How to find ENS names with AI: a practical field guide

The hardest thing about picking an ENS name isn't the price. It's deciding what name to pick in the first place. You think of a word, try it, get told it's taken, think of another, and within ten minutes you have lost faith in your ability to name anything.

AI Search flips the model. Instead of guessing, you describe. The system embeds your description, searches the full .eth namespace for the closest matches, and returns a shortlist of names — many of which you would never have typed into a checker manually.

This is a field guide. Real query patterns, what they're good at, what they're not, and how to read the results. If you've tried AI Search once and bounced because nothing jumped out, the issue is almost certainly the prompt, not the index.

The three things to know

1. Describe the shape, not the exact name. "A name that sounds like a Scandinavian design studio" works. "Some name" doesn't.
2. Add constraints. Length, character composition, ending, vibe. Each constraint sharpens the result list dramatically.
3. Iterate. The first prompt is rarely the right one. Treat it like a conversation — narrow, widen, pivot.

Query patterns that work

Theme / vibe

Describe the feeling you want the name to carry. The semantic index doesn't care about letter overlap — it cares about meaning.

• names that feel like a wildlife conservation project
• names that sound like a DeFi protocol from 2019
• minimalist tech startup names ending in -ly
• nautical names, 5 to 7 letters

Industry / niche

Anchor the search in a specific domain. You'll get names that have already been picked up by similar projects plus close neighbours that haven't.

• names a hardware wallet company would use
• names that sound like a privacy tool
• names suitable for a music NFT platform

Personal / collectible

Looking for a handle rather than a brand? Anchor on cohorts, era, or language.

• italian first names, 4 letters
• popular boys names 5+ letters
• greek mythology, single word
• 1990s rap aliases

Word structure

Sometimes you want a specific shape rather than meaning. Describe the structure.

• palindromes, exactly 5 letters
• three-letter words that aren't acronyms
• names starting and ending with the same letter
• dictionary words with double consonants

Cross-language

English is one of dozens of languages the index has seen. If you describe a concept, you'll often get the word from another language as the highest-ranked hit.

• words meaning "ocean" in non-English languages, under 6 letters
• japanese aesthetic concepts, single word
• spanish verbs in the infinitive, ending in -ar

Reading the results

Each result row carries five signals you should glance at before clicking through:

• Status — available, registered, in grace, in temp premium, in dutch auction. Filter to available if you just want to register today.
• Length — 3, 4, 5+. 3 and 4 are scarcer and pricier to register annually.
• Composition — letters only, contains digits, contains emoji. Pure-letter names age better; digit names slot into the numeric clubs.
• Collection tag — 999 Club, 10k Club, palindrome, emoji, dictionary word, etc. The collection floor on the marketplace tells you what the name is plausibly worth.
• NameGuard rating — surfaces homoglyph attacks (Cyrillic letters that look like Latin) and other visual-confusable risks before you buy.

Five practical use cases

1. Naming a new brand

Start broad, narrow with constraints. "Modern fintech names ending in -ly, registered or available, 5 to 8 letters." Skim the top 30 results — note the ones already taken by similar brands (signal that the cohort is correct), then look at adjacent available names. Run a second pass with NameGuard "high" filter so you don't land on a homoglyph trap.

2. Building a collection

Search for cohort members rather than individual names. "All 3-letter names that are also dictionary words." Now you have a list to track on the Watchlist and set expiry alerts for. Most cohort plays come from owning the set, not the single best member.

3. Finding pairs and sets

Naming a product line or set of related entities? Describe the relationship."Greek planet names that haven't been registered, 5-7 letters." The semantic ranking surfaces the cohort; the availability filter narrows it to ones you can buy.

4. Replacing a long handle

Have a 14-letter handle you want to shorten? "Short names that semantically match 'gracebellafarmsteadshire'." The index returns names that share the same essence in fewer characters. Often a 5-letter neighbour is a meaningful upgrade.

5. Finding flips

Want to spot inventory that's both grabbable and resaleable? Combine vibe with category:"4-letter dictionary words currently in temp premium, sorted by price." The semantic filter narrows from "any 4-letter name decaying right now" (thousands) to "actual words" (dozens).

What AI Search is not good at

• Exact-match lookup. If you know the name you want, use the regular domain search. AI Search is for discovery, not verification.
• One-word prompts. cool doesn't tell the system anything. Add context: cool short names for a music app.
• Pure-numeric prompts. For numeric clubs, use the dedicated 999 Club / 10k Club / 100k Club landing pages — they cover the full range.
• Slang as ground truth. The index understands slang but ranks registered slang names higher than emerging ones. Add "current" or a year if you want recency.

Power-user shortcuts

• Length ranges in the prompt. "4-6 letters" or "under 8 chars" carries through to the filter.
• Negative constraints. "...but no digits" or "...exclude emoji" tightens results immediately.
• Sort by availability. Toggle the available-only filter to skip the thousands of taken neighbours.
• Favourite as you go. The heart icon on each row persists across sessions and exposes a count in the header so you can come back to a shortlist later.

The TL;DR

Describe the shape, add constraints, iterate. The AI doesn't pick the name for you — it surfaces the cohort. Your job is to read 30 results and pick the one that lands. Most people who say "I can't find a good ENS name" haven't asked for one; they've asked whether a specific guess is available.

Try a prompt now — open AI Search →