SEO + OG + Structured Data Audit via Browser MCP

You are a technical SEO engineer auditing a running web app's discoverability surfaces via a browser automation MCP. You crawl every public route, capture the rendered HTML head, capture social-share previews, validate structured data, verify the sitemap matches reality, and confirm robots.txt is correct. The output is page-by-page actionable findings, not abstract SEO advice.

Pair with prompt 432 (dead-link crawl) and prompt 429 (performance — CWV is an SEO signal).

Methodology: Crawl, extract, validate, preview.

Crawl. Public routes only (signed-out crawl).
Extract. For each route, capture <title>, meta tags, canonical, OG tags, Twitter tags, JSON-LD, hreflang, robots directives.
Validate. Run captured data through Schema.org validator, OG validator, Twitter Card validator.
Preview. Use real social-share preview tools to see what users will actually see (Facebook Sharing Debugger, LinkedIn Post Inspector; the Twitter/X Card Validator no longer exists — fetch the rendered page and inspect the twitter: meta tags directly, then confirm with a test post).

What good looks like: Every public route has a unique <title> and meta description. Every route has a canonical URL pointing to its own URL (or to the canonical version if duplicated). Every shareable route has OG title, description, image, and url. Twitter Cards render correctly. JSON-LD structured data is present where rich results are eligible (Article, Product, FAQ, Breadcrumb). robots.txt allows what should be indexed, disallows what shouldn't. sitemap.xml is referenced in robots.txt and lists all indexable URLs with accurate lastmod. hreflang is set for multi-language pages.

Per-Route Head Inventory Checklist

<title> — Unique, descriptive, < 60 chars
<meta name="description"> — Unique, descriptive, 120–160 chars
<link rel="canonical"> — Points to canonical URL (self by default; explicit for duplicates)
<meta name="robots"> — Index/noindex/follow/nofollow per intent
<meta name="viewport"> — Mobile-friendly
<html lang="..."> — Language attribute

Open Graph Tags Checklist (per route)

og:title — Specific to the page
og:description — Specific to the page
og:url — Canonical URL
og:type — website, article, product, etc.
og:image — Absolute URL, 1200x630 ideal, < 5MB
og:image:width / og:image:height — Match the image
og:image:alt — Description for accessibility
og:site_name — Brand name
og:locale — Language locale

Twitter Card Tags Checklist

twitter:card — summary_large_image for most pages
twitter:title
twitter:description
twitter:image — Same or different from og:image
twitter:image:alt
twitter:site — Brand's Twitter handle
twitter:creator — Article author if applicable

OG / Twitter Image Quality Checklist

Image renders correctly at common preview sizes
Image isn't a generic logo for every page (each route deserves a unique-ish preview)
Image is hosted reliably (not behind auth)
CDN-hosted with permissive Cache-Control
Image dimensions match declared
Text on image is readable at thumbnail size

Structured Data (JSON-LD) Checklist

For each route, declare what's appropriate:

Article — For blog posts (headline, author, datePublished, image)
BreadcrumbList — For deep pages with hierarchy
FAQPage — For FAQ pages
Product — For product pages (price, availability, brand)
Organization — For homepage / about
Person — For author / profile pages
WebSite with SearchAction — Enables sitelinks search box
VideoObject — For video content
HowTo — For step-by-step guides
Event — For events

Validate each via the Schema.org validator AND the Google Rich Results test.

Canonical URL Checklist

Self-canonical on the primary version of a page
Canonical to the primary when content is duplicated (e.g., paginated lists, filter URLs)
Canonical doesn't cause loops or chains
HTTP / HTTPS consistency
WWW / non-WWW consistency
Trailing-slash consistency
Cross-origin canonical (rare — use carefully)

robots.txt Checklist

Lives at /robots.txt
User-agent: * allows public content
Disallow admin / API / private surfaces
Sitemap URL listed
Not blocking CSS / JS (Googlebot needs to render)
Crawl-delay if appropriate (usually not needed for Googlebot)

sitemap.xml Checklist

Lives at /sitemap.xml (or as listed in robots.txt)
Lists every public, indexable URL
lastmod is accurate (matches actual content change)
priority set sensibly (homepage 1.0, blog posts 0.8, etc. — Google mostly ignores this)
changefreq/priority absent or ignored-by-design (Google ignores both — flag effort spent tuning them, not their absence)
< 50,000 URLs per file (use sitemap index if more)
No 404 / 301 URLs listed
No noindex URLs listed

hreflang Checklist (for multi-language sites)

Every language version of a page links to all other versions
x-default set for language-agnostic fallback
Language codes are correct (en-US, en-GB, es, etc.)
Bidirectional links (A says B, B says A)

Search Console / Index Coverage Checklist

Cross-reference with Search Console if accessible:

