Skip to main content
Generalfusengine

seo-sxo

Use when optimizing search experience (SXO). Covers intent matching, user personas, user stories, page-type analysis, dwell time, scroll depth, pogo-sticking prevention.

Stars
13
Source
fusengine/agents
Updated
2026-05-17
Slug
fusengine--agents--seo-sxo
View on GitHubRaw SKILL.md

// install — copy + paste into any project

mkdir -p .claude/skills && curl -fsSL https://raw.githubusercontent.com/fusengine/agents/HEAD/plugins/seo/skills/seo-sxo/SKILL.md -o .claude/skills/seo-sxo.md

Drops the SKILL.md into .claude/skills/seo-sxo.md. Works with Claude Code, Cursor, and any agent that loads SKILL.md files from .claude/skills/.

Search Experience Optimization (SXO)

Intent Matching (5 layers, 2026)

Layer Intent Expected Page Key Signals CTA
L1 Informational Guide, tutorial Quick answer, TOC, depth Newsletter / next article
L2 Navigational Homepage, brand page Logo, navigation, search Direct link to feature
L3 Commercial Comparison, reviews Tables, pros/cons, ratings Trial / demo / comparator
L4 Transactional Product, pricing CTA, price, trust signals Low-friction purchase
L5 Post-conversion Docs, FAQ, troubleshoot Steps, search, error fixes Support / upsell / community
  • >80% of queries carry mixed / multi-layer intent — map the dominant layer, then serve secondary layers below the fold.
  • Conflict rule: when query modifiers disagree with the live SERP layout, trust the SERP layout (it reflects Google's resolved intent).

Local vs Global Intent (2026)

Dimension Local Global / National
Dominant intent Immediate transactional + proximity navigational Informational + commercial → deferred transactional
SERP trigger Map Pack (3-Pack) + local AI Overviews Organic results + AI Overviews
Behavioural signal "[service] near me", "[service] [city]", direct call/tap "[service] reviews", "best [service]", "how to choose [service]"
Conversion speed Fast (ready to act now) Slow (nurture across buyer journey)
AI Overview CTR erosion Low — Map Pack resists (physical intent can't be answered by AI) High — informational keywords absorbed (-40% to -58% in 2026)
  • Diagnose intent geography first: a query with implicit local intent must serve a location page + Map Pack assets, not a blog post.
  • Protect against erosion: shift global informational pages toward entity signals + structured data so they earn AI citations instead of losing the click.

User Persona Method

  1. Define 2-3 personas (role, goals, pain points)
  2. Map each persona to query types they search
  3. Build pages addressing their specific intent
  4. Validate with user testing (heatmaps, session recordings)

User Stories Format

As a <persona>, I search "<query>" because I need <goal>.
I expect to find <expected content> within <time>.
I'll engage by <action: scroll, click, share, convert>.

SXO Metrics

  • Dwell time: > 1 minute for informational, > 3 for in-depth guides
  • Scroll depth: > 50% reach mid-page, > 25% reach bottom
  • Pogo-sticking: < 10% (users returning to SERP)
  • Click-through (next step): > 30% click expected CTA
  • CWV → conversion: speed is UX — every 1s of extra delay = ~-7% conversions (keep LCP < 2.5s, INP < 200ms)

Conversion Levers (local transactional pages)

Transactional intent = a minimum-friction action CTA above the fold. Answer-first text serves the AI Overview and the click; the CTA + proof serve the conversion.

  • Explicit CTA above the fold, repeated (top + after pricing): "Call now — [phone]" / "Get a free quote".
  • Native click-to-call tel:[phone] — most urgent local traffic is mobile.
  • Urgency / live availability: "Available now", real-time slots.
  • Friction reduction next to the CTA: "No commitment · Reply in 2 min · Pay after service".
  • Trust proof glued to the CTA: rating [x.x]/5 ([n] reviews), warranty/guarantee, license [number].

Anti-Patterns

  • ❌ Above-fold pop-ups (delay 30s+ or scroll-trigger)
  • ❌ Wall of text (paragraphs > 4 lines)
  • ❌ Generic headlines (specific > clever)
  • ❌ Buried answer (quick answer should be in first 100 words)
  • ❌ Aggressive selling on an informational page (intent mismatch) — match the CTA to the intent layer