nextjs-seo skill

Next.js SEO Optimization

Start with evidence and useful content

Read the installed Next.js version and relevant node_modules/next/dist/docs/ guides before changing framework code; use current official docs when local docs are unavailable. Use Vercel MCP documentation search for platform behavior when available, but check retrieved examples against the installed framework. Do not copy a legacy cache API from a search snippet into a newer app.

Give the H1, subtitle and section headings distinct jobs; do not repeat the same phrase in each for SEO. Keep UI copy concise, concrete and useful.

• Solve the reader's task first. A short useful page does not need padding.

or TL;DR just to make a page look optimized. Add content only when it answers a real question. Preserve necessary explanations and disclosures.

• Do not add generic introductions, keyword blocks, decorative badges, a FAQ

maintainable value. A swapped name/logo and duplicate list is insufficient.

• Create team, location or category pages only when each offers distinct,

leave pre-launch copy on an active service or change timestamps for freshness.

• Describe seasons, availability and update times from current data. Do not

traffic, not indexing status, Google's canonical choice or CWV. Use the relevant report or URL Inspection; mark unavailable checks as unverified.

• Separate evidence: a Search Console performance export shows queries and

indexing, rich-result or AI-citation guarantees.

• Report findings and verification before predicting impact. No ranking,

Quick SEO Audit

Run this checklist for any Next.js project:

1. Check robots.txt: curl https://your-site.com/robots.txt
2. Check sitemap: curl https://your-site.com/sitemap.xml
3. Check metadata: View page source, search for and
4. Check JSON-LD: View page source, search for application/ld+json
5. Check Core Web Vitals: Use PageSpeed Insights (pagespeed.web.dev) and the Search Console CWV report for field data — Lighthouse is lab-only and can't measure INP

Essential Files

app/layout.tsx - Root Metadata

import type { Metadata, Viewport } from 'next';

// Viewport must be a separate export. `themeColor`, `colorScheme` and
// `viewport` inside `metadata` are dropped: Next.js 16 emits no tag, only an
// "Unsupported metadata" build warning.
export const viewport: Viewport = {
  width: 'device-width',
  initialScale: 1,
  maximumScale: 5,
  userScalable: true,
  themeColor: [
    { media: '(prefers-color-scheme: light)', color: '#ffffff' },
    { media: '(prefers-color-scheme: dark)', color: '#0a0a0a' },
  ],
};

export const metadata: Metadata = {
  metadataBase: new URL('https://your-site.com'),
  title: {
    default: 'Site Title - Main Keyword',
    template: '%s | Site Name',
  },
  // ~150-160 chars is a guideline, not a limit — Google truncates per device/query
  description: 'Compelling description with target keywords',
  // No `keywords` field: Google ignores the keywords meta tag entirely
  openGraph: {
    type: 'website',
    locale: 'en_US',
    url: 'https://your-site.com',
    siteName: 'Site Name',
    title: 'Site Title',
    description: 'Description for social sharing',
    images: [{ url: '/og-image.png', width: 1200, height: 630, alt: 'Site preview' }],
  },
  tw

app/sitemap.ts - Dynamic Sitemap

import type { MetadataRoute } from 'next';

export default async function sitemap(): Promise<MetadataRoute.Sitemap> {
  const baseUrl = 'https://your-site.com';
  const posts = await getPosts(); // your CMS/DB

  return [
    {
      url: baseUrl,
      images: [`${baseUrl}/og-image.png`], // Image Sitemap entry
    },
    { url: `${baseUrl}/about` },
    ...posts.map((post) => ({
      url: `${baseUrl}/blog/${post.slug}`,
      lastModified: post.updatedAt, // real content timestamp
    })),
  ];
}

lastModified must reflect the content's actual last change (CMS updatedAt, file mtime, git commit date) — Google uses lastmod only when it's consistently accurate, and new Date() on every build marks everything "just changed", which teaches Google to ignore it. Skip changeFrequency and priority: Google ignores both.

