cmux/web/i18n
Lawrence Chen 46589f531c
Fix SEO indexing: hreflang, canonicals, sitemap, trailing slash (#2193)
* Fix SEO indexing: add hreflang, canonicals, sitemap per-locale entries

Google Search Console showed 380 not-indexed vs 86 indexed pages.
Root causes: missing hreflang tags on rendered pages (only in sitemap),
no canonical on homepage, inconsistent canonicals wiping parent hreflang,
sitemap only listing English URLs, trailing slash duplicates, and
_next/static chunks being crawled as pages.

Changes:
- Add buildAlternates() utility for consistent canonical + hreflang
- Add hreflang tags to all pages via alternates.languages in metadata
- Add self-referencing canonical URLs to every page (homepage had none)
- Expand sitemap to emit separate entries for each locale
- Add missing /docs/custom-commands to sitemap
- Remove skipTrailingSlashRedirect to normalize trailing slashes
- Block /_next/ in robots.txt to stop chunk crawling

* Add per-page alternates to docs sub-pages and blog index

Docs sub-pages and blog index only returned title/description in
generateMetadata, so they inherited the parent layout's alternates
(pointing to /docs or /blog). Now each page sets its own
buildAlternates() with the correct path so canonical and hreflang
point to the actual page URL.

* Derive openGraph.url from buildAlternates to avoid drift

* Redirect non-English legal pages to English, remove from sitemap

Legal pages (privacy policy, TOS, EULA) are untranslated English content.
Serving them under every locale creates 54 duplicate URLs. Now:
- Middleware 301-redirects /ja/privacy-policy etc. to /privacy-policy
- Sitemap only includes English URLs for legal pages (no locale variants)
- Legal page metadata uses static English-only canonical

* Fix legal page redirect to only match /<locale>/<page> paths

endsWith matched too broadly (e.g. /docs/eula). Now only redirects
when the path after the first segment is an exact legal page match.

* Skip next-intl for legal pages to prevent locale redirect loop

Without this, a Japanese user hitting /privacy-policy could be
redirected by next-intl to /ja/privacy-policy, which our middleware
redirects back to /privacy-policy, creating a loop.

* Rewrite legal pages to /en/ instead of NextResponse.next()

Pages live under app/[locale]/, so skipping next-intl entirely
would break route resolution. Rewrite to /en/privacy-policy etc.
so Next.js can resolve the [locale] segment correctly.

---------

Co-authored-by: Lawrence Chen <lawrencecchen@users.noreply.github.com>
2026-03-26 15:10:39 -07:00
..
navigation.ts Internationalize website with next-intl for 19 languages (#1216) 2026-03-12 05:36:58 -07:00
request.ts Fix locale page crashes under Google Translate (#1956) 2026-03-22 16:55:00 -07:00
routing.ts Internationalize website with next-intl for 19 languages (#1216) 2026-03-12 05:36:58 -07:00
seo.ts Fix SEO indexing: hreflang, canonicals, sitemap, trailing slash (#2193) 2026-03-26 15:10:39 -07:00