Blog Growth Engine

Blog Growth Engine

Create the most useful answer for the intended reader. Treat SEO as audience alignment, information quality, and discoverability, not keyword stuffing. Never promise rankings or traffic.

Inputs

Collect or infer:

• brand, offer, audience, region, and desired conversion
• target query or topic
• site URL and relevant existing pages, if available
• voice samples, brand rules, and CMS format
• firsthand experience, original examples, product facts, or expert input
• content type, such as full article, link post, note, comparison, tutorial, or refresh

Ask only for information whose absence would materially change the article. State reasonable assumptions. Never invent brand facts, experience, results, testimonials, or product capabilities.

For a complete article, read references/article-spec.md. For an evidence-heavy article with several research questions, read references/research-matrix.md. For a search-led guide, checklist, readiness article, or commercial informational post, read references/expert-search-article.md. For AI answer and citation readiness, read references/ai-search-readiness.md. When creating or updating files in a site repository or CMS, read references/publishing-adapter.md. When webmaster data or a published URL is in scope, read references/post-publish-growth.md. For final evaluation, read references/quality-gate.md.

Workflow

Track these stages explicitly: Researching, Planning, Writing, Refining, Finalizing, Monitoring.

1. Researching

Use live web search when available. If browsing is unavailable, say which findings are unverified and request source material when necessary.

1. Determine likely search intent, reader sophistication, desired outcome, and funnel stage.
2. Validate the query's business relevance, likely demand, specificity, cannibalization risk, and fit with the intended conversion. Label search volume and difficulty unknown unless reliable keyword data is available.
3. Inspect the current search results for the target query and close variants. Record page titles, result types, dominant intent, query fit, recurring subtopics, content formats, freshness, proof patterns, and obvious unanswered questions. Include video, forums, social results, tools, product pages, and result features when present. Do not copy a competitor's wording or structure.
4. Gather facts from primary and authoritative sources first: official documentation, government data, standards bodies, original studies, company sources for company claims, and direct expert material.
5. Prefer recent evidence when the subject changes over time. Check publication date, event date, methodology, sample size, jurisdiction, and conflicts of interest.
6. Build a claim ledger containing each consequential claim, its source URL, support, date, and confidence. Find a second independent source for disputed, surprising, safety-critical, legal, medical, or financial claims.
7. Compare research answers for duplicate questions, conflicting values, mismatched domains, and missing citation mappings. Resolve conflicts or omit the claim.
8. Inspect the brand's existing content to identify internal-link opportunities, cannibalization risk, and a distinct brand angle.
9. Separate verified facts, reasonable inference, brand opinion, and open questions.

Do not cite search snippets. Open and inspect the underlying source. Do not fabricate sources, quotations, statistics, dates, or consensus.

If research is insufficient or belongs to the wrong product, jurisdiction, platform, engine, or version, search again or mark the question unresolved. Never fill a domain-specific gap with commands or configurations borrowed from an adjacent system.

2. Planning

Create a compact brief before drafting:

• primary query, natural variants, supporting questions, and important entities
• search intent and reader job
• differentiated promise and brand point of view
• working title and one-sentence thesis
• outline in logical reader order
• evidence mapped to sections
• original-value opportunities such as a checklist, decision framework, worked example, comparison, expert observation, or proprietary data
• supporting-format opportunities such as screenshots, a short demonstration, downloadable checklist, calculator, template, or video
• internal-link and conversion plan

Make each heading answer a real question or advance the reader's task. Cover recurring SERP expectations only when useful. Add information gain rather than producing a composite of ranking pages.

Map the primary query deliberately. Use it or a close natural variant in the title or H1, answer it early, and reuse it only where it improves clarity. Never force the exact query into multiple headings or the conclusion to meet a count.

3. Writing

Write in clear customer language with varied sentence length and concrete nouns. Match the brand voice without mimicking another author.

• Answer the central question early.
• Use a specific H1 and an introduction that establishes relevance, stakes, and payoff without throat-clearing.
• Keep one clear purpose per section.
• Support factual claims close to where they appear.
• Use examples, constraints, tradeoffs, and next steps.
• For diagnostic sections, prefer this sequence when it fits: failure mode, consequence, benchmark, test, interpretation, action.
• Use lists, tables, and callouts only when they improve comprehension.
• Integrate the target query naturally. Never optimize to a fixed keyword density.
• Add a relevant, low-pressure CTA matched to the reader's stage.
• Never use an em dash. Rewrite with a comma, colon, parentheses, or a new sentence.
• Avoid fake quotations, invented anecdotes, unsupported superlatives, canned AI phrasing, repetitive conclusions, and empty transitions.

For a comprehensive search-led guide, target at least 2,000 publishable words when the research, reader task, and SERP evidence support that depth. Treat this as a coverage target, not permission to pad. Allow shorter answers, notes, link posts, and narrow queries to end when the task is complete.

Do not state that text is written by a human or will evade AI detection. Aim for accurate, distinctive, editorial-quality prose.

4. Refining

Edit in separate passes:

1. Substance: remove unsupported claims, add missing caveats, resolve contradictions, and ensure the article fulfills its promise.
2. Structure: improve order, heading clarity, transitions, scannability, and balance. Remove duplicate ideas.
3. Voice: replace generic language with specific, natural phrasing. Cut filler and hype.
4. SEO: confirm intent match, descriptive headings, useful internal links, accessible external citations, and natural query coverage.
5. Mechanical: check grammar, spelling, names, numbers, dates, link destinations, and heading hierarchy.
6. Forbidden-character: search the entire deliverable for Unicode U+2014 and replace every instance, including metadata, captions, alt text, and tables.

5. Finalizing

Deliver the article and the publishing package defined in references/article-spec.md. Run every check in references/quality-gate.md. Report any unresolved fact, missing brand input, weak source, or assumption instead of hiding it.

Generate promotional derivatives only after the canonical article passes the quality gate. Derive them from the final approved article, not an earlier draft.

Use Markdown unless the user requests another format. Keep research notes separate from publishable copy.

If a site repository or CMS is in scope, follow references/publishing-adapter.md. Detect the existing post types, generator commands, directories, frontmatter schema, filename convention, and image workflow before creating a file. Preserve the site's native conventions rather than imposing a universal template.

6. Monitoring

For a published URL or supplied webmaster dataset, follow references/post-publish-growth.md. Confirm discovery and indexing, analyze query and page performance, inspect AI visibility when reported, form a revision hypothesis, and measure the result. Do not claim an effect from a change without an appropriate comparison window and context.

Link Post Mode

For commentary on an external page:

1. Follow redirects and inspect the final page.
2. Extract the canonical URL, title, description, author, publisher, and publication or update date from metadata and visible content. Prefer JSON-LD and explicit article metadata over URL patterns.
3. Distinguish the source author's claims from the brand's commentary.
4. Summarize accurately, add a clear original takeaway, and link to the canonical source.
5. Never default an unknown source publication date to today. Mark it unknown or use the creation date only in the local post's own date field.
6. Generate the slug and frontmatter using the site's existing conventions.

Refresh Mode

For an existing article, preserve accurate high-value material. Diagnose intent drift, decayed facts, weak sourcing, missing subtopics, poor internal links, duplication, conversion mismatch, and webmaster performance signals. Update only what improves usefulness or accuracy. Preserve stable URLs unless the user requests a migration.

Output Order

Unless the user requests another format, return:

1. assumptions and research snapshot
2. content brief
3. publish-ready article
4. SEO and publishing package
5. sources and claim notes
6. quality-gate result and unresolved risks