| name | geo-seo-audit |
| description | SEO + GEO audit and implementation workflow — Core Web Vitals, HTML structure, semantic writing, E-E-A-T, content intent, Wikipedia/Wikidata, JSON-LD schema, meta tags, crawl budget, robots.txt, sitemap, IndexNow, topic cluster architecture, link building, brand signals, conversion optimization, analytics, internationalization, platform presence, defensive review |
SEO + GEO Audit
This skill covers both traditional SEO and Generative Engine Optimization (GEO) — ensuring advocacy content ranks in search engines AND appears in AI answer systems (ChatGPT, Perplexity, Google AI Overviews, Claude, Gemini, Bing Copilot).
The core insight: Only 17-32% of AI Overview citations come from pages ranking in the organic top 10. Domain Authority correlates with AI citations at r=0.18; topical authority (r=0.40) and branded web mentions (r=0.664) are the real predictors. 80% of URLs cited by AI assistants do not rank in Google's top search results. This means SEO and GEO have different — but overlapping — optimization paths.
When to Use
- When building or reviewing any public-facing advocacy website
- Before launching content that needs to be discoverable by search engines or AI systems
- When diagnosing why a site is not appearing in Google results or AI responses
- When implementing a content hub or topic cluster for an advocacy campaign
- As a pre-launch checklist for any new Open Paws platform page
Process
Step 1: Audit Core Web Vitals
Measure using real field data from Google Search Console (CrUX), not lab tools — Google ranks on field data. Current thresholds (as of March 2026):
| Metric | Good | Needs Improvement | Poor |
|---|
| LCP | ≤ 2.5s | 2.5–4.0s | > 4.0s |
| INP | ≤ 200ms | 200–500ms | > 500ms |
| CLS | ≤ 0.1 | 0.1–0.25 | > 0.25 |
Check: Is LCP element preloaded with <link rel="preload">? Are explicit width and height on every image/video/iframe (prevents CLS)? Is TTFB under 200ms? Are JavaScript long tasks (>50ms) broken up with scheduler.yield() or setTimeout? Is font-display: swap used for web fonts? Are below-fold images lazy-loaded? Are modern image formats (WebP/AVIF) used?
INP fix priority order: (1) audit third-party scripts, (2) lazy-load non-critical features, (3) optimize event handlers with scheduler.yield(), (4) reduce JavaScript payload with React Server Components, (5) virtual lists and CSS containment.
Next.js rendering check:
| Strategy | TTFB | Best For |
|---|
| SSG (default cached fetch) | Sub-100ms | Static content — blogs, docs |
ISR (next: { revalidate: 60 }) | Fast, cached | Periodically updated content |
SSR (cache: 'no-store') | 200ms+ | User-specific real-time data |
| PPR (experimental) | Fast shell + streamed | Mixed static/dynamic pages |
Sites with INP above 200ms average -0.8 position drops. Pages with LCP above 3 seconds experience 23% more traffic loss. 43% of sites still fail the 200ms INP threshold — the most common CWV failure.
Step 2: Audit Technical SEO
Rendering:
- Verify SSR or SSG — client-side-only rendering increases crawl cost and is invisible to AI crawlers
- Test with Google's URL Inspection tool to verify rendered content matches expectations
- AI crawlers (GPTBot, ClaudeBot, PerplexityBot) generally do not execute JavaScript
Mobile-first:
- Audit using Googlebot Smartphone user agent, not desktop
- Google uses mobile scores as the primary ranking signal for all results
- 53% of mobile visitors abandon sites taking more than 3 seconds to load
Crawl budget:
- Check for faceted navigation URL explosion (filter/sort parameter combinations)
- Verify robots.txt blocks low-value parameter URLs, internal search, admin pages
- Check for redirect chains (sequential redirects waste crawl budget)
- Verify XML sitemap contains only canonical, indexable URLs with accurate
<lastmod> dates
- Audit pages with zero organic traffic over 12 months — candidates for consolidation, noindex, or removal
- Verify HTTP status codes: 200 for live content, 301 for permanent redirects, 404 for missing, 410 for intentionally removed
Security headers (set via Next.js middleware):
- CSP with nonces for scripts/styles
Strict-Transport-Security: max-age=63072000; includeSubDomains; preloadX-Content-Type-Options: nosniffX-Frame-Options: DENYReferrer-Policy: strict-origin-when-cross-originPermissions-Policy: camera=(), microphone=(), geolocation=()Cross-Origin-Opener-Policy: same-origin- Remove
X-Powered-By header
Supply chain:
- Pin exact dependency versions, use
npm ci in CI pipelines
- Scan with Socket.dev or Snyk
- Enable 2FA on npm accounts
Step 3: Audit HTML Structure
Check each page for:
- Exactly one
<h1> tag containing the primary topic
- Logical heading hierarchy (
h1 > h2 > h3) with no skipped levels
<h2> headings phrased as questions where the section answers something (7× citation impact for smaller sites; also improves Featured Snippet and PAA selection)- First paragraph after each heading: direct 40-60 word answer
- Paragraphs of 2-4 sentences (40-60 words); content sections of 120-180 words
- Semantic elements:
<article>, <section>, <nav>, <aside>, <header>, <footer>, <main>
lang attribute on <html> tag- Descriptive
alt text on every <img>; descriptive file names with keywords
- Meaningful anchor text on every
<a> — never "click here"
<table> for comparison data; <ol> and <ul> for lists; <blockquote cite="..."> for quotations<time datetime="YYYY-MM-DD"> for dates; <dfn> for definitions; <abbr title="..."> for acronymsid attributes on all <h2> and <h3> elements
Flag any content rendered exclusively by JavaScript.