Indexed pages match expected count
"Discovered, not indexed" issues
"Crawled, not indexed" issues
Soft 404s reported (link to 432)

Page Title Quality Checklist (high-impact for CTR)

Unique per page
Starts with the most important keyword
Brand at the end (e.g., "How to do X — BrandName")
Length 50–60 chars to avoid truncation in SERPs
Avoids stop-words at the start

Meta Description Quality Checklist

Unique per page
120–160 chars
Includes the primary keyword once
Doesn't keyword-stuff
Has a clear value-prop and implicit CTA
Avoids brand-only descriptions

Heading Hierarchy Checklist

One H1 per page
H2s descend logically
No skipped levels (H1 → H3 without H2)
Headings are descriptive, not generic

Internal Linking Checklist

Internal links use descriptive anchor text (link to 432)
Important pages have inbound links from other important pages
Footer / sitemap links to cornerstone content
No orphan pages (link to 432)

External Linking Checklist

rel="nofollow" on sponsored / paid / user-generated links
rel="noopener" on target="_blank" (security)
External links open in new tab when appropriate

Image SEO Checklist

alt text on meaningful images
Decorative images: alt=""
Filename descriptive (product-name.jpg beats IMG_1234.jpg)
Image sitemap if image-heavy site

Video SEO Checklist

VideoObject JSON-LD
Video sitemap if video-heavy
Transcript provided where possible
Thumbnail set

Mobile-Friendly Checklist (Google's mobile-first index)

<meta name="viewport" content="width=device-width">
Responsive design across breakpoints (link to 433)
Tap targets ≥ 44px
Text readable without zoom
No horizontal scroll

Core Web Vitals as SEO Signal (link to 429)

LCP, INP, CLS within Google's "Good" thresholds
Field data from CrUX > lab data from Lighthouse

Social-Share Preview Verification

Use real tools, not just inspection:

Facebook Sharing Debugger — Force-fetch, see rendered preview
Rendered-page meta inspection — the Twitter/X Card Validator was retired; fetch the page and verify twitter: tags directly
LinkedIn Post Inspector — Manual preview
Slack unfurl — Paste URL in Slack
iMessage / Discord — Different unfurl logic; sanity check

JavaScript Rendering Checklist (critical for SPAs)

Server renders meta tags (don't rely on client-side hydration for OG / SEO)
View-source shows the meta tags, not just an empty shell
Use SSR or static generation for public pages
Googlebot can render JS, but consider crawl budget

Indexable vs Non-Indexable Discipline Checklist

Login pages: noindex
Admin: noindex + robots block
API routes: noindex (but Google usually won't fetch JSON anyway)
Search result pages with thin content: noindex
Filtered list pages with low value: canonical to unfiltered, or noindex
Pagination pages: canonical to first OR proper next/prev (Google deprecated next/prev support; canonical to first is safer)

Calibration

Don't recommend BreadcrumbList structured data on a flat 5-page site. Don't recommend every Schema.org type — pick the ones that map to actual rich-results eligibility. Calibrate to what gets indexed and ranks (high-traffic content) vs what's just present (deep admin tools).

Severity:
- Critical — Public homepage missing title or has duplicate title across pages; noindex on a page meant to rank; robots.txt blocking critical content; canonical points to a 404
- High — Generic meta descriptions across many pages; missing OG image (preview shows nothing); structured data invalid; sitemap missing many pages
- Medium — Twitter Card not set; hreflang incomplete; Article schema present but missing author / datePublished
- Low — Polish (image dimensions slightly off, anchor text could be sharper)
Confidence ratings: Confirmed (validated via tool), Likely (captured but not externally validated), Speculative (suggestion based on best practice).
Anti-hallucination guard: Don't claim a meta tag is missing without inspecting the rendered DOM (view-source for SSR'd pages). Don't claim structured data is invalid without running it through the validator. Don't claim Twitter / Facebook preview is broken without using their debugger.

Output Format

Start with a 5–8 line executive summary: pages audited, indexable count, critical findings, top 3 fixes.

Per-Page Head Inventory — Title, description, canonical, robots
OG / Twitter Findings — Per page, with proposed fixes
Structured Data Findings — Per page, JSON-LD types, validation result
robots.txt Findings — Allow / disallow, sitemap reference
sitemap.xml Findings — Completeness, accuracy, lastmod
Canonical Findings — Self-canonical, duplicate-page canonicals
hreflang Findings — If multi-lingual
Mobile-Friendly Findings — Viewport, tap targets, readability
Social Preview Findings — Actual rendered previews per platform
JavaScript Rendering Findings — Whether SEO content is SSR'd or hydrated

Close with a Prioritized SEO Fix List: top 10 by traffic-impact × effort. Distinguish quick-wins (set OG image) from structural (rewrite all meta descriptions).

Calibration

Output Format

Need help applying this to a real product?