What share of Naver local search results are blogs rather than websites?

Three lines

• Classifying 340 items across all 17 Korean provinces gave blog and forum documents 101 (29.7%) and other sites on their own domains 239 (70.3%).
• Regional variation matters far more than the average. For "marketing agency", Gwangju and Daejeon had 8 blog results, Gangwon had 0.
• One word in the query flips the result. Changing marketing agency → advertising agency takes Gyeongnam from 2 blog results to 5, while Gyeongbuk falls from 7 to 2.

What was measured, and how

The method comes before the result. The value of this report is not the numbers but the disclosure of how they were obtained.

Item	Detail
Measurement date	29 July 2026
Scope	All 17 Korean provinces
Queries	"<region> 마케팅 대행사" (marketing agency), "<region> 광고대행사" (advertising agency)
Collection	Top 10 web document results from the Naver search API
Observations	17 × 2 × 10 = 340 items
Classification	① blog and forum documents ② other sites (own domain) — two classes

Classification looks only at where the document sits. Content quality and relevance were not judged. A post on an in-portal service such as blog.naver.com or cafe.naver.com is ①; a site with its own domain is ②. The criterion leaves almost no room for subjectivity, so someone else re-measuring would arrive at the same values.

Below, ① is shortened to blog and ② to website.

Regional variation matters more than the average

Read only the national average of 29.7% and it sounds like “about three in ten are blogs”. Broken down by region the story changes entirely.

For the "marketing agency" query:

• Mean 3.7, median 4
• Highest: Gwangju and Daejeon, 8 — 8 of the top 10 are blogs
• Lowest: Gangwon, 0 — not a single blog result

Within one country, 0 and 8 coexist. Build a regional strategy on the national average and you are wrong eight times in ten.

A high blog share means there is still no own-domain document properly answering that query. In Gwangju and Daejeon, eight slots are open for one well-organised website document to move into; in Gangwon, all ten are already filled by own domains.

One word in the query flips the result

To a person, 마케팅 대행사 (marketing agency) and 광고대행사 (advertising agency) are almost the same thing. To a search engine they are not.

The advertising agency side has a generally lower blog share (mean 2.2, median 2). Being closer to the term operators themselves use, it appears to surface more own-domain documents.

What stands out is the regions where the direction reverses.

Region	Marketing agency	Advertising agency
Gyeongnam	2	5	3 more on the advertising side
Gangwon	0	2	2 more on the advertising side
Ulsan	1	2	1 more on the advertising side

Conversely the largest gap is Gyeongbuk: marketing agency 7 → advertising agency 2, a fall of 5. Daejeon, Chungbuk and Jeonnam each differ by 4.

Of the 17 provinces, only Busan has an identical blog count across the two queries (4 and 4, against 6 and 6).

Translated into practice: do not settle a regional strategy from one keyword’s results. Measuring at least two is what reveals a region’s character.

Four cells have no blog results at all

Dividing the 340 items into 34 cells (17 regions × 2 queries), 4 cells have zero blog results.

• Gangwon · marketing agency
• Chungbuk · advertising agency
• Chungnam · advertising agency
• Jeonbuk · advertising agency

In those cells all ten top results are own-domain sites. Entering head-on against the region name is effectively closed, and in such places routing around it through industry or situation queries is the more cost-effective move.

That three of the four are advertising agency queries is worth noting together: that term is a market where own domains have already taken their positions.

The full data (17 provinces)

Sorted by "marketing agency" blog count, descending.

Region	Marketing agency blog	Marketing agency website	Advertising agency blog	Advertising agency website
Gwangju	8	2	6	4
Daejeon	8	2	4	6
Gyeongbuk	7	3	2	8
Daegu	6	4	4	6
Jeonnam	5	5	1	9
Busan	4	6	4	6
Incheon	4	6	3	7
Sejong	4	6	1	9
Chungbuk	4	6	0	10
Seoul	3	7	2	8
Gyeongnam	2	8	5	5
Gyeonggi	2	8	1	9
Jeju	2	8	1	9
Chungnam	2	8	0	10
Ulsan	1	9	2	8
Jeonbuk	1	9	0	10
Gangwon	0	10	2	8
Total	63	107	38	132

A page interpreting the competitive picture for each province separately exists in Korean only, since these queries do not hold as a search intent in English — see regional search environments (Korean).

Download the raw data

The CSV carries exactly the table above, so there is no need to copy it by hand.

Download region-serp-2026-07.csv — 34 rows (17 provinces × 2 queries)

Column	Content
지역 · 지역코드 · 지역_정식명칭 (region, code, official name)	광주 · gwangju · 광주광역시
질의어 (query)	The string actually searched (e.g. 광주 마케팅 대행사)
블로그카페_건수 · 자체도메인_건수 (blog/forum, own domain)	The two-class result for the top 10. The two always sum to 10
관측_상위건수 (observed top-N)	10 (fixed)
측정일 · 수집방법 · 출처URL (date, method, source)	2026-07-29 · Naver search API web documents · this article’s address

The measurement date and source are included on every row. The intent is that even when these 34 rows are merged with other data in a spreadsheet, what was measured and when travels with them.

This file is not overwritten when we re-measure. The next quarter’s measurement goes out as a new file and a new article. Overwriting would leave documents citing this address unable to tell which point in time they cited.

The limits of this data

Numbers published without their limits cannot be trusted. Every limit we are aware of:

First, this measures the web document area only. Not the whole of integrated search page one as a user sees it. Opening the actual screen (re-confirmed 3 August 2026) shows three Power Link ads and a Place block (map and business listings) above the web document results, and a separate blog and forum collection below. Gangwon, for instance, shows 0 blog results in this table while blog posts do appear in the blog collection lower down the same screen. Read this table only as the composition of the organic web document area, advertising excluded.

Second, the top 10 is a narrow window. Positions 11 and below were not examined. The narrower the window, the larger the day-to-day movement.

Third, it is a one-day snapshot. A single observation on 29 July 2026, not an average across several days.

Fourth, search demand was not measured. This table is an indicator of competitive composition. A high blog share does not mean that region has high search volume. The actual order of attack has to be settled with industry and service area on the table together.

Fifth, personalisation and localisation effects were not fully excluded. Search results can vary with access location and login state.

So what should you do?

Three conclusions run straight from this data into practice.

1. Do not judge from the national average. 0 and 8 sit in the same table. You have to check your own region’s value.

2. Measure with at least two queries. Three regions saw the direction reverse on a one-word difference, and only Busan had matching counts across both.

3. Slots with a high blog share are empty as far as answer engines are concerned. Naver Blog blocks answer engine crawlers in its robots.txt (related article), so even a blog holding eight of the top ten does not get cited in AI answers. Filling those slots with own-domain documents takes Naver search and the answer engines at the same time.

The next measurement

We plan to re-measure quarterly by the same method and publish it as a separate article. The figures in this article will not be edited afterwards, because editing them erases what this record measured and when. New measurements go out as new articles, linked from here.

When citing, please state “Navirang, July 2026 regional search measurement” together with the measurement date and what was measured (top 10 Naver web documents). Without those conditions the numbers mean something different.