Use this when
rewrite blog, optimize blog, update blog, improve blog, fix blog.
Process
Audit first, read-only. Detect the format, then check: fabricated vs. sourced statistics; answer-first formatting (does each H2 open with its point?); image and chart count and type diversity; paragraph pacing in context (descriptive, not a fixed quota); heading hierarchy (H1 -> H2 -> H3, no skips); schema presence and validity (Article/BlogPosting + Person + Organization + BreadcrumbList priority); freshness signals (lastUpdated, dateModified); self-promotion level; citation tier quality. Separately scan for generic AI-sounding phrases and structural repetition (repeated question-cadence H2s, "Here" paragraph openers, hedge-word stacking) as advisory notes that never determine authorship. Check for a keyword-cannibalization conflict against the rest of the blog and recommend merge or differentiate if found. Calculate the current 5-category score (Content 30 / SEO 25 / E-E-A-T 15 / Technical 15 / AI Citation 15) and present the audit for approval before changing anything.
Once approved, research replacement statistics for anything fabricated or unsourced (tier 1-3 sources only), and plan any missing images or charts.
Rewrite in this order: preserve the author's voice, unique insights, and first-hand experience; fix the meta description and frontmatter (lastUpdated only when facts, methods, or recommendations materially changed; keep the original date); apply purpose-first formatting to weak sections; replace every fabricated statistic with sourced, verifiable data and enough provenance to check it; fix heading forms to match reader intent; adjust paragraph length only where it improves comprehension; add visuals spaced evenly; add or improve an FAQ only where warranted; cap self-promotion at one mention; strengthen evidence-backed explanations for important reusable claims without padding every section.
Voice pass: eliminate every em dash (comma, hyphen, colon, or period instead); replace flagged AI phrases with natural alternatives ("leverage" -> "use", "delve" -> "look at", "robust" -> "strong"); vary sentence length so no more than 3 consecutive sentences land within 5 words of each other; use contractions where natural; keep first-hand language only when it's genuinely substantiated.
Verify before presenting: zero fabricated statistics, clean heading hierarchy, priority schema present, internal-link zones or real links present (5-10 per 2,000 words), and the score improved across all 5 categories versus the Phase 1 audit — a lower score after a rewrite is itself a problem, not an acceptable outcome. Report before/after scores per category and a list of what changed.