app/robots.ts - Robots Configuration

import type { MetadataRoute } from 'next';

export default function robots(): MetadataRoute.Robots {
  const baseUrl = 'https://your-site.com';

  return {
    rules: [
      {
        userAgent: '*',
        allow: '/',
        disallow: ['/api/', '/admin/'],
        // Never disallow /_next/: crawlers need render-critical CSS/JS.
        // A named group (Googlebot, OAI-SearchBot…) ignores these rules; see mistake 11.
      },
    ],
    sitemap: `${baseUrl}/sitemap.xml`,
  };
}

host was omitted intentionally — it's a non-standard directive Google ignores. Use canonical URLs / 301s to declare the preferred host instead. See references/sitemap-robots.md.

app/manifest.ts - Web App Manifest

Same MetadataRoute family as sitemap/robots, placed at the root of app/. Not an SEO requirement — a PWA-completeness nicety with no ranking effect; skip it unless the site is (or may become) a PWA. Full example in references/metadata-api.md.

OG / Twitter Images

Three ways to set social images — prefer the file conventions over hand-syncing URLs in the metadata object:

1. External URL in metadata (the openGraph.images / twitter.images examples above) — fine for externally hosted images.
2. Static file convention (recommended default): drop opengraph-image.(jpg|jpeg|png|gif) and/or twitter-image.* into a route segment (app/opengraph-image.png for the root, app/blog/opengraph-image.png for /blog). Next.js auto-emits og:image/twitter:image + :type/:width/:height. A deeper, more specific image overrides one above it. Add alt text with a sibling opengraph-image.alt.txt. Build fails if the file exceeds 8 MB (OG) / 5 MB (Twitter).
3. Dynamic generation with ImageResponse (per-page/per-post images): an opengraph-image.tsx in the route segment exporting alt, size, contentType and a default Image({ params }) (params is a Promise in v16) that returns new ImageResponse(, { ...size }). Renders via Satori — flexbox only, no display: grid, 500 KB bundle cap (JSX, fonts, images), fonts ttf/otf/woff only; statically optimized at build time unless it reads request-time data. Full example, fonts, generateImageMetadata and the favicon/icon.tsx/apple-icon conventions: references/metadata-api.md.

Key Principles

Cache Components & SEO

With cacheComponents: true, use "use cache" for data/components that can be shared and whose freshness requirements permit caching. Static content needs no extra cache directive merely for SEO:

// app/(home)/sections/hero-section.tsx
import { cacheLife, cacheTag } from "next/cache";

export async function HeroSection() {
  "use cache";
  cacheLife("hours");   // SEO content that changes a few times/day; see profiles below
  cacheTag("hero");     // Invalidate via updateTag("hero") in a Server Action

  const data = await fetchData();
  return <div>{/* SEO-visible content */}</div>;
}

Choose cacheLife from the product's freshness requirements and the installed Next.js documentation. Do not infer a cache lifetime from the page category; marketing and legal pages also need timely publication and invalidation.

Key rules:

• "use cache" must be the first statement in the function body (or at the top of the file for file-level caching)
• No cookies()/headers()/runtime searchParams inside a plain "use cache" scope. Keep public content separate from personalized data; do not accidentally cache private information for all users. Check the installed docs before using experimental private caching.
• Invalidate with updateTag("hero") inside a Server Action (read-your-writes; it throws outside one), or revalidateTag("hero", "max") from a Route Handler / webhook (the one-argument form is deprecated)
• Route segment exports revalidate, dynamic and fetchCache error with Cache Components enabled; without it, follow the installed version's cache model. Verify next build, the served content and publish-time invalidation.
• A database read does not by itself make a sitemap update after deployment. Configure and verify its refresh/invalidation path. Metadata can be static or runtime-dependent; inspect the route rather than assuming.

Rendering Strategy for SEO

These are rendering trade-offs, not ranking tiers. A Client Component can still be server-prerendered; "use client" does not mean its content is absent from HTML.

Core Web Vitals Targets

each metric separately; this does not mean the same 75% of visits pass all three simultaneously. Lighthouse is a lab diagnostic, not an INP field score.

