claude skills / Claude Code Web Skills / technical-seo-setup / en français

Technical SEO setup.

The SEO plumbing set up once in the code, so every new page inherits it.

See what it produces
  • Claude Code
  • Next.js
  • Google Search Console
  • Bing Webmaster Tools
  • Lighthouse
  • Rich Results Test
9steps, from the initial state to submission in Search Console
2.5 starget LCP, Google's "good" threshold for Core Web Vitals
2measurements for each speed fix: before and after

orders of magnitude, depending on your volume

the problem

What gets in the way today.

  • 01A staging robots.txt has been blocking the whole site for weeks and nobody noticed.
  • 02The main content is rendered in JavaScript, so neither Google nor AI assistants really read it.
  • 03The same titles repeat across ten pages and the sitemap dates change with every deploy.

input, output

You give this. You get this.

What you give Claude: the code, the domain, the page types and the facts about the business

Company: Lumen Ledger, bookkeeping software
Codebase: Next.js App Router 15, static and server-rendered, pricing calculator client-side
Domain: www.example.com, English only
Page types: home, 4 product pages, pricing, 18 guides, legal (31 pages)
No index: /signup/thanks, /search, /legal/cookies-settings
AI crawlers: we want to be recommended by assistants
Access: Search Console and Bing Webmaster Tools

$ claude "Set up the technical SEO of this Next.js site: metadata, robots, sitemap, structured data, llms.txt and Core Web Vitals."

Technical SEO setup, report excerptexample

Technical SEO setup: www.example.com (Lumen Ledger)

Framework: Next.js App Router 15 | Date: 2026-11-19

BEFORE > AFTER

Titles: 9 of 31 pages share one title > 31 unique titles, 50 to 60 characters

Meta descriptions: 14 missing > 31 unique, trimmed at sentence end

Canonicals: none > self-referencing on all indexable pages

robots.txt: "Disallow: /" copied from staging > allow, private paths disallowed, AI policy written

Sitemap: missing > index with pages and articles, real lastmod from git

Pricing table: client-side only > server-rendered, calculator stays client-side

404 status: returned 200 > returns 404

llms.txt: missing > generated from content, 12 main pages

The robots.txt finding was urgent: production had blocked all crawlers since a deploy on 2026-10-02. Approved and fixed first.

CRAWL POLICY

Owner's decision: open. All listed AI agents allowed. The CDN bot protection challenged PerplexityBot, an exception was added and tested: 200 response.

STRUCTURED DATA

All pages: Organization, WebSite, WebPage, BreadcrumbList > no errors

Product pages: SoftwareApplication with offers as shown on pricing > eligible

Guides: Article with author page > no errors

PERFORMANCE (mobile, lab)

Home: LCP 4.6 s > 2.1 s | CLS 0.18 > 0.02 | score 58 > 93

Pricing: LCP 3.9 s > 2.3 s | TBT 610 ms > 160 ms | score 49 > 88

Main fixes: hero image 1.2 MB > 140 KB, chat widget loaded on interaction, fonts self-hosted.

NEEDS INPUT

Official social profile URLs for sameAs.

NEXT CHECK

2026-12-10: indexing report, sitemap status, field Core Web Vitals.

the method

What Claude does, in order.

  1. step 1Record the starting point

    For each page type: status, title, meta, canonical, robots directive, number of h1s, structured data, content present without JavaScript, and mobile Lighthouse.

  2. step 2Set metadata and crawling

    One helper for title, meta, canonical and Open Graph. A robots.txt that allows by default, an AI policy written with you, staging never indexable.

  3. step 3Sitemap and structured data

    A sitemap built from the same data as the pages, with real dates. One JSON-LD graph per page type, matching the visible content and validated.

  4. step 4Speed up, fix, submit

    Main image preloaded, unused scripts removed, self-hosted fonts, a real 404 page, single-hop redirects, then the sitemap submitted to Search Console and Bing.

demo in French / click, try it

Your turn.

An interactive demo of the skill. It is in French, the pack works in any language: ask Claude to answer in English.

the film / 44 seconds, with sound

The pack film.

the pack film

In 8 seconds.

the guardrails

What the skill never does.

  • Never block indexing by accident: any change that removes pages from search requires your explicit approval.
  • Structured data describes what is visible: no fake reviews, no invented ratings, no made-up dates.
  • Real dates: lastmod and dateModified come from the actual content change, never from the build date.
  • No magic tag for AI, and no promise of rankings, traffic or citations.

Questions

Does the skill work outside Next.js?

Yes. The code templates are written for Next.js App Router, with notes for other stacks. The rules (canonicals, sitemap, JSON-LD, 404) stay the same.

Does llms.txt get you into ChatGPT?

No. It is a cheap courtesy for agents, not a ranking lever. What matters: clean HTML, clear answers, crawl access and consistent facts.

Should I block AI crawlers?

It is a business decision Claude makes with you: open, partial or closed. It is written explicitly in robots.txt, then tested with each user agent.

When will I see results?

Re-read the indexing report two to four weeks after submission. Field Core Web Vitals need about 28 days of data.

the pack

In the pack: Claude Code Web Skills.