GSC vs Naver SearchAdvisor vs Cloudflare — Same Site, Three Datasets, Three Stories

TL;DR: For a Korean-targeted site, GSC (Google Search Console) shows you only about 10–20% of the real traffic. The main channels are Naver SearchAdvisor and Cloudflare Web Analytics; I'd treat GSC as the side dish, not the main course.

The trigger — panic over GSC

This was about a month after I migrated to a new domain (taystudios.com). I opened GSC and saw this:

Last 3 months:
- Total clicks: 2
- Total impressions: 777
- Avg CTR: 0.3%
- Avg position: 14.8

Impressions peaked on 5/14 and had dropped to nearly zero by 5/24. I'll admit that when I saw that graph, my first thought was "is the site dead?"

But at the same moment, Cloudflare Web Analytics told a completely different story:

Last 24 hours:
- Page views: 79
- Visits: 55

About 55 people were coming in every day. The two tools seemed to contradict each other head-on, and it took me a while of staring at it to realize the obvious: the two tools measure entirely different things in the first place.

What each tool actually measures

So GSC falling while Cloudflare stays stable simply means that traffic from non-Google sources is holding steady. It wasn't a contradiction at all — the two were just measuring different channels.

Actual numbers — same site, one month

Google Search Console (last 30 days)

Clicks: 2
Impressions: 777
CTR: 0.3%
Avg position: 14.8

The top 5 keywords surfacing the site looked like this:

• "image watermark" (impressions 2, clicks 1, position 8.5)
• "net salary 80M KRW" (impressions 4, position 6.8)
• "Local Tax Act housing property tax rate 2026" (impressions 3, position 10.3)
• "MOIS car acquisition tax 7%" (impressions 2, position 11.5)
• "background remover site" (impressions 2, position 48.5)

Naver SearchAdvisor (last 30 days)

Look at the same month from Naver's side and the scale changes entirely:

Total clicks: 230 (= 115× GSC)
Total impressions: 26,000 (= 33× GSC)
Avg CTR: 0.9% (= 3× GSC)

Impressions spiked from 5/22, settled at an average of 2,000–3,000 a day, and peaked at 7,600 on 5/27.

Cloudflare Web Analytics (24 hours)

Page views: 79
Visits: 55
Country: Korea 53 / China 1 / US 1

Visits by source:
- m.search.naver.com: ~18
- search.naver.com: 16
- search.daum.net: 11
- chatgpt.com: (yes)
- Total Naver+Daum = 45 (79%)

Top page: /tools/baby-formula/ (16 visits)

Finding 1 — Google is a 20% channel for Korean sites

Naver = 33–115× GSC

Google's search share in Korea is usually quoted at 30–35%, yet our Google traffic came in even smaller than that. The three reasons I suspect:

a. Google sandbox (the 6–12 week probation for new domains)

The migration was on 5/9, so as of 6/2 we were at week 3.5. 122 pages were indexed, but rankings still seemed throttled — in my experience a new domain has to pass through this window once before it loosens up.

b. AI Overview kills clicks

A search like "net salary 50M KRW" gets answered directly by Google's AI, so the click never reaches our site. Google's grasp of Korean policy data (Gemini) feels stronger than Naver's here.

c. Korean NLP

For Korean policy queries like "parental leave benefits calculator," Naver matched our pages more accurately. My impression is that Google's Korean NLP is simply weaker than Naver's homegrown model.

The upshot: if you only watch GSC, you miss 80% of the channel mix outright.

Finding 2 — Cloudflare's real data is inside "Visits detail"

The genuinely useful data in Cloudflare Web Analytics isn't on the main screen. You have to click the Visits card and go into the detail view:

Visits by source (Referrers):
- m.search.naver.com
- search.naver.com
- search.daum.net
- chatgpt.com

Hosts:
- taystudios.com

Top paths:
- /tools/baby-formula/: 16
- /tools/salary/: ...
- /image/bg-remove/: ...

Browsers, OS, Country, Device types

This detail is where the real sources show up. On a day when GSC reads "0 Google clicks," Naver mobile and desktop plus Daum were still bringing in more than 50 visits combined.

What caught my eye most was the chatgpt.com referrer. That's a signal that LLM search is starting to cite the site — and it's something neither GSC nor Naver will ever show you.

Finding 3 — Top-page mismatch

The interesting case was the baby-formula page. It doesn't show up in GSC at all, yet it was the #1 visited page in Cloudflare. It turned out Naver's "infant formula amount" queries matched it well, and because that Naver traffic never appears in GSC, baby-formula was entirely invisible from Google's angle.

So if you want to know which pages are actually popular, you have to look at Cloudflare. The GSC data really is just the tip of the iceberg.

How to use each tool

After running all three side by side, here's how I'd describe each one's job.

GSC — for Google evaluation + keyword discovery

• Which queries surface the site
• Indexing status (how many pages indexed)
• Average position
• Per-page index exclusion reasons
• Don't use it as a traffic gauge — Korean sites only see the tip

Naver SearchAdvisor — the main channel for Korean users

• Most important for Korean-targeted sites
• Sitemap submission, manual page collection requests
• Impression/click trends (covers almost all Korean traffic)
• robots and page diagnostics
• Without registration, the engine only finds the site by chance — very slow

Cloudflare Web Analytics — real traffic + sources

• All channels combined (Direct + every engine + referral)
• Sources detail — where visitors came from
• Top paths — the actually popular pages
• Country, device, browser breakdown
• LLM referrers (chatgpt.com · perplexity.ai) starting to appear
• Privacy-friendly — no personal tracking

Our action priorities

Once I combined all three datasets, the order I'd tackle things in came out like this:

1. Register Naver Webmaster + submit sitemap — biggest ROI (79% of traffic)
2. Request indexing via GSC URL Inspection on changed pages — partially route around sandbox
3. Bing Webmaster (1-click GSC import) — +1–3%
4. Daum Webmaster — +5–10%
5. Check Cloudflare sources regularly — track the LLM-referrer trend

One-liner

Korean sites can't be evaluated on GSC alone. Naver + Cloudflare, side by side.

GSC shows you "Google search data," and nothing more. For a Korean-targeted site the real traffic is Korean engines plus AI search plus Direct — that's the main current. Judge the site on a single dataset and you'll miss the picture, and honestly I think that's the most common trap new-domain operators fall into.

Related

• 12 Core SEO·Search-Engine Concepts — sandbox · E-E-A-T · DA · CTR · AI Overview
• (Coming) Adding llms.txt — site metadata for the AI-search era