Startup domain names — the naming-to-acquisition playbook.

Naming and domain acquisition are the same project. Founders who treat them separately end up either with a name they love and can't afford the domain for, or a cheap domain they don't love and have to rebrand around. The companies that get this right run both tracks in parallel from week one.

The five steps below are the sequence we run with funded AI startups. Each step kills candidates that fail the next step's constraints — by the time you're at step 5, you're acquiring a name that's already passed trademark, .com availability, and budget gates.

Don't brainstorm. Generate against a constraint matrix. Define the constraints first:

• Length — target 5–8 characters before TLD.
• Syllables — 1–2 syllables only.
• Phonetics — no ambiguous letters (ph/f, c/k/q, ks/x).
• Semantic field — abstract or invented, not category-locking.
• Avoid — "ly" suffix, "AI" prefix, numbers, hyphens.

Generate 50+ candidates against these constraints. Use a structured method (combinatorial blends, latin/greek roots, invented coinages) rather than vibes. The constraint matrix is what makes the next steps actually filter — not just compare aesthetics.

Kill candidates that fail any of these:

1. USPTO TESS search for live marks in your International Class (Class 9 for software, Class 42 for SaaS).
2. EUIPO search for European launch markets.
3. Linguistic clearance — Google translate against your top 10 target-market languages. Real businesses have shipped names that mean "diarrhea" in Italian and "incompetent" in Portuguese.
4. Social handle availability — Instagram, X, LinkedIn, TikTok. Not blocking criteria but reduces optionality.
5. Negative-meaning search — Google the bare word. What's on page one? What are people complaining about?

Expect to cut your 50 candidates to 10–15. Anything failing trademark in your largest market is dead — don't try to negotiate around an existing mark in a class you'd compete in.

For each surviving candidate, check the .com in this order:

1. Free. Register immediately at Cloudflare or Porkbun. Don't tell anyone the candidate list before this step — drop-watchers exist.
2. Premium-listed on Sedo/Afternic/Dan with a buy-it-now price. Note the price; this is your floor.
3. Owned, no listing. WHOIS lookup. Identify owner type (portfolio investor vs. operating business). This is broker territory.
4. Operated by a real business. Functionally locked. Move on.

By the end of step 3 you'll typically have 3–5 acquirable candidates with rough price expectations.

The single highest-leverage moment to spend at the top of your band is the gap between term sheet signing and funding announcement — before the seller can google your round size.

For free names: register and move on. For listed names: buy at acceptable buy-it-now or negotiate via the platform. For owned names: anonymous outreach (broker or non-company email), comparable-sale anchoring, escrow, registrar transfer, DNS cutover.

Full taken-domain acquisition playbook →

1. Misspellings that worked in 2007 don't work now. Lyft, Flickr, Tumblr launched when the domain market was thin and the naming style was novel. Today, misspellings signal "we couldn't afford the real word" and erode trust.
2. The "ly" suffix is saturated. Calendly, Bitly, Grammarly were early. Every new "-ly" name in 2026 reads like a placeholder.
3. AI- prefix collision. Twenty AI startups using "AI" as a prefix are indistinguishable from each other in a press release. The naming convention that worked in 2022 is now a commodity signal.
4. Product-locking names. "Docsly" works for one of three products. "Notion" works for a docs tool, a CRM, or a calendar. Pick the second kind.
5. Trademark conflicts in your largest market. Naming around a known mark inside your competitive class is a litigation timer, not a defensible position.

Selected public and reported transactions, anonymized where private.

• Cursor.com (acquired post-launch from cursor.so) — reported high-six-figures.
• Perplexity.ai — registered (free) at founding; .com still owned by a separate entity.
• Character.ai — registered (free) at founding; matching .com remains separate.
• Linear.app → linear.com — eight figures, post Series B.
• Notion.so → notion.com — high seven to low eight figures (reported, undisclosed).
• Brex.com — reportedly mid-six figures pre-launch.
• Ramp.com — reportedly low seven figures post-Series A.
• Glean.com — high six figures, broker-acquired pre-launch.
• AI.com → OpenAI — approximately $70M, 2024.
• Anthropic.com — registered at founding (anthropic was an unowned coined word).

The pattern: the companies that picked invented words got the .com free; the companies that picked real words paid for the upgrade at the moment they could afford it.

What domain should a startup buy first?

Two-word brandable .com is the default — it's the shape most funded startups land on and where the most acquirable inventory exists at reasonable prices ($15K–$250K). One-word .com is the upgrade path (Series A and beyond). One-word .ai is the credible launch alternative if the .com number doesn't pencil at seed stage.

How much should a startup spend on a domain?

Loose budget heuristic: 0.5–2% of your seed raise on the domain. A $3M seed supports a $15K–$60K domain. A $10M Series A supports $50K–$200K. Spending less than 0.5% means accepting a worse name and rebranding later (more expensive over the lifecycle). Spending more than 2% at seed is usually founder ego, not investment logic.

Do I need to own the .com to raise?

No, but you need a credible path to it. VCs in 2026 are comfortable funding companies on .ai, .io, or .so for the first round. By Series B, almost every well-positioned company has either acquired the matching .com or has a documented plan to. Operating long-term on a non-.com is viable; operating without any premium domain at all is the actual red flag.

What are the worst naming mistakes?

Five repeat offenders. (1) Common-word misspellings (Lyft, Flickr — these worked because they were 2007; they don't anymore). (2) Trailing 'ly' (saturated, weak). (3) AI- prefix collision with twenty other startups. (4) Names that lock the company into one product. (5) Names trademark-blocked in your largest target market. Each of these forces a rebrand inside three years, which costs more than the right domain would have.

How long does the naming + acquisition process take?

Naming itself: 2–4 weeks if you run a structured process, 3+ months if you don't. Trademark and linguistic clearance: 2 weeks. Domain acquisition once the name is chosen: 1 week (registered new) to 8 weeks (off-market broker negotiation). Plan 6–12 weeks total from 'we need a name' to 'we own the domain'.