12 Core SEO·Search-Engine Concepts — sandbox · E-E-A-T · DA · CTR · AI Overview on One Page

Investigate indexing, ranking and impressions separately. A new domain alone does not establish a fixed waiting period or automatic recovery.

About a month after migrating the domain (taystudio.github.io → taystudios.com), I opened GSC and saw 2 clicks and 777 impressions. Honestly, I looked at those numbers and assumed something was broken with the site, and it threw me off for a while. Then Cloudflare showed a steady ~55 visits a day, and once I registered Naver SearchAdvisor I found 26,000 impressions and 230 clicks over the month.

In the end, I'd just panicked after looking at one side of the data.

This post collects the 12 SEO concepts I actually bumped into over that month. I'll cover the traps new-domain operators most commonly fall into, along with the quirks specific to the Korean market.

The basic flow

Site → [Crawl] → [Index] → [Rank] → Search results

If any one of these stages stalls, traffic goes to zero. After our 5/9 migration there was almost no indexing until 5/22, and then it started in earnest from 5/22 — about 13 days in total. In my experience, a fresh domain is simply slow to get indexed in the first place.

1. Google Sandbox — not a diagnosis

An informal label for weak search visibility on new sites. These observations do not establish a particular algorithm or its application to this site. Check the URL response, indexing directives, canonical URL and latest crawl first.

Google’s recrawl guidance explains that requesting a crawl does not guarantee indexing.

2. E-E-A-T — Google's four quality pillars

Experience · Expertise · Authoritativeness · Trustworthiness

This matters especially for YMYL (Your Money, Your Life) content — taxes, medical, legal, finance — which Google evaluates very strictly. A single wrong number can be enough to drag down the trust score of the whole site.

More than half of our 38 calculators are YMYL (salary · property-tax · inheritance, etc.), so I made it a rule that every number is sourced directly from the government's official sites (law.go.kr, NTS, MOEF). Not a single blog or news outlet is cited. The idea was to shut down the E-E-A-T risk from the start.

3. Domain Authority (DA)

This is a 0–100 score for how much search engines trust a site. It's measured by the likes of Moz and Ahrefs — it's not an official Google score, just an estimate.

The factor that really decides it is backlinks: the quantity and quality of other sites linking to you.

• DA 70+ → Google ranks you high
• DA 0~10 → back of the results
• DA grows only with time + naturally accumulated backlinks

A new domain like ours has a DA close to zero. I suspect some authority carried over from the old taystudio.github.io, but either way it's something that needs 3–6 months to build up.

4. AI Overview — Google's new threat

Since 2024, Google has been showing AI answers directly at the top of search results.

Query: "net salary on 50 million KRW"
↓
AI Overview: "About 3.54M per month, after the 4 social insurances..."
↓
The user never clicks our site

For a calculator site like ours, that's a direct hit. Naver's Korean-policy AI is still fairly weak, so at least there the impact is minor.

5. Zero-Click Search

This is when users search but never click through — they get the answer from the search results page itself.

• Featured Snippet
• Knowledge Panel
• AI Overview
• "60%+ of Google searches are now zero-click" (Similarweb 2024)

The way to counter it is differentiated content that goes beyond a simple fact: simulations, matrices, case studies, visualizations — depth that AI can't condense into a single line.

That's exactly why we attached matrices to our tools (1-home-owner FMV ratio 9×4 / DSR back-calculation / 6+6 split scenarios).

6. The Korean search market

Korean queries — taxes, calculators, policy — lean heavily toward Naver. So judging a site by Google alone is a mistake here, and that's the thing I felt most acutely over a month of running it.

7. Sitemap · robots.txt · llms.txt — the contracts with crawlers

These days the standard is to explicitly Allow GPTBot · ClaudeBot · PerplexityBot · Google-Extended in robots.txt, and that's how I have it set up.

8. Schema.org / JSON-LD — the rich-snippet engine

This is a structured-data vocabulary that search engines understand, and JSON-LD is the most common format for it.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [{
    "@type": "Question",
    "name": "What's the net salary on KRW 50M?",
    "acceptedAnswer": { "@type": "Answer", "text": "About 3.54M per month..." }
  }]
}
</script>

Drop this in and Google starts showing ⭐ ratings, expandable FAQs, and breadcrumbs in the results, which in turn lifts CTR.

9. CTR Optimization — turning impressions into clicks

Showing up doesn't help if no one clicks. The things that drive CTR are these:

• <title> — direct match with the query, and attractive
• <meta name="description"> — core value in ~150 chars
• URL readability — /tools/salary/ > /p?id=12345
• Rich snippets — ⭐, FAQ expand
• Site credibility (HTTPS, domain age)

For us, the combination of a new domain and AI Overview pressure left GSC CTR at 0.3% (normal is 2–5%). On the exact same site, Naver showed 0.9% — a 3× difference. It's a number that confirmed Korean engines match Korean content better.

10. Keywords — short-tail vs long-tail

New sites simply can't win short-tail. The path forward is patiently accumulating long-tails, and the matrices and simulations we keep adding are exactly that long-tail capture strategy.

11. Three analytics tools — looking at one is missing the big picture

Combining all three, the picture came out like this:

• GSC (3 months): 2 clicks / 777 impressions
• Naver (30 days): 230 clicks / 26,000 impressions
• Cloudflare (24h): 57 visits

Naver turned out to be 33–115× GSC. If I'd only looked at Google, I might well have written the site off.

12. GEO (Generative Engine Optimization) — SEO for the AI era

This is about optimizing so that ChatGPT · Claude · Perplexity · Gemini cite your site when they answer.

• llms.txt — markdown meta at site root
• Structured data (Schema.org)
• Source citations (author · date · source)
• Fact-rich content

Cloudflare already shows chatgpt.com as a referrer, which means LLM exposure has already begun. I expect updating llms.txt to speed that up.

Operational checks

Check technical access and content quality separately. Record deployment and Google crawl times, compare the same URLs before and after changes, and do not interpret visibility changes as proof of a fixed waiting period.

Related

• (Coming) GSC vs Naver vs Cloudflare — same site, three different stories
• (Coming) Adding llms.txt — site metadata for the AI-search era