Technical SEO for a Next.js site, done once and correctly

Metadata comes from data, never from strings

The App Router's generateMetadata runs on the server with the same params the page uses, so the title, description and canonical can be derived from the record the page is about. A hand-written title in a page file is a second copy of a fact, and second copies drift.

One canonical per language

The Arabic page and the English page are not duplicates of each other — they are alternates. Each one canonicalises to itself. Pointing the Arabic page's canonical at the English URL, which happens constantly, tells a search engine that the Arabic version should not be indexed at all.

hreflang has to be reciprocal

• Every page lists every language version of itself, including its own.
• The declarations must agree in both directions — if A names B, B must name A.
• Use absolute URLs. A relative alternate is ignored.
• Include an x-default pointing at the version a visitor with no language match should get.
• Never generate alternates for a page that does not exist in that language.

The sitemap is generated, not maintained

Build it from the same arrays that produce the routes, so a new case study appears in the sitemap the moment it exists and an unpublished draft never does. Include lastModified from the record's own date rather than the build time — a sitemap where every URL changed today says nothing.

robots.txt, and what you are deciding

Disallowing a path does not remove it from an index — it prevents crawling, which can leave a URL indexed with no description. Use noindex for pages that should not appear, and robots.txt for paths that should not be fetched. They are different tools and using one for the other's job is the most common technical SEO mistake there is.