optimise-seo

• IS: crawlability, metadata, structured data, canonicals, redirects, hreflang, Core Web Vitals, programmatic SEO, security/privacy headers, and error-page status behaviour for Next.js App Router apps.
• IS NOT: visual redesigns or layout changes (use ui-design), or page-level UI quality review (use ui-audit). Making a site readable by AI agents (llms.txt, MCP discovery, WebMCP) is out of scope; research current specs rather than relying on static SEO guidance.

Allowed surface: metadata, structured data, semantic HTML, internal links, alt text, app/sitemap.ts, app/robots.ts, next.config.ts redirects and headers, error pages, performance tuning. Never touch component styling or layout.

Workflow

Copy and track this checklist:

SEO progress:
- [ ] Step 1: Inventory routes and decide index intent per route
- [ ] Step 2: Fix crawl/index foundations (sitemap, robots, canonicals, redirects, status codes)
- [ ] Step 3: Implement metadata + structured data
- [ ] Step 4: Improve semantics, internal links, and Core Web Vitals
- [ ] Step 5: Validate with references/seo-checklist.md and report evidence

Before writing code for steps 2-4, read references/nextjs-implementation.md: App Router patterns (Metadata API, generateMetadata, sitemap index, JSON-LD component, OG image generation, headers()/redirects() config).

Must-have on every site

• app/sitemap.ts lists all public URLs; app/robots.ts links to it
• Canonical URL set and consistent on every page (one host, one casing, one trailing-slash policy)
• Unique title and description per page via metadata or generateMetadata
• OpenGraph + Twitter Card tags with a 1200x630 image
• JSON-LD: Organization and WebSite on the homepage, BreadcrumbList on inner pages, plus Article/Product/FAQ where the content type matches. Define each entity once in a @graph with a stable @id and reference it by @id instead of duplicating it inline (@id conventions and the per-type code: references/nextjs-implementation.md)
• One h1 per page with logical h2-h6 hierarchy
• Descriptive alt text, internal links between related pages, CWV in target (LCP < 2.5s, INP < 200ms, CLS < 0.1)

Programmatic SEO (pages at scale)

• Validate search demand for the pattern before generating pages
• Each page needs unique value backed by defensible data; templated text swaps are doorway pages
• Clean subfolder URLs, hub-and-spoke linking, breadcrumbs on every page
• Index only strong pages; noindex the long tail and monitor indexation and cannibalisation in Search Console
• Applies to single pages too: noindex a thin page and omit it from the sitemap until it carries unique content, then add an internal link and index it

SEO audit (triage order)

1. Crawl/index: robots, sitemap, stray noindex, canonicals, redirect chains, soft 404s
2. Technical: HTTPS, Core Web Vitals, mobile/desktop parity
3. On-page: titles/H1 uniqueness, internal links, remove or noindex thin pages
4. Orphans: indexable pages (metadata, canonical, breadcrumb) with no internal link or sitemap entry. Either give them a crawl path (nav/footer link plus sitemap) or noindex and drop from the sitemap

Redirects and indexing policy

• Permanent moves use 301/308, temporary use 302/307. Never chain; point straight to the final URL.
• A missing page must return a real 404, not 200 with a friendly message: 200-for-missing is a soft 404 and won’t index.
• Give every route an explicit index decision: public pages default to index, follow; staging, admin, thin, or private routes get metadata.robots noindex (HTML) or X-Robots-Tag (non-HTML, whole environments).

Internationalisation (multi-locale sites)

One URL pattern for all locales; reciprocal hreflang with self-reference plus x-default; translate metadata, not just body; never auto-redirect by IP or Accept-Language. Full rules and the generateMetadata pattern: references/internationalisation.md.

Technical hardening (security, privacy, resilience)

Security headers (HSTS, CSP, nosniff, frame-ancestors, Referrer-Policy, Permissions-Policy), SRI, cookie flags, security.txt, privacy policy and opt-in consent, correct 404/500/503 behaviour, web app manifest. Read references/technical-hardening.md for header, cookie, consent, or error-page tasks.

Gotchas

• Don’t over-generate thin or doorway pages: indexation drops and sitewide quality signals suffer.
• Don’t let canonicals conflict across variants (trailing slash, www, uppercase): ranking signal splits between duplicates.
• Don’t block crawlers via robots.txt, noindex, or auth walls on routes meant to rank; check before shipping, not after traffic drops.
• Don’t move indexable content behind a client-only render ("use client" plus a fetch in useEffect): the crawler indexes the empty shell, so the page ranks for nothing. Server Components already render on the server, so this is a regression you introduce, not a default to fix.
• Don’t change URLs without 301/308 redirects: link equity and crawl budget are lost.
• Don’t add JSON-LD that doesn’t match visible content; Google treats it as spam and may demote the page.
• Don’t ship hreflang that isn’t reciprocal across every alternate; search engines ignore non-mutual sets.
• Don’t serve a maintenance window with 200 or 404: return 503 + Retry-After so the site isn’t deindexed.
• Don’t add Strict-Transport-Security with preload/includeSubDomains before every subdomain is HTTPS; it’s effectively irreversible.
• Don’t set Person as publisher on articles: rich results expect an Organization publisher with a logo. Keep Person as author.
• Don’t stop at required schema fields: Search Console flags missing recommended fields as rich-result warnings. Clear warnings, not just errors (per-type field lists: references/nextjs-implementation.md).
• Don’t hardcode the sitemap lastModified: it goes stale and signals dead content. Derive it from the most recent content date.
• Don’t leave indexable pages orphaned: no internal link wastes crawl equity and the page may never be discovered or ranked.

References

• references/nextjs-implementation.md: App Router code patterns; read before steps 2-4
• references/internationalisation.md: URL strategy, hreflang, localised metadata; read for multi-locale sites
• references/technical-hardening.md: security headers, privacy/consent, resilience; read for hardening tasks
• references/seo-checklist.md: pass/fail validation checklist; copy during step 5

Related skills

• ui-design: visual direction, palettes, typography, landing-page CRO
• ui-audit: page-level UI quality and rendered i18n behavior (locale formatting, plurals, RTL); optimise-seo owns hreflang and localized metadata
• multi-tenant-architecture: per-tenant routing and custom domains. It owns the constraint that robots.txt, sitemap.xml, and llms.txt vary by tenant and never come from /public; this skill owns their content once routing works.

Validation (step 5, evidence required)

Copy references/seo-checklist.md, mark every item pass/fail, then attach command evidence:

Report remaining blockers with exact URLs and owner/action.