SEO often appears as a giant directory of keywords, titles, content, crawling, links, hreflang, AI search and tools. A directory shows what exists, not what comes first. An operational system needs the job, dependencies and acceptance evidence for each module. Without them, teams start dozens of tasks and cannot explain which change created value.
Layer one: strategy decides which pages should exist
Strategy covers business objective, market, audience, demand, competing pages, intent, topic map and roadmap. Keyword research should produce a query/problem-to-intent-to-page-type-to-URL-to-conversion map, not a list. SERPs, Search Console, sales questions, support records and communities provide demand evidence; third-party volume remains a dated estimate.
Layer two: on-page foundations communicate meaning
- Title and H1: describe the page task without keyword-swapped duplicates
- Meta description: help click decisions without assuming it will always display
- Headings: organise questions, steps and decisions rather than keyword repetition
- URL: stable, readable and maintainable, with mappings for migrations
- Internal links: connect services, evidence, insights and next actions
- Images: accurate alt text, dimensions and provenance rather than stuffing
Layer three: content fulfils different decision jobs
Build topic clusters around user jobs: explanation, diagnosis, tutorial, comparison, alternatives, procurement, evidence and service pages. Pillars map the field; supporting pages complete specific tasks and link coherently. Evergreen pages still need fact maintenance. Programmatic SEO should scale only when every page has distinct data, a real task and quality controls.
Layer four: optimisation is not keyword density
Optimisation begins with intent and page type, then entities, terminology, conditions, evidence, examples, FAQs and decision steps. NLP or content scores suggest gaps but do not replace editorial judgement. CTR work should not use promises the page cannot fulfil. For weak pages, choose keep, improve, consolidate, reposition or remove before editing phrases.
Layer five: technical SEO delivers the content
- Crawl and status: priority pages are accessible with correct responses
- Index eligibility: visible, distinct content without accidental noindex
- Architecture: priority pages are reachable through navigation and clusters
- Sitemap: preferred URLs only, with meaningful lastmod
- Robots: crawl management, not an index-removal mechanism
- Canonical: signal consolidation, not a substitute for redirects and status
- Schema: matches visible facts without fabricated entities or reviews
- Performance: separate lab diagnostics from field data
Layer six: authority comes from evidence and relationships
Off-page SEO is not link purchasing. Durable assets include original research, tools, data, expert commentary, permissioned cases, resource pages, digital PR and organic mentions. Guest posts, resource links and broken-link outreach only add value when relevant, editorial and useful. Backlink audits must distinguish spam noise, normal mentions and real risk rather than automatically disavowing every weak domain.
Layer seven: business model changes the page system
SaaS often needs feature, use-case, integration, alternative, comparison, free-tool and BOFU procurement pages; ecommerce relies on categories, filters, products, inventory and structured data; local businesses need location and service proof. Templates cannot merely swap names—every URL needs a distinct job and evidence.
Layer eight: international SEO is localisation
International SEO covers language and market URLs, hreflang, canonicals, translation and local review, intent differences, local proof, currency, regulation and conversion. Redirects should not block access to alternate versions. An English page is not a translated Chinese keyword list; market demand must be revalidated.
Layer nine: AI search builds on SEO foundations
AI Overviews, AI Mode and ChatGPT add prompt sets, brand facts, entity relationships, citable passages, source evidence and answer monitoring; they do not remove crawl, index, content and authority foundations. Record prompts, answers, citations and dates per platform, separating crawler access, indexing, citation and recommendation. Structured data cannot guarantee inclusion.
Layer ten: analytics connects SEO to growth
- Search Console: query, page, impression, click and index evidence
- Analytics: source, landing page, content path and key events
- Server logs: crawl requests without automatically trusting user agents
- Rank tracking: fixed market, device and query trends
- CRM: qualified enquiries, opportunities and sales feedback
- Reporting: separate implementation, platform processing, search and business outcomes
Layer eleven: audits need decisions
Technical, content, link, competitor and indexing audits should record coverage, evidence, impact, confidence, owner, dependency, fix and verification. The core output is a decision ledger: keep, improve, consolidate, reposition or remove. An audit without implementation ownership rarely changes outcomes.
Layer twelve: growth comes from verifiable experiments
After foundation issues close, test titles, links, structures, templates, tools and conversion paths. Every experiment needs a hypothesis, unit, window, guardrails and stop rule. Automation helps collection, QA and workflow; it should not mass-publish indexable pages without review. An SEO win is a reproducible improvement, not a short-term screenshot.
A practical order of operations
- 1. Define objective, market, audience and conversion
- 2. Map queries/problems to pages and services
- 3. Fix crawl, index, canonical and localisation foundations
- 4. Improve service, evidence and topic clusters
- 5. Build legitimate third-party authority and mentions
- 6. Connect search consoles, analytics and CRM
- 7. Update, consolidate and experiment from evidence
- 8. Add GEO prompt and answer monitoring to the same system
Should SEO start with technical work or content?
Fix blockers to access, crawling, indexing or conversion first while maintaining essential content. Scale clusters after the foundation is sound.
Do more SEO tools produce better results?
No. Every tool needs a decision role, stable definitions, ownership and workflow. Duplicate tools add cost and conflict.
Does AI SEO require a separate website?
Usually not. AI discovery still relies on accessible content, entity facts, evidence and search foundations. Extend the existing system with prompt structure, sources and monitoring.
How do we choose between updating and consolidating a page?
Assess whether it completes a distinct task, overlaps in query and intent, and retains links or business value. Consolidation needs a survivor URL, content migration, redirect and verification plan.
Source:Google Search Central: SEO Starter Guide
Source:Google Search Central: Technical requirements
