Keyword Research → Optimized Landing Pages: The Workflow

The workflow has four stages: understand what's ranking → build briefs → generate pages → automate iteration.

Stage 1: Understand What's Ranking (Competitive Intelligence First)

Before generating a landing page for a keyword, run a competitive intelligence pass on the pages that currently rank. From competitive intel prompt, a structured Opus 4.6 prompt can extract comprehensive competitor profiles in minutes using only public data:

• Proxy signals to pull: job postings (feature roadmap hints), review sites like G2/Capterra (customer language and pain points), conference talks (positioning language), and the competitor's own landing page copy
• What to extract per competitor: positioning, value props, CTA language, page structure, social proof types, objection handling
• Critical practice: explicitly flag confirmed facts vs. AI inference — the prompt template from competitive intel prompt builds this in

For content intelligence specifically, use the cf-crawl skill (from cf crawl scheduled knowledge base) to batch-crawl competitor landing pages into markdown in one shot:

npx claude-code-templates@latest --skill utilities/cf-crawl

This gives Claude direct access to competitor page content as context when generating your brief — not just a description of it.

Stage 2: Build the Content Brief (Plan-First)

From Workflow: unless it's literally a one-line change, there's always a plan.md first. For landing pages, the brief is the plan. Each keyword gets its own brief file:

plans/lp-[keyword-slug].md

What to include in the brief (assembled from Workflow and claude code seo workspace):

• Primary keyword + semantic variants
• Search intent classification (informational / transactional / navigational)
• Competitor positioning gaps (from Stage 1)
• Customer language mined from review sites
• H1 formula, H2 structure, target word count
• CTA goal and lead capture mechanic
• Internal linking targets
• Meta title and description formula

Once you have 10+ keyword briefs, run them in parallel sessions — while one session generates the landing page for keyword A, another drafts the page for keyword B (from Workflow). 4–6 parallel Ghostty sessions with bypass permissions enabled is the standard power-user setup.

Stage 3: Generate the Landing Page

The mega-prompt pattern (from vibe coding prompt opus): feed the brief + design constraints into a single comprehensive prompt. Opus 4.6 can generate complete, conversion-ready landing pages one-shot — hero sections, CTAs, and conversion-focused layouts are where it specifically excels (from opus marketing ui).

Design constraint stack (from Vibe Coding and Brand And Design):

1. Copy a layout template from Tailark, Tailwind UI, or shadcn blocks — paste it as a constraint in the prompt ("follow this exact layout structure")
2. Screenshot a reference landing page from landing.love or saaspo.com and include it as visual context
3. Reference your CLAUDE.md style guide (from claude code seo workspace) which encodes brand voice, design tokens, and component preferences

The principle: "give AI constraints and it becomes a weapon" — freestyling produces generic output, constrained generation produces professional output (from Brand And Design).

Copywriting quality: train a /landing-page skill on your highest-converting existing pages (from ai copywriting skill training). Claude learns copy structure and voice from examples rather than just instructions — each page invocation gets better as the skill accumulates reference material.

Stage 4: Automate Ongoing Iteration

Once pages are live, set up monitoring loops (from claude code loop business use cases):

• Competitor monitoring loop (every 20min): Research competitor landing page changes — "more powerful than RSS because you specify semantically what you care about"
• Scheduled cf-crawl (daily): Automatically re-crawl competitor pages and save updated markdown — fresh competitive context without manual work (from cf crawl scheduled knowledge base)

Run /web-setup to grant scheduled tasks access to repos — these run on cloud infrastructure with no local machine required (from Workflow).

Gaps to Flag

Knowledge Engine has no sources on:

• Keyword research tooling (how to generate the initial keyword list)
• SERP analysis (click-through rates, featured snippet targeting, PAA boxes)
• Technical on-page SEO (schema markup, Core Web Vitals, canonical tags)
• Link building or domain authority

The workflow above assumes you arrive with a validated keyword list. The KE coverage is strongest on the generation, design, and automation stages.

Sources cited:

• competitive intel prompt — Structured competitive intelligence to understand what's ranking and why
• cf crawl scheduled knowledge base — cf-crawl skill for batch-crawling competitor pages to markdown
• claude code seo workspace — Dedicated SEO workspace pattern with CLAUDE.md, skills, and tooling
• Workflow — Plan-first brief creation, parallel sessions for batch page generation, scheduled tasks
• vibe coding prompt opus — Mega-prompt pattern: bundle brief + design constraints into single prompt
• opus marketing ui — Opus 4.6 capability for hero sections, CTAs, conversion-focused layouts
• Vibe Coding — Design constraint technique: layout templates prevent AI slop
• Brand And Design — Constraint principle, Dribbble references, component inspiration sources
• ai copywriting skill training — Training landing page skills on high-converting examples
• claude code loop business use cases — Competitor monitoring loops for ongoing iteration