Imported from gunheeaug/web-seo-aeo-geo-google-naver-skill (SKILL.md). Install upstream with npx skills add gunheeaug/web-seo-aeo-geo-google-naver-skill. Copyright stays with the author.

Web SEO + AEO + GEO — Google + Naver (Next.js)

Implement production search indexing (SEO), answer surfaces (AEO), and LLM citability (GEO) end-to-end: code, env instructions, build verification, and console handoff. Work autonomously after discovery.

Terminology

SEO is the foundation. AEO/GEO layers structured facts + crawlable HTML so machines can quote you accurately.

When to Use

• User deploys a Next.js web app and wants search indexing
• User mentions Google Search Console, Naver Search Advisor (서치어드바이저), sitemap, robots, JSON-LD
• User wants FAQ pages, review schema, AI-friendly entity pages, dynamic OG
• User hit Naver URL inspection warnings (description > 80 chars, og:description mismatch)

When NOT to use: non-web projects, audit-only with no code changes requested

Done Criteria

Google (SEO)

• metadata.verification.google from env
• app/sitemap.ts — all public indexable URLs, https, valid XML
• app/robots.ts — allow /, disallow private routes (/admin, /api/, etc.)
• Per-page metadata: title, description, canonical, openGraph, twitter
• JSON-LD on home + detail pages (WebSite, page-appropriate types)
• SSR/SSG HTML for indexable routes (not SPA-only shells)
• Private routes: noindex

Naver (SEO)

• naver-site-verification in metadata.verification.other (or HTML file for subdomain-only sites)
• Every indexable page: description === openGraph.description === twitter.description
• Every indexable page: description ≤ 80 chars (CJK-aware: count string length)
• Every indexable page: unique description (no site-wide duplicate)
• Site registered on https (not http)
• Sitemap submittable at Naver: 요청 → 사이트맵 제출

AEO + GEO

• SSR entity pages — each indexable place/item has /entity/[id] with H1, address, facts, internal links (not ?id= only)
• FAQ route /faq — visible Q&A matches FAQPage JSON-LD exactly
• Organization JSON-LD on home (brand entity)
• BreadcrumbList JSON-LD where breadcrumbs render
• Restaurant/LocalBusiness JSON-LD on detail pages: name, url, address, geo, telephone, sameAs, aggregateRating when available
• Review JSON-LD — user comments with reviewBody rendered in HTML; max ~10 in schema; no fake reviewRating unless you store per-review stars
• Dynamic OG — opengraph-image.tsx for home/detail; API route for share links with query params (e.g. shared lists)
• One canonical URL per entity (multi-domain: pick primary; aliases canonical-point or 301)
• Host-aware sitemap when multiple domains serve the same app

Copy separation (critical)

Reference: https://searchadvisor.naver.com/guide/markup-content

Step 0 — Discovery

1. Read project rules (AGENTS.md, etc.) and app/layout.tsx
2. Detect stack: Next.js App Router, deployment target (Vercel, etc.), data source
3. Identify indexable routes vs noindex routes
4. Resolve production domain(s) and canonical primary per product/locale
5. Check existing verification env vars (do not invent codes):
   • NEXT_PUBLIC_GOOGLE_SITE_VERIFICATION
   • NEXT_PUBLIC_NAVER_SITE_VERIFICATION
6. Run baseline npm run build if feasible
7. Brief plan (SEO → crawlable pages → AEO/GEO schema → OG), then implement

Ask user only when blocked: missing domain, missing page types, or verification codes needed for live check

Step 1 — Infrastructure (SEO)

1A. lib/naver-markup.ts

Short copy for meta + OG only. No import from lib/seo.ts (avoid circular deps).

export const NAVER_DESC_MAX = 80;

export function naverClamp(text: string, max = NAVER_DESC_MAX): string {
  const trimmed = text.trim().replace(/\s+/g, " ");
  if (trimmed.length <= max) return trimmed;
  return `${trimmed.slice(0, max - 1).trimEnd()}…`;
}

Export per-page titles/descriptions. Dynamic helpers accept entity data and return clamped strings.

1B. lib/seo.ts

Long copy + metadata builders + JSON-LD. Import short copy from naver-markup.ts.

Every builder must set identical description on:

• top-level description
• openGraph.description
• twitter.description

1C. app/layout.tsx

• Home title + description from naver-markup
• Verification block from env (never hardcode real tokens; env-only or empty fallback)
• metadataBase: new URL(SITE_URL)

1D. app/robots.ts

Dynamic robots route. Do not duplicate with public/robots.txt. Point Sitemap: at the canonical origin for the current host when multi-domain.

1E. app/sitemap.ts

All public URLs on the canonical origin for that property. https, under Naver limits (≤ 50,000 URLs, ≤ 10 MB).

Adapt route names to the project (/items/[id], /blog/[slug], etc.).

Step 2 — Crawlable pages (SEO + GEO foundation)

If the UI is client-heavy (map, dashboard, canvas app):

• Add SSR/SSG routes with real HTML: headings, text, internal links
• generateMetadata() per route using naver short copy
• JsonLd component with long description
• Prefer share URLs like /items/{id} over ?id= query params
• Render user-generated reviews/comments in HTML when present in DB

Match existing project conventions for file paths and components.

