Korean keyword research built by translating an English list measures a market that does not exist. English concepts split across three Korean word layers, Korean orthography allows several correct written forms of one term, and Korean buyers frame commercial queries on their own pattern. The translated file looks complete, reports demand for phrases nobody types, and omits the phrases that carry the category.
This decision sits above the tooling. The complete guide to Naver SEO for foreign brands covers the channel, and Naver’s DataLab and Search Ads tools produce the volume numbers. This guide covers what those tools cannot decide: which Korean words belong in the research at all, and how to judge the list a vendor hands back.
Why Does Korean Keyword Research Fail on a Translated List?
The guide to Naver keyword research settles why a translated Google list fails and why seeds must originate in Korean. The narrower question here is what translation does to the Korean phrase set: it returns one acceptable equivalent per row, while Korean demand distributes across variants, and three properties of the language make that distribution wide.
One English concept maps to several Korean words. Korean carries native vocabulary, Sino-Korean vocabulary built from Chinese characters, and loanwords written in hangul. A single English category term often has a word in two of those layers, sometimes three, and they behave as separate queries.
Several written forms are simultaneously correct. Korean spacing rules permit more than one admissible way to write the same compound, and loanword spelling has a standard form alongside an established popular form. Naver appears to collapse the spacing variants into one search, while a standard spelling and a popular spelling behave as separate query strings.
The query frame moves with the market. The modifiers Korean buyers attach to a category noun, and the buying stage they represent, follow Korean purchasing behavior. Translating the English modifier set transplants assumptions that do not hold in Korea.
Which Korean Word Layer Do Korean Buyers Type?
Identify which layer owns your category first, because that choice governs every phrase built on top of it. Sino-Korean compounds carry most industrial, technical, legal and administrative category nouns. Loanwords carry much of consumer technology, fashion, beauty and food. Native Korean words carry everyday actions and qualities.
Korea’s own language policy shows how durable the loanword layer is. The National Institute of Korean Language runs a refined-word program that proposes plain Korean replacements for difficult loanwords and foreign terms, and its published list held 18,338 entries when checked on 12 September 2026, including 파운드리 (foundry) replaced by 반도체 위탁 생산, 팹리스 (fabless) by 반도체 설계 전문 회사, and 디지털 리터러시 (digital literacy) by 디지털 활용력. The market frequently keeps the loanword anyway, so both members of those pairs belong in your research, and volume data settles which one you write for.
The failure running the other way costs more. Transliterating an English word into hangul does not create a Korean term, and machine translation does it constantly. Autocomplete is a suggestion index rather than a volume measure, it varies by date, region and login state, and the checks quoted here were run logged-out on one date to show the shape of the gap rather than to size it. Naver’s search autocomplete, checked on 12 September 2026, returned a full commercial suggestion set for 세탁기 (washing machine), covering drum, rental and recommendation variants, while 워싱머신, the transliteration of “washing machine”, returned two suggestions. 에어퓨리파이어, the transliteration of “air purifier”, returned none, while 공기청정기 returned the category’s full set. A page optimized for the transliteration is competing for a fraction of the category’s demand, and the volume tools will show how small.
Business-to-business categories fail the same way at higher cost. 세미컨덕터, the transliteration of “semiconductor”, returned only company names on that same check, led by 온세미컨덕터 (onsemi): suggestions that contain the string rather than category queries built on it. 반도체 장비 (semiconductor equipment) returned the category’s own suggestions. Korean engineers do use English loanwords heavily inside process vocabulary, a separate matter covered in the guide to technical documentation for Korean fabs, and the category noun a buyer searches is usually Sino-Korean.
How Do Spelling Variants Split Your Demand, and Why Doesn’t Spacing?
Korean orthography licenses more than one correct written form of the same term, so the “correct” keyword a vendor delivers is one member of a set. The two kinds of variant behave differently at the engine: spacing forms of one compound resolve to the same search, and spelling forms do not.
The Hangul Orthography standard (문화체육관광부 고시 제2017-12호) states in Chapter 1, Article 2 that each word in a sentence is written with spaces. Article 49 then allows proper nouns other than personal names to be spaced by unit as well as by word, giving 대한 중학교 and 대한중학교 as equally admissible, and Article 50 does the same for technical terms, with 만성 골수성 백혈병 and 만성골수성백혈병. A category term of two or three words therefore has at least two legitimate written forms before anyone searches anything.
Spacing is the variant that does not split demand. On the 12 September 2026 autocomplete check, 공기청정기 and 공기 청정기 returned an identical suggestion list, and every suggestion inside it was written with no spaces at all, which is how an engine that normalizes spacing before matching looks from outside. Write the closed-up form, hold it across every Korean surface, and spend the research budget on the spelling forms.
Loanword spelling splits the same way. The Loanword Orthography standard (문화체육관광부 고시 제2017-14호) sets out in Chapter 1, Article 5 that established loanwords respect customary usage, with the scope and examples set separately, and the institute’s searchable example list held 67,270 entries on 12 September 2026. Where the standard form and the popular form differ, both carry activity: 메시지 and 메세지 returned different autocomplete suggestion sets on that check, as did 액세서리 and 악세사리. Differing suggestion sets are the signal to check both spellings in the volume tools before choosing a primary.
When Do Korean Buyers Type English in Latin Script?
Koreans type Latin script when the term is an acronym, a standard, a model number or a brand name. A translated list can never contain this part of the keyword universe, because translation is the step that removes it.
