SEO Audit Full

Page-Level Checks (in {{page_checks_html}}), output in this exact order:

Core checks: URL Slug · Title Tag · Meta Description · H1 Tag · Canonical Tag · Image Alt Text · Word Count · Keyword Placement · Heading Structure · Internal Links · Schema (JSON-LD)

Schema (JSON-LD) rules:

• Treat Schema as a quality check, not only a presence check.
• Validate JSON-LD parseability, expected @type, required fields, recommended rich-result fields, nested fields, and primary-type conflicts.
• Fail when JSON-LD is invalid, the expected schema type is missing, required fields are missing, or localized schema clearly points to the wrong language/URL.
• Warn when recommended fields are missing, nested fields are incomplete, multilingual pages lack inLanguage, or localized schema cannot be fully confirmed.
• Pass only when the expected schema type is present, required fields are present, no conflicts are found, and localized schema matches the current page language/URL when applicable.
• For multilingual pages, each language version should have its own schema with matching inLanguage, language-specific headline/description where present, and url / mainEntityOfPage pointing to the current localized canonical URL.

★ Full-only additions:

• OG Tags — og:title, og:description, og:image, og:type, og:url presence and validity
• Twitter Card — twitter:card type, title/description/image (with OG fallback detection)

How to Use Script JSON Output

Same rules across full audit modules — map each field's status directly to the report check table:

• status → pass / warn / fail / error → badge in report
• detail → starting point for Evidence line
• Do not contradict script output unless you have additional observable evidence

For check-site.py output:

• staging_subdomains.status → Staging Subdomain Indexation row status
• staging_subdomains.detail → Staging Subdomain Indexation row detail
• staging_subdomains.public_hosts → evidence for public staging/test hosts
• staging_subdomains.similar_hosts → fail evidence for production-like staging duplicates
• robots.status → robots.txt row status
• sitemap.status → sitemap.xml row status
• sitemap_inventory.status → Sitemap URL Inventory module status
• sitemap_inventory.directories[] → table rows with path, url_count, page_type, and example_url
• sitemap_inventory.detail → short explanatory note below the inventory table

For check-schema.py output:

• status → Schema (JSON-LD) row status
• detail → Schema (JSON-LD) row detail
• parse_errors → fail evidence for malformed JSON-LD
• schemas[].fields_missing → fail evidence for missing required schema fields
• schemas[].recommended_missing and schemas[].nested_issues → warning evidence
• localized_schema.status → language/URL alignment status for multilingual schema
• localized_schema.issues → evidence for schema language or URL mismatch

For check-social.py output:

• og.status → OG Tags row status
• twitter_card.status → Twitter Card row status
• og.fields.* → individual field details for the detail cell
• twitter_card.fields.* → individual field details, note fallback fields

For check-pagespeed.py output:

• status → overall Page Speed status
• category_status → Lighthouse Scores badge status
• metric_status → Core Lab Metrics badge status
• final_url → Final URL line
• categories.performance.score → Performance score
• categories.accessibility.score → Accessibility score
• categories["best-practices"].score → Best Practices score
• categories.seo.score → SEO score
• compact_metrics.fcp.display_value → FCP
• compact_metrics.lcp.display_value → LCP
• compact_metrics.tbt.display_value → TBT
• compact_metrics.cls.display_value → CLS
• compact_metrics.si.display_value → Speed Index
• screenshot truthy → Screenshot available

Do not use overall status for the Lighthouse Scores badge. Category scores use Lighthouse thresholds: 90–100 pass, 50–89 warn, 0–49 fail. Example: Performance 60, Accessibility 84, Best Practices 100, SEO 100 means category_status is warn. If LCP or Speed Index fails, metric_status and overall status may be fail while the Lighthouse Scores panel remains warn.

Inside {{pagespeed_checks_html}}, include a short Priority Actions list after the score/metric cards. Keep it to 2–4 concise items based on failing or warning PageSpeed fields:

• LCP slow → optimize hero media, preload critical image, reduce render-blocking CSS
• Speed Index slow → defer non-critical scripts and reduce above-the-fold JS/CSS
• TBT high → split long tasks and delay third-party tags
• Performance score warning/fail → prioritize the biggest Lighthouse opportunities

Always include the official diagnostic link: https://pagespeed.web.dev/

LLM Review Instructions

Core LLM reviews

Resolve every llm_review_required: true field before writing the report: H1 semantic judgment, Title keyword position, URL Slug evaluation, and Meta Description quality must all receive an explicit judgment.

Full-only LLM checks

OG Tags quality (always review):

og:title   : Does it differ meaningfully from <title>? It should be optimized for social sharing.
og:description : Is it compelling for social feeds? Different focus than meta description is OK.
og:image   : Is the URL an actual image path (not a page URL)?

Twitter Card completeness:

If twitter:card is "summary_large_image", twitter:image (or og:image fallback) must be
at least 300x157px. Flag if the image URL looks like a small icon or favicon.

Recommended Workflow

1. Acknowledge full scope — confirm this is a full audit
2. Infer primary keyword — read the page H1, title, and first paragraph unless the user provided one
3. Phase 1: Run core scripts — check-site → check-page → fetch-page → check-schema
4. Phase 2: Run full-only scripts — verify PageSpeed API key exists, then run check-pagespeed → check-social
5. Core checks — 404 handling, URL canonicalization, E-E-A-T trust pages, i18n/hreflang
6. PageSpeed checks — summarize Lighthouse category scores and lab metrics
7. LLM-only advanced checks — E-E-A-T content quality, duplicate content signals, anchor text quality
8. Summarize findings — Evidence / Impact / Fix format
9. Priority actions — top 5 highest-impact fixes with effort/impact tags
10. Render report — save to reports/<hostname>-<slug>-full-audit.html

Report Detail Writing Rules

Use strict formatting:

Pass → one short phrase. No lists, no elaboration.

Warn → one <div class="detail-issue"> with ≤2 bullet points. One <div class="detail-fix">.

Fail → same as Warn. Lead with the exact failure.

Mandatory Finding Format

**Finding: [Finding Title]**

- **Evidence:** [Observable fact, data point, or marked assumption]
- **Impact:** [SEO / UX consequence]
- **Fix:** [Actionable recommendation with example]

For Priority Actions, add effort/impact tags:

1. [High Impact / Low Effort] Fix og:image — social shares currently show no preview.

Reference Files

• Detailed audit modules and field definitions: references/REFERENCE.md
• Final HTML report template: assets/report-template.html
• PageSpeed validation script: scripts/check-pagespeed.py
• Social tags validation script: scripts/check-social.py
• Core scripts: scripts/ (check-site, check-page, check-schema, fetch-page)