• Use the 75th percentile of field measurements, segmented by device. Evaluate

accurate metadata remain necessary; avoid invented numerical ranking weights.

• Good CWV does not guarantee ranking. Relevant content, crawlability and

Verify authorship and credentials rather than generating them.

• Keep mobile content, metadata and structured data equivalent to desktop.

References

• Metadata API — references/metadata-api.md: read when writing generateMetadata, OG/icon files, ImageResponse, the manifest, or when streaming metadata / htmlLimitedBots is in play
• Sitemap & Robots — references/sitemap-robots.md: read for generateSitemaps, image/video sitemaps, multi-group robots rules, static robots.txt/sitemap.xml files
• JSON-LD Structured Data — references/json-ld.md: read before adding any schema type; has the supported/deprecated rich-result list and the @graph pattern
• AI Search (GEO/AEO) & AI Crawlers — references/ai-search.md: read when deciding robots rules for GPTBot/OAI-SearchBot/ClaudeBot etc., or when asked about llms.txt or AI Overviews
• SEO Audit Checklist — references/checklist.md: read when asked to audit a site end to end
• Troubleshooting — references/troubleshooting.md: read when a page is missing from Google, stuck in "Discovered/Crawled – currently not indexed", or indexes fine but never hydrates

Common Mistakes to Avoid

1. Mixing next-seo with Metadata API - Use only Metadata API in App Router
2. Missing canonical URLs - Set a self-referencing alternates.canonical when duplicate/parameterized URLs are a risk; it's a hint, not a requirement — Google may pick its own canonical
3. Using CSR for SEO pages - Use SSG/SSR for indexable content
4. Blocking /next/ in robots.txt - Crawlers need render-critical CSS/JS; never disallow /next/
5. Missing metadataBase - Not a build error: relative OG/Twitter image URLs fall back to the Vercel deployment URL or http://localhost:3000 with only a warning, and relative canonical/hreflang URLs stay relative (Google requires absolute hreflang URLs)
6. Viewport in metadata - themeColor/colorScheme/viewport in metadata are ignored; use export const viewport
7. Mixing metadata object and generateMetadata - Use one or the other in the same route segment
8. Duplicating icons in metadata + file conventions - Prefer favicon.ico/icon./opengraph-image. file conventions; they auto-emit tags and override the metadata object
9. Blanket-blocking AI crawlers - GPTBot disallow: / blocks training but leaves you in AI search; don't accidentally block search bots (OAI-SearchBot, Claude-SearchBot, PerplexityBot). See references/ai-search.md
10. Adding the keywords meta tag for Google - Google ignores it entirely (no indexing or ranking effect); it's noise, not a signal
11. Assuming named robots.txt groups inherit rules - Per RFC 9309 §2.2.1 the group applies only when no group matches, and Google never merges a specific group with *. A { userAgent: 'OAI-SearchBot', allow: '/' } group drops the wildcard's /api///admin/ disallows — repeat them in every named group
12. Trusting browser view for bot metadata - Check status, headers and the complete production response for Googlebot and a relevant HTML-limited bot. Streaming may place metadata outside the initial head. A spoofed User-Agent tests response behavior, not Google's actual crawl access or indexing; use URL Inspection and verified bot logs where available.

Quick Fixes

Add noindex to a page

export const metadata: Metadata = {
  robots: {
    index: false,
    follow: true,
  },
};

Keep the page crawlable for noindex to be seen. Robots disallow is not an index-removal or access-control mechanism. Preview protection and noindex are separate concerns; see references/sitemap-robots.md.

Dynamic metadata per page

type Props = { params: Promise<{ id: string }> };

export async function generateMetadata({ params }: Props): Promise<Metadata> {
  const { id } = await params;            // params is a Promise in current Next.js
  const product = await getProduct(id);
  return {
    title: product.name,
    description: product.description,
  };
}

Canonical for dynamic routes

type Props = { params: Promise<{ slug: string }> };

export async function generateMetadata({ params }: Props): Promise<Metadata> {
  const { slug } = await params;
  return {
    alternates: {
      canonical: `/products/${slug}`,
    },
  };
}