DeepNameSearch

How DeepNameSearch works

DeepNameSearch is a company-name check. You give it a candidate name and a target industry; it runs a fixed pipeline of deterministic checks and web research, then returns a per-dimension score and a go / conditional / hard-stop verdict. It is deliberately thorough — it actively looks for every real risk a name carries — so that when it says a name is clear, you can commit with confidence.

What you give it

  • A candidate company name (one per run).
  • A target industry — this steers the trademark and persona reasoning.
  • Optional sub-brand pillars (e.g. Group, Labs, Studios) to test.

Two modes: Quick check runs domains and social handles only, in a few seconds. Deep audit adds web research, trademark reasoning, confusable-name analysis, linguistics and phonetics, and produces the full report.

The pipeline

  1. Normalize — derive the domain list (per TLD) and the social-handle list from the name.
  2. Domains — an RDAP lookup per TLD, cross-checked against live DNS (SOA / NS delegation), so a parked-but-registered domain reads as taken rather than free. Weighted heavily toward the .com.
  3. Social handles — best-effort checks on GitHub, YouTube, TikTok, X, Instagram, LinkedIn and Facebook, each tagged with a confidence level (some platforms block automated checks and are reported as unverified).
  4. Research — a large language model with live web search looks for real operating companies using the name or a close variant, trademark exposure in the relevant classes, and confusable neighbours (spelling, phonetic, and re-ordered-consonant variants). Every factual claim carries a source link or is labelled unverified.
  5. Format — the written analysis is converted into a strict structured report.
  6. Assemble — the server merges the deterministic checks with the analysis, computes the composite score with fixed weights (it never trusts the model’s own arithmetic), applies the hard-stop and domain gates, and emits the verdict.

What gets scored

The composite verdict weights five dimensions:

  • Trademark (30%) — likelihood-of-confusion risk in the target classes.
  • Domain (30%) — the realistic acquisition picture, .com first.
  • Phonetics (15%) — dictation traps, homophones, sub-brand stutters.
  • Linguistic (15%) — slang and offensive meanings across ten major languages.
  • Persona (10%) — how the name reads to the target buyer.

Two more dimensions — digital namespace (domains and handles combined) and memorability — are shown in the report but sit outside the composite.

Verdict and hard stops

A hard stop forces a red verdict regardless of the score: an active company with the same name in the same category, a critical linguistic hit (profanity or a slur in a major market), a .com owned and used by a direct competitor, or every candidate domain already registered. Absent a hard stop, the verdict follows the composite — and a name can’t come back GO with a green light if the .com isn’t available.

What it deliberately does not do

  • It is not legal advice and not a formal trademark clearance. Trademark risk is a knowledge-based read plus pre-filled deep links into the USPTO, EUIPO, WIPO and UK IPO registers — not a live register search. Confirm anything that matters with a licensed attorney.
  • It does not register anything — no domains, no trademarks, no entities.
  • It stores nothing. There is a short in-memory cache and a per-IP rate limit; there is no account, no history, no database of your searches.

Run a name · See a worked example · FAQ

DeepNameSearch — a company-name check: trademarks, domains, handles, and more, in one report. Public, no login, one name per run, nothing stored.

Not legal advice and not a formal trademark clearance. Trademark risk here is a knowledge-based read plus deep links into the official registers — confirm anything that matters with a licensed attorney.