What A Name Has To Do
A product name does three jobs: it's a memorable handle for the product, it signals what category the product is in, and it leaves room for the product to grow. A name that does the first two but fails the third becomes a ceiling. The number of AI products that named themselves into a corner in 2023 is large.
Names that survived: did all three. Names that dated: did one or two.
Patterns That Aged Well
Pattern A: The Invented Word
Five-letter coined words with no prior meaning (Flowtix, Linear, Notion). Pros: memorable, ownable, easy to trademark. Cons: requires marketing to establish category fit.
Pattern B: The Common Word Reused
A familiar word redirected into a new domain (Drift, Linear, Replit). Pros: easy to remember, evocative. Cons: SEO competition, trademark fights.
Pattern C: The Founder-Memory Name
Tied to something specific from the founder's history (Asana, Stripe). Pros: story to tell, no category constraint. Cons: requires storytelling discipline.
Pattern D: The Compound Word
Two familiar words pushed together (FaceTime, Slack-ish brands). Pros: semi-descriptive without being limiting. Cons: harder to trademark.
Patterns That Aged Badly
- The “GPT” suffix. EmailGPT, LegalGPT, MarketingGPT. Tied to a specific vendor and feels like a wrapper.
- The “.ai” gimmick. Names where the .ai TLD is essential to the brand. The TLD is fine; making it the name isn't.
- The robot mascot. Anything with “bot” in the name signaled “chatbot”. Products outgrew chatbot framing fast.
- The verb-noun. “WriteAI”, “SummarizeBot”. Locks the product into one feature.
- The trademark-violation name. Anything that included a giant company's mark (“GoogleScribe,” etc.) got C&D'd.
- • If you triple your scope in 18 months, does the name still work?
- • Can a non-English speaker pronounce it on first try?
- • Is it 4–9 letters? Beyond 10 is a flag.
- • Does it have a clean .com? Or are you stuck with an alt-TLD?
A Practical Naming Process
- Define the product in one sentence. If you can't, name it later.
- Brainstorm 100 names. Wide net. Most are bad. That's normal.
- Cut to 20. Eliminate the bad patterns above.
- Domain check. Drop anything where the .com is unobtainable or 5-figure.
- Trademark check. USPTO and EUIPO basic search. Don't skip.
- Pronunciation test. Say each remaining name out loud, 10x. The ones that get harder are out.
- Top 5 to a small panel. 5–10 people in your target market, ideally not friends. Ask what they think the product does. Surprising answers are flags.
- Pick. Don't commit to a brand exercise yet. Use the name for 2 weeks and see if it ages.
Validation Beyond Domain Availability
A domain being available is the lowest bar. Better validation:
- Search the name + your category. Are you findable in 6 months?
- Search the name in major search engines. Any uncomfortable associations?
- Search the name on social platforms. Are the handles open?
- Search the name on app stores. Are there confusing existing products?
Naming Inside The Product (Features, Models)
Many AI startups invest in the brand name and ignore feature naming. Bad feature names accumulate. By year 2 the product UI is unreadable. Three guidelines:
- Feature names should describe the action (“Compose”), not the technology (“GPT-4 Writer”).
- Don't give every internal model a public name. Users do not need to know about your three model versions.
- Avoid the “Smart [thing]” prefix. “Smart Search”, “Smart Reply” — signals “the unmarked one is dumb.”
When to Rename, When to Hold
Renaming is expensive: domain transfer, brand redesign, customer confusion, SEO loss. Only rename when one of three conditions is true:
- The current name is actively limiting market expansion.
- The current name has accumulated negative associations.
- A genuine pivot has rendered the name irrelevant.
The best AI product names are ones that stop being notable. After two years, the name is invisible — the product is what users see. That's the goal.
See also branding an AI startup and our design service.
FAQ
Should we use AI to name our AI startup?As an idea generator, fine. As the final decision, no — AI converges on the same patterns.
What about codenames during development?Useful. Don't accidentally ship them.
How much does professional naming cost?$5k–$50k for the process. Worth it for a flagship; overkill for a feature.