Next.js SEO, minus the quiet mistakes.
The App Router gives you great SEO tools, and makes it easy to lose rankings with one line. Here is what goes wrong most often, and how Paradigma catches it on every pull request.
The App Router SEO toolkit. Five files and functions decide how Google and AI see you.
metadataandgenerateMetadata: title, description, canonical (alternates), robots and Open Graph for each page or layout.metadataBasein the root layout: the origin every relative URL in your metadata resolves against.app/robots.tsandapp/sitemap.ts: which crawlers may read what, and which URLs you want found.- Redirects:
redirects()innext.config, orpermanentRedirect()fromnext/navigation. - JSON-LD rendered as a script in a page or layout, and Server Components that put your content in the HTML.
app/layout.tsxBefore
export const metadata: Metadata = {
title: { default: "Acme", template: "%s | Acme" },
};After
export const metadata: Metadata = {
metadataBase: new URL("https://example.com"),
title: { default: "Acme", template: "%s | Acme" },
description: "What Acme does, in one sentence.",
};The one-line root layout change that keeps every canonical and Open Graph URL on your domain.
What breaks rankings. Common App Router mistakes, and the checks that catch them.
A noindex that ships by accident
One robots: { index: false } copied from a draft page, or set in a layout, hides every page under it from Google once it recrawls.
Caught by: A page turned noindex · Noindex served only to crawlers
A robots file that blocks too much
A staging app/robots.ts that disallows everything, or a blanket block on “AI bots” that also removes the crawlers ChatGPT, Claude and Perplexity use to cite you.
Caught by: Search engines blocked in robots.txt · AI search crawlers blocked · AI training crawlers blocked
Canonicals pointing at the wrong host
Without metadataBase, relative canonical and Open Graph URLs resolve against the deployment URL, so search engines can be told your real page lives on *.vercel.app.
Caught by: metadataBase changed or removed · Canonical pointing at the deployment URL · Canonical URL changed or removed
Routes deleted or renamed without a redirect
Old links and search results start returning 404. And redirect() is temporary (307): for a permanent move, use permanentRedirect() or a redirects() entry in next.config.
Caught by: Route removed without a redirect · A page now redirects
Missing or duplicate titles
A page that loses its own title inherits the layout's, and generateMetadata can return nothing when CMS data is missing. Either way, search results and AI answers show the wrong title.
Caught by: Page title or description removed · Duplicate page titles · Rendered pages without a title
Content that only exists in the browser
Fetching the main content in useEffect, or navigating with router.push instead of <Link>, leaves crawlers with a near-empty page and links they can't follow.
Caught by: Content lost on the Preview · Links that need a click to work
Sitemaps and structured data that drift
URLs fall out of app/sitemap.ts, JSON-LD disappears in a refactor, or a hand-built JSON-LD string stops parsing.
Caught by: URLs dropped from the sitemap · Structured data removed · Invalid structured data
Slower pages after a release
A hero image that loses its preload, or heavier client code, can push real-user Core Web Vitals from good to poor.
Caught by: Image loading priority removed · LCP regressions · INP regressions
How Paradigma checks it. Three layers, from the pull request to your live site.
- On every pull request: 27 code checks read your Next.js source and compare it with the base branch. Only blockers fail the GitHub check.
- On your Preview deployment: 13 checks read the HTML the Preview actually serves and compare it with production, which catches runtime metadata and CMS data.
- On your live site: 40 monitoring checks on Search Console, Bing, Core Web Vitals and crawler access: every day on paid plans, weekly on Free.
Fixes arrive ready: suggested changes on the pull request, draft fix pull requests from the dashboard, or Autopilot with your own coding agent. See all 80 rules.
Questions about Next.js SEO
- Does Paradigma work with the Pages Router?
- The code checks are built for the App Router, monorepos included. Pages Router apps and other sites still get rendered-page checks, monitoring, Search Console and AI visibility.
- Do I need Vercel?
- Not for the code checks. Preview checks run when GitHub reports a deployment for your pull request; Vercel Preview is the setup we have tested.
- Will it block my merges?
- Only blockers fail the check, such as a page turning noindex or the whole site blocked in robots.txt. Everything else is a warning you can act on or mute.
- Does it run my code?
- No. Paradigma reads your source statically, in memory, and never stores repository files.