Step 4: Audit Content Intent and Search Positioning
Before auditing content quality, verify content targets the right intent format. Study the top 5 results for each target keyword. Content targeting the wrong intent format does not rank regardless of other optimizations.
| Intent type | Expected format to match |
|---|
| Informational | How-to guides, tutorials, explainers, definitions |
| Commercial investigation | Comparison articles, review roundups, "best of" lists |
| Transactional | Product/service pages, pricing pages, application forms |
| Navigational | Homepage, specific brand/product pages |
Also check: Does the content completely satisfy user intent in one visit (Helpful Content System standard)? Does it demonstrate genuine first-hand knowledge and experience? Does it include specific, dated citations rather than vague attribution?
Step 5: Audit Semantic Writing Quality
AI citation systems retrieve at the sentence and paragraph level. Check:
Entity salience: Is the primary entity the grammatical subject in active voice? Active voice gives subject salience 0.74; passive voice drops it to 0.11. Check for passive constructions and fix them.
Atomic claims: Are sentences self-contained semantic triples? Eliminate vague pronouns — every sentence must make sense in isolation.
Proper noun density: AI-cited text averages 20.6% proper nouns versus 5-8% standard. Check whether content names organizations, researchers, reports, and years specifically rather than using vague attribution.
Content density: Pages under 5,000 characters get approximately 66% of content used; pages over 20,000 characters get only 12%. Check whether pages are focused and dense.
Opening structure: Every major section should open with a 40-60 word declarative answer. Information buried deep in paragraphs is rarely retrieved by AI systems.
Step 6: Audit E-E-A-T Signals
Check for Experience, Expertise, Authoritativeness, Trustworthiness on every content page:
- Original data, case studies, screenshots, specific processes (Experience)
- Detailed author bio with verifiable credentials (Expertise)
- Specific citations from credible sources with names, organizations, and dates (Expertise)
- Third-party coverage, organizational partnerships, industry recognition (Authoritativeness)
- Clear contact information, privacy policy, transparent organizational identity (Trustworthiness)
Author attribution audit: Every content page must have a visible author name, credentials, and link to an author profile page. The author page must have Person schema with name, url, jobTitle, description, image, and sameAs (LinkedIn, external profiles). Content with proper author metadata gets 40% more AI citations.
Step 7: Audit Wikipedia and Wikidata Presence
Wikipedia accounts for 47.9% of ChatGPT's top-10 cited sources. Organizations visible in Wikidata get Knowledge Panels within 7 days and begin appearing consistently in AI answers.
Check:
- Does the organization have a Wikipedia article? Is it accurate, cited, current?
- Does the organization have a Wikidata entry (Q-ID)? Is it complete with: type, founding date, location, founders, official website, social profiles?
- Is the Wikidata Q-ID in the site's Organization schema
sameAs array?
- Is Wikipedia URL in the site's Organization schema
sameAs array?
- Is there an entity web: organization → tools → people → related orgs → policy areas?
- Does the site's structured data match Wikipedia and Wikidata? Inconsistency reduces AI confidence.
If no Wikipedia article exists: document as High finding. Check whether sufficient independent secondary coverage exists to support notability. If yes, creating the article should be prioritized over almost any on-site optimization.
Wikipedia COI policy (mandatory): Never directly edit your own organization's Wikipedia article. Disclose any affiliation on the Talk page. Propose edits through Talk-page requests or experienced neutral editors. Use only independent, reliable sources. Follow Wikipedia's Conflict of Interest and Notability guidelines. Document the disclosure and source list for audit purposes.
Step 8: Audit Structured Data (JSON-LD)
Sites with structured data achieve 41% AI citation rates vs 15% without. Only 12.4% of websites implement it.
