheap-shoes-online.com" is unsearchable as a brand and will lose to one-word competitors.
- Ignoring trademark until you have a logo. Re-branding after a cease-and-desist is brutal. 10 minutes on USPTO TESS now saves three months later.
- Buying every TLD variant. You don't need .pizza. You need .com (or primary) + 1-2 typos + matching socials.
- Paying $40k for a .com pre-revenue. Same money buys you a senior engineer for 6 months, who actually ships the product that justifies the .com later.
- Trusting the registrar's "premium" tag. Every TLD has a "premium" upcharge tier where the registry inflates the price. Cloudflare and Porkbun expose the wholesale price; many other registrars hide it.
Where Cadence fits
Most of this work is unblockable, founder-only territory. You can't outsource the gut-check on whether a name feels right, and you shouldn't outsource the trademark search. But once the name is decided, the actual landing-page-to-product work is exactly the kind of scoped engineering that gets stuck because founders are non-technical or technical-but-stretched.
If your bottleneck is "I have a name, I have a domain, I just need a working v1 by next month," that's a 1-2 week booking on Cadence's founder onboarding flow, not a six-month hire. Every engineer on the platform is AI-native by default, vetted on Cursor, Claude Code, and Copilot fluency before they unlock bookings, so a $1,000/week mid is shipping at roughly the velocity a $4,000/week contractor used to.
For deeper context on the alternative paths, our take on finding a technical co-founder in 2026 and the playbook for building a startup without a technical co-founder both walk through how this fits into the bigger sourcing question.
Try it
Lock the name today. Buy the cheapest TLD that doesn't embarrass you, run a USPTO TESS search, and revisit the premium .com question only after you have paying customers. If the next bottleneck is shipping the product, book your first engineer with a 48-hour free trial: weekly billing, no notice period, replace any week you want.
FAQ
How much should I spend on a startup domain?
At pre-seed: $12-50 total. The .com if cheap, otherwise .ai/.io/.co. After Series A with brand confusion biting: $5,000-50,000 is reasonable as marketing capex. Anywhere in between, negotiate hard or stay on a subdomain.
Is the .com still required in 2026?
No. Roughly 30% of recent Y Combinator cohorts launched on .ai, and about half use something other than .com. Google treats them identically for SEO. .com still matters for consumer brands targeting non-technical buyers; for B2B SaaS it's optional.
What's the safest way to check if a name is trademarked?
Run a free word-mark search at USPTO TESS (tmsearch.uspto.gov), filtered to "live" marks, in your relevant classes (9 for software, 42 for SaaS, 35 for business services). Cross-check WIPO's Global Brand Database for international. Outsource to a trademark attorney once you're filing, not before.
Should I buy a premium domain from a squatter?
Only if you're past seed, the alternative TLDs feel actively wrong, and the price is under $20k. Negotiate from a personal email, open at 20% of the asking price, and use Escrow.com or Sedo above $5k. Most squatters take 25-40% of list if you walk away well.
When does running on a subdomain stop working?
When you start raising a Series A and a lead investor flinches, or when consumers type the URL directly and miss your site. For B2B SaaS sold to engineers via SEO/ads/referral, a subdomain on a credible parent brand can carry a company through $1-5M ARR without issue. (Cadence runs on cadence.withremote.ai today.)
Do I need to buy every TLD variant of my name?
No. Buy your primary, the .com if you ended up on something else, and 1-2 likely typos. Skip the long tail (.pizza, .ninja, .lol). Total defensive spend should stay under $100/year.