Auto mode: Leave sub-services empty — Claude will generate 6 relevant sub-services for this industry automatically.

Hub mode: Every service gets a dedicated page, all linked inside the hub folder.

Full site: 10–20 services work best. AI organizes them into categories — every service gets its own SEO page.

Cluster format: Each line = one cluster. Use Cluster Name :: problem1, problem2, problem3. Each problem becomes a separate SEO page.

How it works: the plan's mode5_pseo.areas lists every area that has its OWN search demand — one page each, targeting that area's real Planner keywords. Claude writes the local profile (neighbourhoods, landmarks, transport, peak season) in one call per area, and coordinates come from OpenStreetMap, never from the model. Business details come from the Business info section above. Target location in the keyword tool: the CITY when the areas are its own districts (Dubai → Al Qusais, Bur Dubai). The STATE or region when the areas are separate towns (Arizona → Kingman, Flagstaff) — every keyword is scoped to that geo, so asking for "plumber kingman" inside Phoenix measured only people sitting in Phoenix and returned 40/mo instead of 880, and put Kingman on a Phoenix street 280km away.

How it works: For each topic we search its REAL primary keyword (from the plan), pull the top-5 ranking pages, and distil them into anonymised facts — brand names, phone numbers and marketing filler are stripped, only data-bearing lines (prices, ranges, standards, timeframes) survive. Those facts + the plan's keywords, PAA questions and must-mention entities go into BOTH the outline and every section, so the article beats the SERP instead of paraphrasing it. Output: E-E-A-T byline, FAQs, cross-linked articles and type-matched schema (HowTo / Article / FAQPage / ItemList + breadcrumbs). Set SERPAPI_API_KEY and SCRAPINGBEE_API_KEY in the builder's GitHub secrets — without them, articles still generate from the keyword + expertise.