For every page verify:
- Organization + WebSite schema on every page:
name, url, logo, sameAs (LinkedIn, Twitter, GitHub, Wikipedia URL, Wikidata URL), description
- Article schema on every content page:
headline, author (with name, url, jobTitle), publisher, datePublished, dateModified, image, description
- FAQPage schema on any page with Q&A content
- BreadcrumbList schema on all non-homepage pages
- Person schema on all author/team profile pages
- Additional applicable types: HowTo, SoftwareApplication, Event, Dataset, VideoObject
Validate all schema at https://validator.schema.org/ and Google's Rich Results Test. dateModified must reflect actual update dates.
Step 9: Audit Meta Tags and Head Elements
For every page:
<title>: Primary Keyword — Brand Name, 50-60 chars, keywords first, unique per page<meta name="description">: 150-160 chars, direct answer to primary query, one statistic, unique<link rel="canonical"> pointing to canonical URL- Open Graph tags:
og:title, og:description, og:type, og:url, og:image, og:site_name
- Twitter Card:
twitter:card, twitter:title, twitter:description, twitter:image
- Article timestamps:
article:published_time, article:modified_time (ISO 8601)
Step 10: Audit Robots.txt
Verify AI citation crawlers are allowed:
- Must allow (power AI answers):
OAI-SearchBot, ChatGPT-User, PerplexityBot, ClaudeBot, Claude-SearchBot, Applebot, Amazonbot
- May block (training only):
GPTBot, CCBot, Google-Extended, meta-externalagent, Bytespider
- Never block:
Googlebot (blocks both Search and AI Overviews simultaneously)
Note: 226+ AI crawlers exist. Verify list is current. Some AI agents use standard browser user-agent strings and ignore robots.txt.
Step 11: Audit XML Sitemap and IndexNow
Sitemap:
- Exists at
/sitemap.xml
- Contains only canonical, indexable URLs
<lastmod> dates reflect actual content update dates (never faked)- Referenced in robots.txt
- Submitted to Google Search Console and Bing Webmaster Tools
- Auto-regenerates when content changes
IndexNow:
- API key file at
https://yoursite.com/{key}.txt
- Ping fires on every publish/update event
- Integrated in CI/CD pipeline or CMS publish hooks
Step 12: Audit Site Architecture and Content Freshness
Architecture:
- URL structure: descriptive hyphenated lowercase, under 75 characters, max 3 levels deep, canonical tags, 301 redirects for changed URLs
- No important page more than 3 clicks from homepage
- Breadcrumb navigation with BreadcrumbList schema
- Topic cluster structure: pillar pages + 8-15 cluster pages with bidirectional links?
- Orphan pages (no internal links pointing to them)?
Content freshness:
- "Last Updated: [date]" visible on every content page using
<time datetime="YYYY-MM-DD">
dateModified in Article schema synchronized with visible dates- Content genuinely updated (not just date-changed) — Google detects date freshness hacking
- 76% of most-cited AI content updated within 30 days; Perplexity gives 3.4× advantage to content under 30 days old
First-mover check: For emerging advocacy topics with few authoritative sources, is the site publishing before competitors claim citation positions?
Step 13: Audit Platform Presence and Brand Signals
85% of AI brand mentions come from third-party pages. Brand mentions now account for 55% of off-page ranking weight.
Check:
- Reddit: Active, authentic presence in relevant subreddits? AI systems have visibility into Reddit's moderation pipeline — inauthentic content carries negative weight.
- YouTube: Channel with transcripts enabled?
- LinkedIn: Active organization page with substantive posts?
- GitHub: Documentation, datasets, or reports published as Markdown?
- Unlinked brand mentions: Monitor with Google Alerts or Ahrefs. Convert high-authority mentions to backlinks — close rate typically above 30%.
- Cross-platform consistency: Key facts (founding date, mission, statistics) consistent across site, Wikipedia, Wikidata, LinkedIn, GitHub?
Platform-specific behavior: Only 11% of domains are cited by both ChatGPT and Perplexity for the same queries. Citation volatility is 40-60% monthly. Build multi-platform presence.
Step 14: Audit Link Profile
Backlinks remain the #2 ranking factor. Topical relevance between linking and linked domains (r=0.4) now outweighs raw domain authority.
Check:
- Natural anchor text distribution: ~40-50% branded, ~20-30% generic, ~15-25% partial match, ~5-10% exact match. Sites below 30% diversity saw -15 average position drops in March 2026 spam update.