The 12 September 2026 autocomplete check shows the pattern. ERP returned ERP뜻, 중소기업ERP (ERP for small and medium enterprises) and ERP정보관리사, an ERP certification. SaaS returned saas 뜻, saas 기업 and saas 솔루션, and CRM returned crm 마케팅. These are Latin-script stems carrying Korean modifiers, written as a single query string, closed up or spaced, and no translation process produces them.
Intent can also split across the two scripts. On the same check, 노트북 returned buying-stage suggestions including recommendation and gaming variants, while laptop returned laptop 뜻 and laptop 가방 (laptop bag). The hangul term carried purchase intent and the Latin term carried definition and accessory queries, which belong on different pages.
Brand names deserve one decision, made once and held. Korean policy runs the same way: the Romanization of Korean standard (문화체육관광부 고시 제2014-42호) states in Chapter 3, Article 7 that personal, company and organization names may keep the spellings they have been using, and the directive on translating public terms (문화체육관광부훈령 제584호, in force 1 July 2026) instructs at Article 6 that an institution’s official English name already in use is the one followed, citing 호텔신라 as The Shilla. Fix your hangul rendering once and research it alongside the Latin form. Branded queries resolve into Naver Brand Search placements, and Latin-script technical demand concentrates on the engine sized in the guide to Google SEO in Korea.
How Does Korean Query Intent Differ From the Source Market?
Korean commercial queries carry a recognizable modifier set, and it does not map one to one onto English buying-stage language.
The autocomplete suggestions returned on 12 September 2026 for consumer categories cluster on 추천 (recommendation), 순위 (ranking) and 렌탈 (rental): 공기청정기추천, 공기청정기순위, 공기청정기렌탈, 세탁기추천, 세탁기렌탈 and 노트북추천. Two things follow. The English comparison query that opens with “best” is served in Korean by recommendation and ranking phrasings, which is a different content format. The rental modifier exposes a purchase model running through several Korean appliance categories that may not exist in your home market, a business-model signal a translated list would delete.
Business-to-business head terms carry a trap of their own. 반도체 장비 returned investor phrasings on the same check: 반도체 장비 대장주 (leading stock), 반도체 장비주 (equipment stocks) and 반도체 장비 etf. A category head term dominated by stock-market intent puts a supplier page in competition with finance content, and the commercially useful demand sits further down the tail. How Korean buyers research and validate is covered in the guide to Korean consumer behavior, and the engine split in the Naver vs Google decision guide.
How Should an Executive Brief and QA a Localization Vendor?
Brief the concept and the buyer, and require a variant set back. A brief that hands over an English keyword spreadsheet and asks for a Korean column has already decided the outcome.
Specify five things: the concept in plain language with the buyer and the buying stage attached, a requirement that seeds originate in Korean from a native speaker, a variant set per concept covering word layer, spacing and spelling forms and any Latin-script form, an intent label on every phrase, and a stated primary for each page. Volumes and competition for those variants come from Naver’s own tools through the workflow in Naver keyword research.
Most of the result can be checked before a Korean speaker sees it. A list with the same number of rows as your English list has not been researched. Confirm that spacing or spelling variants appear, that some entries contain Latin characters, and that no entry is a hangul transliteration of an English word with an ordinary Korean equivalent. Ask for the rejected variants and why each was dropped, because a vendor doing the work holds that file. Then have a Korean speaker who did not build the list read the top twenty phrases and say what they expect to find behind each.
Korea’s public-sector translation directive separates 음역, translating by sound, from 의미역, translating by meaning, defines both at Article 2, and defers to widely used existing translations. A vendor who cannot say which of the two it applied to each term is guessing.
Frequently Asked Questions
Can machine translation produce a usable Korean keyword list? No. Machine translation returns one Korean equivalent per English phrase, and its habit with unfamiliar terms is transliteration into hangul. The 12 September 2026 autocomplete check showed transliterations such as 워싱머신 and 에어퓨리파이어 carrying almost no suggestion activity, while the Sino-Korean equivalents carried the whole category. Treat machine output as a starting glossary and never as a research list.
Should we target the loanword or the Sino-Korean term? Research both and let Naver’s volume data decide, because the answer is category specific. Consumer technology, fashion and beauty often sit in the loanword layer, while industrial, technical and administrative categories sit in Sino-Korean. Government recommendations point one way and usage often points the other: the National Institute of Korean Language’s refined-word list held 18,338 plain Korean replacements on 12 September 2026.
Do spacing variants really need separate treatment? They need one documented decision rather than separate research. Korean orthography admits both the spaced and the closed-up forms of compound proper nouns and technical terms under Articles 49 and 50 of the Hangul Orthography standard. Naver’s autocomplete returned an identical closed-up suggestion set for both the spaced and the unspaced query when checked on 12 September 2026, so write the closed-up form and keep it consistent across every Korean surface.
Does keyword localization matter for Google in Korea too? Yes, identically. Google’s Korean users type the same Korean phrases, so a translated list misses demand on both engines. The difference is composition: Google’s Korean share concentrates in desk-based, technical and English-inflected queries, which raises the value of the Latin-script acronyms and model numbers inside your variant set. Build one Korean phrase universe and target it on both engines.
Keyword localization is the cheapest decision to get right in a Korean program and the most expensive to discover late, because every page, ad group and title inherits it. Inquivix runs Korean keyword research, content and paid search for foreign brands from Seoul, and sets out the execution side in its guide to localizing a global brand for the Korean market. To review the Korean phrase set your current program is built on, write to joon@joonklee.com.
