AEO/GEO Playbook
Intent
Build a repeatable AI-visibility system. Any startup can use this to become the answer ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews give for its category queries.
Inputs
- Target company name and website
- Top 10 queries in its category
- Top 5 competitors
- Current AI visibility baseline (screenshots + data)
Process (SOP)
- Audit the baseline: query 50 category prompts across the target engines and record mentions, citations, and competitors.
- Map every query to a content module and design the pillar-cluster, llms.txt, schema, and semantic HTML.
- Produce 14 content modules: long-form articles, short answers, case studies, and teardowns.
- Distribute: publish the modules, implement llms.txt and JSON-LD, add internal links, and allow AI crawlers.
- Measure weekly and refresh the modules that do not earn citations.
If an engine blocks crawling, fix robots.txt and server access before producing more content. When to stop: if mention rate does not move after 8 weeks, audit query fit and citation sources before scaling production.
Phase 1: Baseline Audit (Week 1)
Query each AI engine with 50 target queries. Record:
- Is the brand mentioned? (Y/N)
- Which competitor is cited instead?
- What sources are cited?
- Screenshots + timestamps + raw responses
Output: baseline-audit.md with raw data + gap analysis.
Phase 2: Content Engine Blueprint (Week 2)
Map every query to a content module. Design:
- Pillar-cluster architecture (topic clusters with internal linking)
- llms.txt (structured index for AI agents)
- JSON-LD schema (Organization, WebSite, Article, FAQ)
- Semantic HTML (proper heading hierarchy, code examples, tables)
Output: content-engine-blueprint.md
Phase 3: Production (Weeks 3-4)
Produce 14 content modules:
- 5 long-form articles (2,000–5,000 words, structured for AI citation)
- 5 short-form answers (200–500 words, direct answers to specific queries)
- 2 case studies (proof of work, citation-ready structure)
- 2 competitive teardowns (structured for citation)
Output: Published content with full schema.
Phase 4: Distribution (Week 5)
- Implement llms.txt
- Add JSON-LD schema to all pages
- Optimize semantic HTML
- Build internal linking web (3-5 internal links per article)
- Submit to AI crawlers (allow GPTBot, ClaudeBot, PerplexityBot)
Output: Live pages with technical optimization.
Phase 5: Measurement (Week 6+)
Weekly tracking:
- AI mention rate per engine (target: high share for category queries)
- Citation count (who cites you, for what)
- Traffic from AI sources (Google Analytics)
- Conversion from AI traffic
Monthly:
- Traffic from AI sources (track trend)
- Conversion from AI traffic (track trend)
Output: measurement-dashboard.md
Outputs
- Baseline audit report
- Content engine blueprint
- 10 published content modules
- llms.txt + JSON-LD + semantic HTML
- Measurement dashboard with 6 weeks of data
Pre-flight checklist:
- 50 baseline queries captured with screenshots
- Every query mapped to a module
- llms.txt and JSON-LD validated
- AI crawlers allowed
- Weekly measurement scheduled
Evidence
AEO/GEO emerged as a distinct discipline as answer engines began citing sources directly. The mechanics differ from SEO: the optimization target is citation, not keyword rank, so content must be structured for machine consumption (llms.txt, schema, clear answer framing).
Examples
Query: "What's the best AI customer support platform?" Before: Brand mentioned in 0/5 engines. After: Brand mentioned in 5/5 engines, cited as a top answer in 4/5.
Metrics
- AI mention rate (weekly per engine). Target: 30%+ of priority queries by week 12.
- Citation count (domains citing you). Target: 10+ distinct domains by week 12.
- Traffic from AI sources (monthly). Target: 20% month-over-month growth.
- Conversion from AI traffic (monthly). Target: 2%+.
Anti-Patterns
- Optimizing for keyword rank instead of citation
- Skipping llms.txt (AI agents read it before crawling page content)
- Skipping measurement (no data means no proof of impact)
- Treating AEO as SEO with a new name (the mechanics differ: citation, not rank)