AEO Prompt Research (Free)

Source: github.com/psyduckler/aeo-skills

Part of: AEO Skills Suite — Prompt Research → Content → Analytics

Discover which prompts and topics matter for a brand's AI visibility — using zero paid APIs.

Requirements

• web_fetch — crawl the target site
• web_search — Brave Search free tier (optional but recommended)
• LLM reasoning — the agent's own model does the heavy lifting

Workflow

Input

The user provides:

• Domain URL (required) — e.g. clearscope.io
• Niche/category (optional) — e.g. "SEO software for content teams"
• Competitors (optional) — e.g. "Surfer SEO, MarketMuse, Frase"

Step 1: Crawl the Site

Use web_fetch on key pages to understand the brand:

Pages to fetch (try each, skip 404s):
- / (homepage)
- /about or /about-us
- /pricing
- /products or /features or /services
- /blog (index only)

Alternatively, run scripts/crawl_site.sh for a batch crawl.

Extract from crawled content:

• Core product/service offering
• Target audience (industry, company size, persona)
• Key differentiators / value props
• Competitor mentions
• Content themes from blog titles

Step 2: Discover the Topic Universe

Using the brand understanding, brainstorm topic categories. For methodology and category types, read references/aeo-methodology.md.

Core prompt categories to generate:

• Problem-aware — "How do I solve [problem]?"
• Solution-aware — "What tools exist for [category]?"
• Comparison — "[Brand] vs [competitor]"
• Best-of — "Best [category] for [use case]"
• How-to — "How to [task the product helps with]"
• Evaluation — "Is [brand] good for [need]?"
• Industry — "[Industry] trends / best practices"

Step 3: Generate Prompts

For each category, generate 5-15 specific prompts people would actually ask an AI assistant.

Guidelines:

• Write naturally — how people talk to ChatGPT, not how they Google
• Be specific — include context (company size, industry, use case)
• Vary intent — research, comparison, how-to, buying decision
• Avoid jargon-heavy or unrealistic prompts

Step 4: Prioritize

Score each prompt (1-5) on:

• Relevance — How closely tied to the brand's core offering?
• Volume potential — How many people likely ask this?
• Winability — Can this brand realistically be the best answer?
• Intent value — Does this indicate buying/conversion intent?

Sort into Tier 1 (≥3.5), Tier 2 (2.5-3.4), Tier 3 (<2.5).

Step 5: Audit Existing Coverage

For Tier 1 prompts, use web_search with site:domain.com [topic keywords] to check if content already exists.

Rate coverage:

• Strong — Dedicated page directly answers the prompt
• Partial — Related content exists but doesn't fully address it
• None — No relevant content found

Step 6: Deliver Results

Output a structured report with:

• Brand summary (2-3 sentences)
• Prioritized prompt list with scores and coverage status
• Content gap analysis (high-priority prompts with no coverage)
• Top 5 recommended content pieces to create first

Tips for Better Results

• If web_search is unavailable, the skill still works — just skip the coverage audit or have the user manually check
• For competitor analysis, crawl competitor sites too and compare topic coverage
• Re-run quarterly — AI prompt trends shift as models and user behavior evolve
• The agent's own knowledge of the industry is a valid research input — use it