← お知らせ

Why a prior trademark search comes first

Filing without a prior trademark search risks refusal, invalidation, and infringement claims. Here is why the search must come first.

When you file a trademark in Korea, the first thing the KIPO examiner checks is similarity with prior marks. If a similar mark is already registered for identical or similar goods, registration is refused under Article 34(1)(7) of the Trademark Act (similarity to a prior-filed or registered mark).

This article covers (1) what costs you incur by skipping the search, (2) how examiners assess sound, appearance, and meaning, (3) practical traps like dominant-portion observation for composite marks, and (4) what AI search catches versus what still requires attorney judgment.

Three things that happen if you skip the search

Filing without a prior search typically leads to one of three scenarios. All three carry unilateral cost — KIPO official fees are never refunded, and attorney fees usually are not either.

  • Refusal — If an identical or similar mark is already registered for identical or similar goods, Article 34(1)(7) bars registration. You can argue via response and amendment, but other than deleting designated goods or amending the mark, refusals are hard to overcome.
  • Invalidation — Even if you slip past refusal and register, a prior rights holder can file an invalidation trial (Article 117) within five years and extinguish your right.
  • Infringement — When the business grows, the rights holder can demand injunction (Article 107) and damages (Article 109). Article 110 even allows the infringer's entire profit to be presumed as the damage. Domain, signage, packaging, and marketing all need replacement.

Filing and examination fees are not refundable

KIPO official fees per class are not returned even after refusal. Attorney fees are usually non-refundable too. Prior search is the cheapest insurance against all this sunk cost.

How examiners judge similarity

Article 34(1)(7) similarity is judged across appearance, sound, and meaning. If any one dimension creates a likelihood of source confusion in trade, the marks are treated as similar. KIPO's examination guidelines call this objective, overall, and isolated observation.

In practice, overall observation is the default. For composite marks (multiple elements combined), examiners also apply dominant-portion observation (the part central to identification) and separate observation (per element).

DimensionSpecific testSimilar example
SoundPronunciation, syllable count, stressBLOOM vs 블룸 (transliteration match)
AppearanceLetterforms, figurative composition, colorBLOOM COFFEE vs BLOSSOM COFFEE (same prefix + same suffix)
MeaningImpression, semantic fieldBlooming vs Full bloom (shared blossom imagery)

Composite mark trap — dominant-portion observation

The Supreme Court has held that when a composite mark has a dominant portion shaping consumer perception, similarity can be judged on that dominant portion alone — without requiring the portion to be separable. Recent rulings (e.g., the November 20, 2025 decision) confirm this.

Common composite-mark mistake

Assuming "FRESH BLOOM COFFEE won't be compared on BLOOM alone" is risky. If BLOOM carries the recognition, FRESH (generic adjective) and COFFEE (industry term) have weak distinctiveness, and your mark may collide with a prior mark whose dominant portion is BLOOM.

Goods similarity — the mark is not the only factor

Even if two marks are similar, refusal is avoided if the designated goods are dissimilar. Conversely, a slight mark difference can be refused when goods are identical or closely related. Goods similarity uses NICE classification but also actual trade reality.

Five factors for goods similarity

Nature
Function, materials, form
Quality
Grade, price range
Use
Same end-use by consumers?
Channel
Same manufacturers and distribution?
Audience
Overlapping consumer base?

What AI similarity search catches automatically

iphere's AI similarity search uses KIPRIS's getAdvancedSearch API to auto-generate queries and narrow candidates. It includes variant spellings you may not think of.

  • Transliteration & compound variants — Latin marks rendered in Hangul phonetics, plus combinations with generic nouns and prefixes.
  • NICE class & similarity group matching — Only registrations in the same NICE class and KIPO similarity-group codes (e.g., G0102) are surfaced and ranked.
  • Risk grading — Sound/appearance/meaning similarity combined with goods distance, in high / medium / low bands.
  • Examiner-perspective comments — The Opus model annotates each candidate with a one-line Article 34 grounds assessment.

AI narrows candidates, attorney judges

AI is strong at candidate discovery and initial grading. Subtleties of dominant-portion vs separate observation and shifts in case law still need an attorney. iphere combines both — you narrow quickly, the attorney finalizes Article 34 review.


Run AI similarity search as you start a trademark application

Enter your mark — KIPRIS query generation, risk grading, and attorney Article 34 review arrive on a single screen.

Start a trademark application

A single search before mark selection prevents tens to hundreds of thousands of won in sunk cost and months of rebranding. The cheapest insurance is the one you buy first.

本記事の原文は ip-here プラットフォームに掲載されています。

ip-here で読む