Claude SEO Audit Skill Guide (2026)

seo-optimizer

Hand the prioritized fixes straight to the optimizer skill to draft the corrected titles and meta.

playwright

Fetch and render the actual pages so the audit reads real DOM, not a guess.

Crawlability and indexation check (robots + sitemap)

Catch the silent killers: a Disallow that blocks Google, a missing sitemap, or pages that never made it into the index.

ForDevs who shipped a robots.txt change and watched traffic drop the next week. This is the first audit area for a reason.

The prompt

Use the seo-audit skill to audit only crawlability and indexation for https://example.com. Fetch /robots.txt and /sitemap.xml. Flag any Disallow rules that block indexable content, any crawl-delay, missing Sitemap directives, and sitemap URLs that 404 or redirect. List how many URLs the sitemap declares versus how many return 200. Output a findings table and a one-paragraph verdict on whether Google can crawl the site cleanly.

What slides.md looks like

## Crawlability & indexation

robots.txt: Sitemap directive PRESENT (/sitemap.xml)
Blocking rules: `Disallow: /app/` — OK, app is gated
Risk: `Disallow: /blog/drafts/` matches /blog/ prefix? No — exact

sitemap.xml: 142 URLs declared
  - 138 → 200 OK
  - 3 → 301 (update sitemap to final URL)
  - 1 → 404 (/old-pricing — remove)

Verdict: Crawlable. Fix the 4 stale sitemap entries.

One-line tweak

Append 'and check for orphan pages: URLs in the sitemap with zero internal links pointing to them' to catch pages Google can find but users can't.

Pairs with

firecrawl

Crawl the full site to compare discovered URLs against what the sitemap declares.

chrome-devtools

Confirm a page renders for Googlebot, not only returns 200, when JS hydration is involved.

On-page audit (titles, meta, heading hierarchy)

Find duplicate titles, missing meta descriptions, multiple H1s, and headings that skip levels, across a batch of pages.

ForContent teams who can't see the markup. On-page is the audit area you can fix fastest with the most ranking upside.

The prompt

Use the seo-audit skill to run an on-page audit across these 8 URLs: [paste list]. For each page extract the title tag (flag > 60 chars or duplicated across pages), meta description (flag missing or > 160 chars), H1 count (flag != 1), heading hierarchy (flag skipped levels like H2 to H4), and whether the target keyword appears in title and H1. Output one row per page in a findings table, then a summary of the three patterns that repeat most.

What slides.md looks like

## On-page findings
| URL | Title | Meta | H1 | Hierarchy | Keyword in title |
|-----|-------|------|----|-----------|--------------------|
| / | 78 chars (trim) | OK | 1 | OK | yes |
| /pricing | dup of /plans | missing | 0 | H2 first | no |
| /blog/x | OK | OK | 2 | OK | yes |

Top repeating patterns:
1. 6/8 pages missing meta description
2. 3/8 pages have 0 or 2+ H1s
3. 2 pages share an identical title

One-line tweak

Swap the URL list for 'crawl /blog and audit every post' when you want the whole content hub graded in one pass.

Pairs with

seo-optimizer

Rewrite the flagged titles and metas with intent-matched copy instead of fixing them by hand.

keyword-research

Confirm the target keyword each page should own before judging title relevance.

Core Web Vitals audit (LCP, INP, CLS)

Grade the three Core Web Vitals against Google's thresholds and trace each failure to the element causing it.

ForFront-end engineers who keep getting 'improve page experience' nudges in Search Console but no specifics.

The prompt

Use the seo-audit skill to audit Core Web Vitals for https://example.com/pricing. Check against the thresholds: LCP < 2.5s, INP < 200ms, CLS < 0.1. Use Chrome DevTools to capture a real trace. For each metric that fails, name the specific element or script responsible (the LCP element, the layout-shifting node, the long task blocking INP) and the fix. Output a findings table plus a 'fix in this order' list ranked by impact-to-effort.

What slides.md looks like

## Core Web Vitals — /pricing
| Metric | Value | Target | Verdict | Cause |
|--------|-------|--------|---------|-------|
| LCP | 4.1s | < 2.5s | FAIL | hero.jpg 1.2 MB, not preloaded |
| INP | 240ms | < 200ms | FAIL | pricing-toggle handler blocks 180ms |
| CLS | 0.04 | < 0.1 | PASS | — |

Fix order (impact / effort):
1. Preload + WebP the hero → LCP ~1.8s
2. Debounce the toggle handler → INP ~120ms

One-line tweak

Add 'compare lab data against the field CrUX data if available' to catch the gap between your fast laptop and real mobile users.

Pairs with

core-web-vitals

Once the audit names the failures, this skill implements the LCP/INP/CLS fixes.

lighthouse

Pull a full Lighthouse performance trace to back the metric numbers.

Schema and structured-data audit

Find pages missing JSON-LD, schema that fails validation, and rich-result types you're eligible for but not using.

