SEO & GEO Checklist

Metadata Basics

• Page <title>: Unique, descriptive, ≤60 characters (fits Google SERP). For React/Next: use dynamic <title> tag or Next.js metadata export.
• Meta description: 120–160 characters, clearly summarizes page value. Should answer "why visit this page?"
• Canonical URL: Prevents duplicate-content penalties. Set <link rel="canonical" href="..."> or via Next.js metadata.metadataBase.
• Open Graph tags: For social sharing and preview cards. Include og:title, og:description, og:image (1200×630 px recommended), og:url, og:type.
• Twitter Card tags: twitter:card, twitter:title, twitter:description, twitter:image for Twitter/X sharing preview.

Structured Data (JSON-LD)

• Format: Use JSON-LD in <script type="application/ld+json"> blocks (preferred by both Google and modern LLM crawlers).
• Schema types: Choose applicable types: Organization (company info), Article (blog posts/news), Product (e-commerce), BreadcrumbList (site navigation), FAQPage (Q&A content).
• Why it matters: Enables Google rich results (star ratings, recipes, job listings) AND helps LLM crawlers understand semantic structure for more accurate citations/excerpts.
• Validation: Test with schema.org examples and Google's Rich Results Test.

Technical SEO

• Sitemap.xml: Submit to search engines via robots.txt or Search Console. For Next.js: use next-sitemap or manually generate and serve from /public/sitemap.xml.
• robots.txt: Define crawl budget and disallowed paths. Basic rule: User-agent: * + Allow: /. For Vercel: robots.txt auto-generated; check /robots.txt.
• Semantic HTML: Single <h1> per page, no skipped heading levels (h1 → h2/h3, not h1 → h4). Use <header>, <main>, <nav>, <article>, <section> landmark elements. Overlaps with accessibility-first HTML — see typescript-conventions Accessibility section.
• Duplicate content: Avoid multiple URLs serving identical content. Use canonical tags to point to preferred version.
• Index coverage is a separate question: Everything above is "is the page well-formed". Whether a URL is actually in Google's index — and why not — is diagnosed differently: Discovered – currently not indexed is a crawl-budget signal, Crawled – currently not indexed is a content-quality verdict, and only GSC URL Inspection reports either. For that, plus the inspection quota arithmetic, stratified sampling, sitemap segmentation for per-family coverage, and the soak-window rule, load seo-index-census.

Core Web Vitals (CWV)

Target thresholds (Google's metric for rank/user experience):

• LCP (Largest Contentful Paint): ≤2.5s. Optimized images, preload critical resources.
• INP (Interaction to Next Paint): ≤200ms. Break long JavaScript tasks, defer non-critical work.
• CLS (Cumulative Layout Shift): ≤0.1. Reserve space for late-loading fonts/ads, set image dimensions.

Common React/Vite culprits:

• Unoptimized images (use <img> with width/height or lazy-load). In Next.js, use next/image for automatic optimization.
• Late-loading web fonts (specify font-display: swap or preload via <link rel="preload">).
• JavaScript bundle size; code splitting with dynamic imports for route-based or intersection-observer-based chunks.
• Hydration cost on initial page load (minimize render work in client-side React, consider streaming SSR).

GEO (Generative Engine Optimization)

• llms.txt: An emerging convention (not universally adopted yet) analogous to robots.txt for AI crawlers. Place at /.well-known/llms.txt or /llms.txt. Specifies which content is quotable/extractable and any usage policies. Example: Allow: /blog/* (LLMs may cite blog posts). Check llms.txt specification for current status before relying on it.
• Quotable content: Structure facts and claims clearly near the top of the page. LLMs prioritize declarative, extractable statements over marketing fluff. Avoid burying key information in prose; use lists, tables, or Q&A formats.
• Q&A sections: Direct question-answer content maps well to LLM-answer-engine excerpting. Pair each Q with a concise A. Markup with <dl> + schema FAQPage or QAPage JSON-LD for best results.
• Byline and metadata: Attribution, publication date, author expertise signals improve trustworthiness in LLM citations.

Next.js-Specific

• Metadata export: Use Next.js 13+ metadata export in page.tsx instead of HTML <meta> tags.

  export const metadata: Metadata = {
    title: "Page Title",
    description: "Page description",
    openGraph: { title: "...", description: "...", images: [{ url: "..." }] },
  };

• Sitemap and robots.txt: Auto-generated by Next.js. Customize in app/sitemap.ts and app/robots.ts (App Router).
• next/image: Always use for images to enable automatic format negotiation, responsive sizing, and lazy-loading (improves CWV).
• Dynamic routes: If using generateStaticParams(), ensure all variants are covered for SEO discoverability.