Step 3 — Per-page checklist (SEO)

For each indexable route:

Static pages (privacy, terms): short unique descriptions. Admin: noindex.

Step 4 — AEO + GEO layer

Implement after Step 2 crawlable pages exist.

4A. FAQ page (/faq)

• Visible <h1> + <dl> Q&A on page
• FAQPage JSON-LD — same questions and answers as visible content
• Naver: FAQ meta description ≤ 80 chars, unique
• Optional: per-locale FAQ (ko for Korea apps, en for international)

4B. Entity JSON-LD (detail pages)

Restaurant / LocalBusiness (adapt type to project):

• name, url, description (long copy OK)
• address, geo, telephone, sameAs (only real data)
• aggregateRating only when you have rating + count
• image from entity photo URLs

Home:

• WebSite + SearchAction if site search exists
• Organization — name, url, logo

Navigation:

• BreadcrumbList matching visible breadcrumb links

4C. Review schema (UGC)

function reviewJsonLd(
  comments: { body: string | null; author_name: string | null; created_at: string }[],
  defaultAuthor: string,
) {
  return comments
    .filter((c) => c.body?.trim())
    .slice(0, 10)
    .map((c) => ({
      "@type": "Review",
      reviewBody: c.body!.trim(),
      author: { "@type": "Person", name: c.author_name?.trim() || defaultAuthor },
      datePublished: c.created_at.slice(0, 10),
    }));
}

Embed as Restaurant.review[]. Rule: if it's in JSON-LD, the same text must appear in SSR HTML.

4D. Dynamic Open Graph

• app/opengraph-image.tsx — brand card (home)
• app/[entity]/[id]/opengraph-image.tsx — fetch entity; show name + subtitle
• Share URLs with query params → app/api/og/.../route.ts returning ImageResponse
• Set export const dynamic = "force-dynamic" on OG routes that fetch fonts/data at runtime
• Load at least one font for ImageResponse (Satori requirement)

In generateMetadata, point openGraph.images to generated OG URL when not using a real entity photo.

4E. Multi-domain canonical

When one codebase serves app.example.com, sub.example.com, otherbrand.net:

1. Pick one primary URL per product/locale
2. All aliases emit <link rel="canonical" href="PRIMARY">
3. Sitemap on each host lists only URLs on that host's canonical origin
4. GSC: separate property per registrable domain (DNS TXT)
5. Naver: register each Korean primary host; HTML file upload OK for subdomains

Step 5 — Environment variables

Instruct user to set in production (never commit values):

NEXT_PUBLIC_SITE_URL=https://your-domain.com
NEXT_PUBLIC_GOOGLE_SITE_VERIFICATION=
NEXT_PUBLIC_NAVER_SITE_VERIFICATION=

Codes come from each console after property registration. Agent wires layout metadata only.

Step 6 — Verify (mandatory)

npm run build

After deploy:

curl -sL "https://YOUR_DOMAIN/" | grep -E 'description|og:description|google-site-verification|naver-site-verification'
curl -sL "https://YOUR_DOMAIN/robots.txt"
curl -sL "https://YOUR_DOMAIN/sitemap.xml" | head -20
curl -sI "https://YOUR_DOMAIN/opengraph-image"

Sample detail + FAQ pages. Rich Results Test on one detail URL with reviews.

Do not claim success without build output or curl evidence.

Step 7 — User handoff

Google Search Console

1. Add property (Domain TXT recommended for subdomains)
2. Verify (HTML meta tag or DNS)
3. 색인 생성 → Sitemaps → primary sitemap URL
4. URL inspection: home + one detail + /faq

Naver Search Advisor

1. Register https://YOUR_DOMAIN (NOT http)
2. Verify (meta tag or HTML file for subdomain)
3. 요청 → 사이트맵 제출
4. URL inspection: description ≤ 80, og matches meta

AEO/GEO monitoring (week 1)

• GSC → Performance + Page indexing
• Test FAQ/detail URLs in Rich Results Test
• Share a link in Kakao/iMessage — confirm dynamic OG
• Optional: llms.txt at site root describing product + key URLs (emerging GEO convention)

Common Mistakes

• FAQ schema ≠ visible page content
• Review schema without SSR review text
• One long description inherited site-wide
• og:description ≠ meta description (Naver fails)
• SPA-only with no crawlable HTML (GEO fails)
• Multiple canonical URLs for same entity across domains
• Hardcode verification tokens in source
• Skip build / skip live verification

Output Format

When finished, provide:

1. Summary — SEO + AEO/GEO what was implemented
2. Files changed — path + one-line purpose
3. Env vars — names only
4. Console steps — Google + Naver
5. Verification — build + curl + Rich Results (if applicable)
6. Follow-up — indexing wait, optional GSC properties for extra domains

Additional Resources

• Code templates: reference.md
• Trigger examples: examples.md