ForTeams chasing rich results (FAQ, Product, Breadcrumb) who don't know which pages already have valid markup.

The prompt

Use the seo-audit skill to audit structured data on https://example.com. For the homepage, a product page, and a blog post, extract any JSON-LD, validate the required and recommended properties for each @type, and flag schema that would fail Google's Rich Results Test. List rich-result types the page is eligible for but does not implement (e.g. a blog post with no Article schema, a product page with no Product/AggregateRating). Output a table of @type, status, and missing properties. Note: implementation belongs to the schema-markup skill.

What slides.md looks like

## Structured data findings
| Page | @type found | Status | Missing / eligible |
|------|-------------|--------|---------------------|
| / | Organization | valid | + WebSite + SearchAction |
| /product/x | Product | invalid | offers.price missing |
| /blog/y | none | — | add Article + author |

Eligible but unused: FAQPage on /pricing (has Q&A content)
Handoff: pass these to the schema-markup skill to write the JSON-LD.

One-line tweak

Add 'and flag any schema that contradicts on-page content (e.g. AggregateRating with no visible reviews)' — that mismatch is a manual-action risk.

Pairs with

schema-markup

The audit only finds gaps; this sibling skill writes and validates the JSON-LD.

seo-optimizer

Tie schema gaps back into the broader on-page fix list.

Internal linking and site-architecture audit

Map the internal link graph, surface orphan pages, and find money pages buried too many clicks from the homepage.

ForSEOs who suspect link equity is pooling on the blog while product pages starve. Authority is the fifth audit area.

The prompt

Use the seo-audit skill to audit internal linking for https://example.com. Crawl up to 50 pages and build the internal link graph. Report: orphan pages (zero inbound internal links), pages more than 3 clicks from the homepage, the 5 pages with the most inbound links, and any generic anchor text ('click here', 'read more') pointing to important pages. Output a findings table and a short list of 5 specific link additions that would help the highest-value buried pages.

What slides.md looks like

## Internal linking findings
Orphan pages (0 inbound): /pricing/enterprise, /case-studies/acme
Depth > 3 clicks: /product/integrations (4), /security (5)
Top hubs: / (47 in), /blog (38), /docs (31)
Generic anchors: 12 "read more" → vary them

Recommended additions:
1. Link /pricing → /pricing/enterprise (depth 4 → 2)
2. Add /security to the footer (depth 5 → 1)

One-line tweak

Add 'weight the graph by which pages have the best Search Console impressions' to prioritize links toward pages already close to ranking.

Pairs with

firecrawl

Crawl the site to build the actual link graph instead of inferring it.

draw-io

Render the site architecture as a diagram for the audit report's appendix.

Content quality and E-E-A-T audit

Grade pages on depth, topical coverage, and the E-E-A-T signals Google's quality raters look for, beyond keyword density alone.

ForContent leads whose pages have the right keywords but still lose to thinner competitors with more authority signals.

The prompt

Use the seo-audit skill to run a content quality audit on these 5 blog posts: [paste URLs]. For each, assess depth versus the query intent, topical coverage gaps (subtopics a thorough page would cover but this one skips), and E-E-A-T signals: visible author with credentials, publish/update date, citations to primary sources, and first-hand experience markers. Flag thin content (< 300 meaningful words) and outdated stats. Output a per-post scorecard and a prioritized 'what to add' list.

What slides.md looks like

## Content quality — /blog/seo-basics
Depth: 620 words vs top result 2,100 — THIN
Coverage gaps: no section on Core Web Vitals, no examples
E-E-A-T: no author byline, no update date, 0 citations
Experience signals: none (reads like a definition list)

What to add (priority):
1. Author byline + credentials + 2026 update date
2. A worked example section (first-hand experience)
3. Cite Google's CWV thresholds (primary source)

One-line tweak

Add 'rewrite any sentence that reads like generic AI filler' to catch the spun-content smell that tanks E-E-A-T.

Pairs with

humanizer

Strip the AI-writing tells the audit flags so the page reads like a person wrote it.

keyword-research

Surface the subtopics a thorough page should cover to close the coverage gaps.

Competitor SEO diff (you versus a ranking rival)

Audit your page and a competitor's ranking page side by side, then list the gaps that explain why they outrank you.

ForTeams stuck at position 6-10 for a keyword a competitor owns at position 1-3. The diff is the most actionable artifact.

The prompt

Use the seo-audit skill to run a comparative audit: my page https://example.com/guide versus the rival ranking #1 for 'topic', https://rival.com/guide. Compare word count, heading structure, schema present, internal links pointing in, page speed, and topical subtopics covered. Output a two-column diff table (mine vs theirs) and a ranked list of the 5 gaps most likely to be closing the position difference.

