Claude Code SEO Checklist

A free technical SEO checklist in dependency order: crawlability, indexing, canonicals, sitemaps, metadata, structured data, internal links, Core Web Vitals and search engine integration.

Plain Markdown, no sign-up. Drop it straight into a repository.

Download the checklist

A technical SEO checklist you can hand to Claude Code and work through in order. Every item is something a machine can verify, which is the point: this list deliberately stops where SEO becomes guesswork about ranking factors nobody outside Google can measure.

The order matters. Crawlability gates indexing, indexing gates everything else. Fixing metadata on a page Google cannot reach is wasted effort.

A note on scope. An LLM is excellent at the verifiable half of SEO — status codes, canonicals, markup, link graphs, heading structure — and unreliable at the half that requires external data, such as search volume, keyword difficulty, or competitor backlinks. If a tool tells you a keyword gets a specific number of monthly searches without querying a data source, that number was invented. Get those figures from Search Console, Bing Webmaster Tools, or a keyword tool with an API.

1. Crawlability

robots.txt returns 200 and is valid
Nothing needed to render the page is disallowed — CSS, JS, fonts, images
No blanket Disallow: / left over from staging
Sitemap URL declared in robots.txt
Crawl-blocking is intentional everywhere it exists, and you can say why for each rule
Server does not rate-limit or block legitimate crawler user agents
No infinite crawl spaces: calendars, faceted filters, session IDs in URLs

2. Indexability

Rendered HTML contains no <meta name="robots" content="noindex"> on pages meant to rank
No X-Robots-Tag: noindex response header on those pages
Check the rendered page, not the template — noindex is often injected
Staging, preview and password-protected environments *are* noindexed
Thin, duplicate or utility pages are deliberately excluded, not accidentally included
Search Console shows the important pages as "Indexed", not "Discovered — currently not indexed"

3. Canonicals

Every indexable page has exactly one <link rel="canonical">
It is absolute, HTTPS, and uses the production hostname
It is self-referencing unless a duplicate genuinely exists
Parameterised URLs (?utm_, ?sort=, ?variant=) canonicalise to the clean URL
Paginated pages self-canonicalise; they do not all point at page one
Canonical target returns 200, not a redirect or a 404
No conflicting signals: canonical says one thing, sitemap or internal links say another

4. URLs and redirects

One canonical hostname; the other variants redirect to it in a single hop
HTTP redirects to HTTPS sitewide
No redirect chains longer than one hop, no loops
Permanent moves use 301/308, temporary use 302/307 — and the choice is deliberate
Old URLs from any previous site redirect to the closest equivalent, not all to the homepage
URLs are lowercase, hyphenated, readable, and stable
Trailing-slash behaviour is consistent and enforced by redirect

5. XML sitemap

Reachable, returns 200, valid XML
Contains only canonical, indexable, 200-status URLs
Contains no redirects, 404s, noindexed pages, or non-canonical variants
Under 50,000 URLs and 50MB uncompressed per file; index file used above that
<lastmod> reflects real modification dates, or is omitted
<priority> and <changefreq> are ignored by Google — do not agonise over them
Submitted in Search Console and Bing Webmaster Tools, showing "Success"
On a hosted platform, validate the generated sitemap rather than replacing it

6. Titles and meta descriptions

Every indexable page has a <title>; no page shares one with another
Titles lead with the query intent, not the brand
Roughly 50–60 characters so they are not truncated
Every page has a meta description; none are duplicated
Descriptions are roughly 140–155 characters and read as a reason to click
No keyword stuffing, no pipe-separated keyword lists
No unresolved template variables or HTML entities in either
The homepage title is not the raw domain or the platform default

7. Headings and on-page structure

Exactly one <h1> per page, matching the page's actual subject
Heading levels descend without skipping (no h2 straight to h4)
Headings describe the section, and would work as a table of contents
The primary topic appears naturally in the first paragraph
Content answers the query it targets in the first screenful
No hidden text, no text-coloured-as-background, no keyword blocks

