SEO kit

Every page's <title>, description, canonical and Open Graph come from one of two builders: content pages get the article-shaped one, every other route the marketing one.

Metadata

export async function generateMetadata({ params }: PageProps): Promise<Metadata> {
  const { locale } = await params;
  const t = await getTranslations({ locale, namespace: "contact.meta" });
  return buildPageMetadata({
    locale,
    pathname: "/contact",
    title: t("title"),
    description: t("description", { product: config.siteName })
  });
}

Canonical and hreflang

Article archives

An article section has exactly one crawlable pagination surface - the ?page=N ladder on its own hub - so its documents never occupy two indexable addresses.

• /{section}?page=1 permanently redirects to /{section}; a non-numeric or out-of-range page 404s from the body and the metadata.
• Rungs 2+ carry a de-optimised title and no description, so they never compete with the hub for the section's own query.
• Both ladder controls are plain navigation: Load more articles is the next numbered link with a friendlier label, so every rung is a real server-rendered document.
• Every other listing pages in place, and sorting never changes a URL anywhere, so only the ladder mints a second address - see Content sections.

Breadcrumbs

buildBreadcrumbList(items, locale, path) and <Breadcrumbs items> render one trail from buildBreadcrumbItems twice: as visible navigation and as a BreadcrumbList node. Content pages, contact, the legal pages and /download use both, so no page emits breadcrumb schema without showing the trail.

Structured data on content pages

<ContentSchema doc locale section> emits one connected graph per page, picked by its template.

Every graph also carries BreadcrumbList, the author Person, and FAQPage when the document has an faq. Review, rating and offer nodes are never built from content data.

OG images

Two route handlers generate 1200x630 cards on demand: /og/{locale}/{section} for a listing, /og/{locale}/{section}/{slug} for a document, each drawing the title, section label, site name and a bundled font.

• A document with image advertises that hero and never names the generated URL.
• Every content route declares its own openGraph.images, archives included; nothing is inherited from a parent segment.

Markdown renditions

Every content page also serves Markdown at <path>.md, and at its canonical URL under Accept: text/markdown - see AI Agents & Crawlers. /llms.txt lists the static routes, then one line per section listing with content, titled from content.sections.{slug}, minus noindex sections.

The alternatives rendition puts the frontmatter first and the body after (the verdict as a list, the grouped matrix and the pricing as tables, one block per rival, FAQ as question and answer): for that template the body alone is close to empty. An alternatives section serves its page at /{section}.md, the way a page section serves its root, rather than a listing of documents it does not have. Every side's pricing is printed - ours from pricingConfig.plans, theirs from the compare fact sheet each item names - above a starting-price line naming each tool.

Preferred sources (Google)

A reader who marks a site as a preferred source sees it more often in Google Top Stories, with a "preferred" badge. Ships ON.

It renders as a compact button in the footer brand column, between the support email and the social icons (site-wide), and once in an article's sticky right rail, directly under the share controls (desktop widths, where that rail exists). Nowhere else: a section hub asks a reader who has read nothing yet, and commercial and app surfaces are not Top Stories candidates.

Google's translated badge images are a download from Google, not a file in this repo: add one to your own pages if you want it.

The required-fields guard

A document missing a required field gets indexable: false: noindex, follow, out of the sitemap and llms.txt, hidden from hubs and listings in production, rendered degraded, and named in the build summary. A whole section goes out with noindex: true, which changes robots, sitemap and llms.txt only, never listing membership. See the editorial checklist.