Prompt 01
Lock the buyer, the awareness stage, and the seed keywords before any tool fires.
You are an SEO strategist. Interview me to extract the inputs I need to do keyword research for my business.
Ask, one at a time:
1. What does the business do (1-2 sentences)?
2. Who is the buyer (role, company size, what they currently use)?
3. What are the top 3 buyer pains in their own words (case-study quotes if you have them)?
4. What awareness stage are most prospects at (problem-aware, solution-aware, product-aware)?
5. Which 5 seed keywords represent how the buyer would describe their problem to a friend?
6. Who are the 3-5 competitors the buyer compares against?
When I'm done, summarise everything as a locked brief I can re-feed into future prompts.
Prompt 02
Pull seed-aligned keywords + question variants + related-search expansions, then dedupe into one buyer-aligned table.
Using the locked brief from Prompt 01, run the following DataForSEO Labs calls in parallel:
1. dataforseo_labs_google_keyword_ideas — all 5 seed keywords, location US, language en, limit 300, filter search_volume > 100, order_by search_volume desc.
2. Same call again, limit 200, filter regex ^(what|why|how|can|does)\s — capture question keywords only.
3. dataforseo_labs_google_related_keywords — one call per seed, depth 2, limit 100 each, filter search_volume > 200.
After data lands: dedupe across all result sets, build one table sorted by volume desc, columns = keyword | volume | KD | intent | CPC | source endpoint. Apply a B2B-relevance filter to exclude consumer noise (image/video/girlfriend/character/homework/detector). Target 80-120 rows.
Save to .planning/keyword-research/keyword-table.md.
Prompt 03
Quantify LLM-channel demand and identify which domains/pages get cited inside ChatGPT/Claude/Perplexity for the same seeds.
Now use the DataForSEO AI Optimization endpoints to see how this topic shows up in AI search. Use the seed keywords from Prompt 01.
1. ai_optimization_keyword_data_search_volume — pull LLM-era search volume for each seed, location United States, language en.
2. ai_opt_llm_ment_top_domains — for each seed, return the top 15 domains that LLMs (ChatGPT, Claude, Gemini, Perplexity) cite most often when answering questions about this topic.
3. ai_opt_llm_ment_top_pages — top 15 most-cited individual pages.
Return three tables, one per call. Then write a 4-bullet summary:
- Which domains dominate AI answers in this niche
- Which page formats LLMs prefer (docs, blog posts, GitHub, YouTube)
- Where my domain appears (or doesn't) in the citation list
- 2 or 3 specific pages we should study and beat
Prompt 04
Turn the merged keyword set into shippable content clusters with a pillar page + supporting posts each.
Take the buyer-aligned table from Prompt 02 and cluster every keyword into 6-10 topical groups (e.g., Pricing & Access, Install & Setup, Tutorial & How-to, Features Deep-Dive, Comparisons, Integration & SDK, Models, Status & Monitoring).
For each cluster, output:
- Cluster name
- Total monthly volume
- Average KD
- AI-search demand (yes / no, with the LLM volume number from Prompt 03)
- Pillar page brief: target keyword, intent, suggested H1, 3-5 supporting H2s
- Supporting blog post list: title, target keyword, volume, KD, suggested word count
Sort clusters by total volume desc. Flag any cluster whose top keyword has KD ≤ 15 — those are quick-win content priorities.
Prompt 05
Find the keywords where direct competitors rank in the top 20 and theaiarchitects.com doesn't rank at all.
Two-step competitor gap analysis for theaiarchitects.com:
Step 1: Use serp_competitors to identify the top 10 domains that compete with theaiarchitects.com across the keyword clusters from Prompt 04. Pick the 3 most relevant direct competitors.
Step 2: Use domain_intersection (intersections: false) to compare theaiarchitects.com against those 3 competitors. Show the top 30 keywords where any competitor ranks in the top 20 organic results AND theaiarchitects.com does not rank at all.
Return columns: keyword, competitor domain, competitor rank, monthly volume, KD, search intent, suggested cluster (from Prompt 04). Sort by volume descending.
Prompt 06
Hand a writer (or a Claude Code skill) a single document they can act on tomorrow morning.
Save the full SEO strategy as a markdown file at docs/SEO-STRATEGY-V2.md. Structure it as:
1. Discovery summary (business, buyer, awareness stage, seed keywords from Prompt 01)
2. Executive summary (3 bullets: total monthly search opportunity, number of clusters, top 3 quick wins)
3. Cluster overview table (name, total volume, target page count, average KD, AI-search demand y/n)
4. Per-cluster sections with:
- Pillar page brief (target keyword, intent, suggested H1, 3-5 supporting H2s)
- Supporting blog post list (title, target keyword, volume, KD, suggested word count)
5. AI-search section: top 5 pages we should write specifically to win LLM citations (from Prompt 03)
6. Quick wins section: top 10 keywords with KD under 25 and volume above 500
7. Gap-analysis appendix: full competitor gap list from Prompt 05
Format it for someone who will hand this to a writer (or a Claude Code skill) tomorrow morning.