8. Structured data

Valid JSON-LD in the rendered HTML
One Organization (or Person) entity, consistent across the site
WebSite entity present where appropriate
Article/BlogPosting on articles, with headline, datePublished, dateModified, author, publisher
Author is a real, identifiable entity — a Person where a person wrote it
Product markup on product pages matching visible price, currency, availability
BreadcrumbList matching the visible breadcrumb hierarchy
No duplicate entities of the same type on one page
No Review/AggregateRating markup without real reviews. This is a manual-action risk
Validated against Google's Rich Results Test on the live URL

9. Internal linking

Every indexable page is reachable from the homepage in a few clicks
No orphan pages
Anchor text is descriptive; no "click here", "read more" as the only link text
Anchor text varies naturally; no sitewide exact-match keyword links
Pillar pages link down to their supporting pages, and supporting pages link back up
Related pages within a topic cluster link to each other
Navigation and footer links are HTML anchors, not JavaScript handlers
No broken internal links and no internal links to redirects

10. Content quality

Each page targets one search intent, and no two pages target the same one
Cannibalisation checked: search site:yourdomain.com <query> and confirm one clear winner
Content demonstrates first-hand experience where the topic calls for it
Claims are supported; statistics have a source; commands and APIs are real
Author is identified, with credentials that are true
Publish and update dates are real
No scaled or templated pages produced primarily for search engines
Every page would still be worth publishing if search engines did not exist

11. Images

Descriptive filenames
Meaningful, non-stuffed alt on informative images; empty alt on decorative ones
Served at rendered size, compressed, modern format with fallback
width and height set
Lazy-loaded below the fold, eager for the LCP image
Images appear in the sitemap or are otherwise discoverable, if image search matters to you

12. Performance and Core Web Vitals

LCP ≤ 2.5s, CLS ≤ 0.1, INP ≤ 200ms, measured on the live site
Measured on mobile as well as desktop
Field data checked in Search Console once enough traffic exists — lab data is a proxy
Render-blocking resources minimised
Layout shift eliminated at the source (dimensions, font fallbacks, reserved ad slots)
Third-party scripts audited; each one still earns its cost

13. Mobile

Mobile rendering is the one that matters — indexing is mobile-first
Content parity: mobile shows the same content and links as desktop
No horizontal scroll at 320px
Tap targets comfortable and not overlapping
Interstitials do not obscure content on arrival from search

14. International (only if you serve multiple locales)

hreflang annotations are reciprocal and use valid language-region codes
An x-default exists
Each locale's URLs are distinct and canonical to themselves
Skip this section entirely if you serve one language — a broken hreflang is worse than none

15. Search engine integration

Google Search Console verified for the exact production hostname
Sitemap submitted; status "Success"; discovered URL count looks right
Coverage report reviewed for unexpected exclusions
Live URL inspection run on the homepage and one key page
Bing Webmaster Tools verified, sitemap submitted
IndexNow key file reachable at the site root and returning the key as plain text
IndexNow submissions return 200 or 202 — note that Google does not use IndexNow
URL submission reserved for genuinely new or changed pages, not run on a timer

16. Ongoing validation

A crawl script you can re-run after every deploy
It fails loudly on a partial run rather than reporting a clean result
Duplicate title and description detection is part of it
Structured data re-validated after template changes
Search Console checked monthly for new coverage errors and manual actions
Query and page performance reviewed against what you actually publish

What this list does not do

It does not tell you what to write about, and it will not rank a page that has nothing to say. Technical SEO removes obstacles; it does not create demand. Every item here is a way of making sure a genuinely useful page is not held back by something mechanical.

The Claude Code Website Launch System turns this checklist into prompts and audit scripts that produce evidence rather than opinions.

Licence

Free to use in any project, including client work. Do not resell or republish it as your own.