AEO Strategy

Produce a research-backed content strategy for search + AI visibility. Output: cluster taxonomy, keyword gap analysis, competitor content audit, prioritised article queue with target keywords, and a 90-day publishing timeline. Level 1 strategy skill — produces what to write; the Level 2 aeo-content skill produces the actual content from this queue.

When to run

User says "AEO strategy", "content roadmap", "keyword gap analysis", "what should [company] publish", "90-day content plan for search", "article queue for [company]", or "what content will get us cited by AI?". Skip if they want a single article (/aeo-content), channel-agnostic plan (/content-strategy), or pure SEO audit (web search). Full trigger list: the premium reference.

Inputs

Required: company context (product, features) · 3-5 competitors with URLs · GSC exports (Queries CSV, Pages CSV, Coverage) · GA4 exports (Traffic acquisition, Landing pages). Without GSC/GA4, all keyword data must be marked [ESTIMATED] and strategy confidence flagged in the executive summary.

Recommended: ICP research · existing content audit · positioning/messaging · competitor-research · content-strategy · win-loss-analysis · transcript-analysis (last two enable Phase 2.5 query seeding).

Optional: prior keyword research · monthly content target (override default 20) · AirOps brand kit.

Full export instructions (exact GSC/GA4 click paths, columns, used-in-phase mapping) + validation checklist: the premium reference.

Steps

Eight phases. Each phase has sub-steps, output templates, and checkpoints in the premium reference — read it before producing the output.

1. Phase 1 — Input validation. Load company context · confirm 3-5 competitors with geography tags · set parameters (default: 20/month, 90 days, 50/20/30 BOFU/MOFU/TOFU split; adjust by maturity).
2. Phase 2 — Cluster taxonomy. Build 5-8 product-mapped pillars (not generic categories), each with 5-10 sub-clusters from product docs + ICP pains + competitor content.
3. Phase 2.5 — Transcript query seeding (only if win-loss/transcript outputs exist). Mine buyer questions, convert to natural-language queries, tag by cluster + source, merge into aeo/query-index.md (or seed it).
4. Phase 3 — SEO keyword gap analysis. Pull client baseline (GSC) + competitor keywords (DataForSEO → Apify → Exa fallback) · identify gaps · group by cluster · separate by geography · pull AirOps AI-citation baseline · write 1-paragraph competitor keyword strategy summaries.
5. Phase 4 — Competitor content analysis. Enumerate each competitor's pages free with mcp__spider__spider_links (per .claude/rules/crawl-cost-discipline.md), triage to in-scope pages, then Firecrawl only the kept set · LLM-classify pages by funnel stage (TOFU/MOFU/BOFU) AND cluster · build funnel-stage and cluster comparison tables · identify zero-coverage clusters and underweight stages.
6. Phase 5 — Competitor best performing content. Exa + Firecrawl + DataForSEO traffic estimates → pick top 5 per competitor · tag each with funnel stage + cluster · write 1-paragraph strategy analysis per competitor.
7. Phase 6 — Content type strategy. Define types per stage (BOFU: comparisons, branded, integrations, pricing, demo · MOFU: how-to, "Best X for Y", deep-dives, use cases, compliance · TOFU: definitions, industry guides, regulatory, thought leadership) · set % allocation grounded in Phase 4 gaps · write rationale per type.
8. Phase 7 — Article queue. Generate specific titles per type with target keywords. BOFU: 1 comparison per competitor + branded pages. MOFU: how-tos mapped to features + "Best for [year]" listicles, flag UPDATE vs CREATE. TOFU: definitions targeting highest-volume gap keywords + industry/regulatory guides.
9. Phase 8 — 90-day timeline. Assign articles to months (M1: comparisons + high-volume TOFU + initial MOFU · M2: BOFU deep-dives + continue MOFU/TOFU · M3: remaining BOFU + integrations + TOFU depth) · build period × stage summary table.

Anti-hallucination guardrails (must apply during execution): keyword volumes sourced or marked [ESTIMATED: based on SERP analysis] · page counts from crawl data only · no invented competitor content (verify via Exa/Firecrawl) · article titles flagged as suggestions · explicitly note approximated research. Full guardrails + per-phase checkpoint lists: the premium reference and the premium reference.

What good looks like

Output structure (the premium reference for full template): # [X]-day content roadmap for [Company] → Executive summary (2-3 sentences + content-type table) → Research (A. clusters · B. keyword gaps · C. competitor content · D. competitor best performing) → Strategy and content examples (rationale + article tables) → Timeline (monthly + summary) → Iteration prompts.

Design principles: executive summary table first (10-second scan) · research before strategy · clean tables over prose walls · specific ready-to-brief titles with target keywords (not categories) · simple monthly timeline.

Pre-delivery quality gates (full checklists in the premium reference):

• Research: 5-8 clusters mapped to product features · keyword gaps with volumes · competitor crawl page counts · pages classified by stage AND cluster · top 5 per competitor.
• Strategy: allocation sums to 100% · monthly counts realistic · rationale grounded in Phases 3-5 (not generic) · types appropriate for client maturity.
• Execution: specific titles (not categories) · target keywords per article · UPDATE vs CREATE flags · monthly priorities actionable immediately · summary totals mathematically correct.

Evaluations: allocation percentages add to 100 · article-count totals match Phase 6 allocation across Phases 7 and 8 · every keyword volume tagged [VERIFIED] or [ESTIMATED] · every competitor page count traceable to Firecrawl crawl or marked [ESTIMATED] · ≥3 sources per major Exa-derived claim with [VERIFIED: exa_search, {url}, accessed {YYYY-MM-DD}] per .claude/rules/exa-protocol.md.

Final ship gate

Run /premortem --output before ship. See /premortem skill for the 5 execution domains (will-it-resonate / will-it-convert / will-it-stay-on-brand / will-stakeholder-push-back / will-it-degrade-over-time) and output template.

Trivial-case escape: ## Premortem\nNo failure modes — trivial change satisfies the contract for genuinely trivial outputs.