- Links in editorial context (body content), not footers/sidebars/author bios
- Top-performing link building tactics: original research/data (156% more links vs generic how-to), digital PR campaigns, unlinked mention conversion
- Any links from PBNs, sponsored guest posts on generalist sites, or niche edits on thin domains — these were specifically devalued in March 2026
Step 15: Audit Content Patterns
Check whether content templates implement high-citation patterns:
Citable paragraph: every major claim follows [fact]. [statistic with attribution]. [elaboration]. [Source: Org, Date]
FAQ block: <h2> with exact question, 40-60 word direct answer immediately following, FAQPage schema. Also improves Featured Snippet and People Also Ask selection.
Definition block: <section id="what-is-term"><h2>What is [Term]?</h2><p><dfn>[Term]</dfn> is [direct definition].</p></section>
Original data: Pages with proprietary data get 4.31× more citations per URL. Check whether any unique organizational data can become dedicated, permanently-addressable pages.
VideoObject schema: If any video content exists, verify VideoObject schema with name, description, thumbnailUrl, uploadDate, and contentUrl.
Step 16: Audit Conversion Optimization and Analytics
Donation page audit:
- 3-4 preset amounts with middle pre-selected and impact descriptions?
- Monthly giving pre-selected? (31% of nonprofit online revenue; 64% of orgs still default one-time)
- Single-step form? (multi-step forms see 52% drop in completions)
- Site header navigation removed during donation flow?
- Form embedded on-site, not redirecting to third-party processor?
autoComplete attributes on all form fields?
Analytics:
- Using cookieless analytics (Plausible or Umami) as primary?
- AI referral traffic tracked with custom channel group? (ChatGPT 78% of AI traffic; mobile app drops referrer)
- Key conversion events marked in GA4:
donation_completed (with value), newsletter_signup, volunteer_form_submit?
Internationalization check (if multilingual):
lang and dir attributes set correctly on <html>?- Hreflang tags self-referencing and reciprocal on every page?
- Subdirectory URL strategy (
/en/, /hi/, /ar/) for domain authority consolidation?
- ICU MessageFormat used for plural/gender forms (Arabic requires 6 CLDR plural categories)?
Step 17: Defensive Review
Check for techniques that would violate platform guidelines:
Hidden text: Verify no content is hidden via display:none, visibility:hidden, white-on-white text, zero-size fonts, negative positioning, or invisible Unicode (U+E0000 to U+E007F). Google's SpamBrain and PhantomLint compare parsed text against OCR-rendered images.
Agent-aware cloaking: Verify the server does not serve different content to AI crawlers vs human visitors based on User-Agent detection. Explicitly prohibited; domain-wide penalty risk.
Scaled AI content: Verify all published content has meaningful human review. Manual actions issued by Google since June 2025 for "scaled content abuse."
User-generated content injection: If the site hosts comments or forum posts, verify they are sanitized before being served to crawlers.
Findings Report
Document findings by priority:
Critical (blocks launch)
- Primary content rendered JavaScript-only
- Missing
<h1> or broken heading hierarchy
- Missing canonical tags
- robots.txt blocking AI citation crawlers
- Missing HTTPS or expired SSL
- LCP > 4s or page weight > 3MB
- Agent-aware cloaking detected
High (implement before significant content investment)
- LCP > 2.5s (failing Good threshold)
- INP > 200ms (43% of sites fail this)
- No Wikipedia article when sufficient notability sources exist
- No Wikidata entry
- Missing structured data (JSON-LD schema)
- Content siloed with no topic cluster architecture
- Pages not SSR or SSG
- Missing
sameAs links to Wikipedia/Wikidata in Organization schema
- Content not matching search intent format for target keywords
Medium (implement in next sprint)
- Question-format headings not used
- Answer-first paragraph pattern not followed
- Missing entity salience (passive voice, vague pronouns, weak proper noun density)
- Content density outside optimal range
- Missing E-E-A-T signals (author attribution, sourcing, experience signals)
- Missing author profile pages with Person schema
- Missing IndexNow integration
- Outdated or inconsistent
dateModified values
- No Reddit, YouTube, or LinkedIn presence
- Key facts inconsistent across platforms
- Anchor text distribution outside recommended ranges
- Donation page using multi-step form or missing monthly pre-selection
- No AI referral traffic tracking in analytics
- Missing cookieless analytics (relying solely on GA4 with consent banner)
- Missing hreflang tags on multilingual pages
Low (optimize over time)
- llms.txt not implemented (low current value but zero-cost)
- Individual content not following citable paragraph pattern
- Missing comparison tables on decision-relevant pages
- No proprietary data or original research pages
- VideoObject schema not implemented for video content
- Security headers incomplete (CSP, Permissions-Policy, COOP)
- Supply chain scanning not in CI pipeline
For each finding: the specific page or component affected, what is missing or incorrect, and the exact implementation needed.