What slides.md looks like

## Competitor diff — 'topic'
| Signal | Mine | Rival #1 |
|--------|------|----------|
| Word count | 900 | 2,400 |
| Schema | none | Article + FAQ |
| Inbound internal links | 2 | 19 |
| Subtopics covered | 4 | 11 |
| LCP | 3.6s | 1.9s |

Top 5 gaps to close:
1. 7 missing subtopics (theirs cover the full intent)
2. Add Article + FAQ schema
3. 17 more internal links needed

One-line tweak

Swap the single rival for 'the top 3 ranking pages' to find the subtopics ALL of them cover that you skip — those are non-negotiable.

Pairs with

competitor-analysis

Go deeper than a page diff into the rival's keyword and backlink strategy.

ahrefs

Pull the rival's real backlink and keyword data instead of inferring from the page.

Pre-launch / migration SEO checklist

Audit a staging site before launch (or a redesign before cutover) so you don't ship a traffic-killing regression.

ForEngineers shipping a redesign or domain migration. The cheapest audit is the one that runs before deploy, not after.

The prompt

Use the seo-audit skill to run a pre-launch SEO audit on the staging site https://staging.example.com against production https://example.com. Check: are noindex/nofollow tags accidentally left in place, are canonical tags pointing to staging URLs, do all production URLs have a 301 to their new equivalent, are titles and metas preserved, and is the sitemap updated. Output a go/no-go checklist where every red item blocks launch.

What slides.md looks like

## Pre-launch checklist — staging
[BLOCK] <meta name="robots" content="noindex"> still on 40 pages
[BLOCK] canonical tags point to staging.example.com
[BLOCK] /blog/* has no 301 map to new slugs
[OK] titles/metas preserved
[OK] sitemap.xml regenerated

Verdict: NO-GO. Clear the 3 blocking items first.

One-line tweak

Add 'and diff the rendered HTML of 5 key pages between staging and prod' to catch content that silently dropped in the rebuild.

Pairs with

playwright

Fetch both staging and prod to diff the actual rendered markup.

chrome-devtools

Verify the staging robots meta tag renders the way Googlebot would see it.

Recurring audit you re-run every release

Turn the one-off audit into a saved prompt you re-run after each deploy, so regressions surface in the PR, not in Search Console.

ForTeams that fixed everything once and want it to stay fixed. The audit is most valuable when it runs on a schedule.

The prompt

Use the seo-audit skill to run our standing SEO regression check on https://example.com. Re-audit only the items we marked critical last time: sitemap freshness, no stray noindex tags, all canonical tags self-referential, LCP < 2.5s on the homepage and /pricing, and zero duplicate titles across the top 20 pages. Compare against last week's seo-audit-report.md if present. Output only what CHANGED — new failures, newly fixed items — as a short delta, and exit clean if nothing regressed.

What slides.md looks like

## SEO regression delta vs 2026-06-16
NEW FAILURE: /pricing LCP 2.9s (was 1.8s) — hero image swapped, not WebP
NEW FAILURE: duplicate title on /features and /product
FIXED: sitemap now lists all 142 live URLs
NO CHANGE: canonical tags, noindex — all clean

Verdict: 2 regressions. Block the release or file 2 fixes.

One-line tweak

Wire this prompt into a CI step or a scheduled Claude run so the delta lands in the deploy channel automatically.

Pairs with

core-web-vitals

Auto-fix the LCP/INP regressions the delta surfaces before the release ships.

lighthouse

Capture a fresh performance trace each run so the LCP comparison uses real numbers.

Community signal

Three voices on running SEO audits with an AI agent. The first confirms the scope, the second makes the audit-then-fix case, the third is the honest limit — Claude covers most checks, not all, and only with the right input data.

“Run a complete technical SEO audit using Claude in 30 minutes. 12 checks covering crawlability, schema, Core Web Vitals, indexing, and AI visibility.”

Kulbhushan Pareek · Blog

Independent practitioner confirming Claude covers the exact audit surface this skill claims — crawlability, schema, Core Web Vitals, indexing. Concrete scope corroboration from outside the project.

“By bridging the gap between identifying a problem and executing code, marketers can now perform deep technical SEO audits and deploy fixes in minutes rather than months.”

Stormy AI blog · Blog

The pro-AI case for a code-agent audit: the agent both finds and fixes. Directly supports the audit → prioritized findings → implement loop the cookbook leans on.

“Claude can perform 9 of the 12 technical SEO checks in this list when given the right input data.”

Kulbhushan Pareek · Blog

An honest limit from a Claude-for-SEO advocate: it covers most but not all checks, and only with the right input data. The skill is a starting report, not a replacement for the tools that supply that data.

The contrarian take

Not every AI SEO audit is worth the tokens. The sharpest critique comes from Taha Bilal (Aristral):

“Most 'AI SEO audit' tools are classic site crawlers with a language model bolted on to write the recommendations in plainer English.”

Taha Bilal (Aristral) · Blog

From an SEO vendor's critique of AI audit tools.