Content quality & EEAT rewrites: humanized rewrites across 40+ pages

[psyduckler]

Summary

Resolves the content-quality and EEAT audit findings with humanized, practitioner-voiced rewrites across the entire site. Roughly +1,300 lines of net new content, replacing templated boilerplate with first-person observations drawn from Bernard's Clearscope experience.

TypeScript clean. 148/148 vitest tests passing.

What changed

Use-case pages (30 pages)

• Fixed critical curl-example bug: image and audio use-case pages were rendering text-payload boilerplate (type:"text", content:"Paste content here") instead of the modality-specific request body. Now correctly uses the modality's actual sample.
• All 30 entries rewritten with genuine domain depth (KDP narration-AI policy, eBay/Etsy seller-verification flows, ACORD/SIU references for insurance claims, Reddit API changes since 2023, FTC testimonial rules, etc.)
• Extracted to a dedicated src/useCases.ts module for maintainability
• New UseCase schema fields: whatWeveSeen (Bernard's practitioner observation), domainNuance (vertical-specific concepts), realExample (concrete setup/result)

Text-detection API cluster (6 pages)

The audit's biggest red flag — six near-duplicate URLs with identical structures. Each now has a distinct primary job:

• /ai-detection-api: category overview ("authorship-likelihood vs workflow-routing")
• /ai-content-detector-api: CMS pre-publish gate (WordPress save_post hook example)
• /ai-written-content-detection: RAG/training-data hygiene (batch curation pattern)
• /ai-generated-content-detection: UGC moderation triage (user-trend aggregation)
• /ai-written-content-detector: agent-builder integration (bounded revise loop)
• /ai-generated-text-detector: auto_revise deep-dive (rescore pattern)

Each gets distinct code samples, distinct FAQs, distinct cost-modeling sections, and a distinct codeLabel so they no longer all share the "Copy-paste routing example" heading.

Comparison + alternatives pages (10 pages)

• ComparisonPage schema extended with practitionerNote, whereCompetitorWins, whereVeracityWins
• All 4 /vs/* pages rewritten with first-person observations on the category split between authorship-likelihood detectors (GPTZero, Originality.ai, Copyleaks) and workflow-routing APIs
• All 6 /alternatives/* pages rewritten with honest "it depends on..." framing instead of templated alternative-positioning copy
• Each comparison explicitly names 3–4 dimensions where the competitor genuinely outperforms VeracityAPI

Integration pages (real framework code)

• /integrations/langgraph: full StateGraph implementation with bounded revise loop (prevents the infinite-loop bug Bernard observed costing teams hundreds of dollars on a single document)
• /integrations/claude: Anthropic SDK direct tool_use pattern alongside MCP (most teams need both)
• /integrations/openai-actions: Custom GPT system-prompt fragment with reliability patterns

Methodology / How-it-works split

• /methodology rewritten as the epistemology page: what VeracityAPI claims, what it explicitly doesn't, the trust contract by field, policy table by intended_use, and an explicit "when the right answer is don't use VeracityAPI" section
• /how-it-works rewritten as the technical pipeline page: model, temperature, schema, pipeline steps, threshold table with actual numbers, version changelog, v0.2 roadmap
• Two pages now answer distinct questions for distinct readers per the audit recommendation

Blog (4 new substantive posts + EEAT scaffolding)

• "What 10 years at Clearscope taught me about AI slop"
• "The most expensive bug in agent content workflows"
• "Why the AI-detection category is splitting"
• "Why we don't publish competitor benchmarks (yet)"
• BlogPost interface extended with author bio, reading time, word count
• Blog post pages now include byline, "About the author" card, and JSON-LD schema with Person author
• Legacy posts retained at original URLs with author backfill

Visual hooks

• Small CSS additions for .practitioner-note (hot-pink-bordered to distinguish from regular cards), .byline, .author-bio

Test plan

• npx tsc --noEmit clean
• npx vitest run — 148/148 passing (one test updated to reflect new blog-post count)
• Manual: deploy to a preview environment and click through a sampling of pages (use-case pages, text-detection cluster, comparison pages, the two blog posts with Bernard byline) to confirm visual rendering and the practitioner-note styling
• Manual: confirm the curl-example fix on an image and audio use-case page renders the correct payload
• SEO: confirm sitemap still lists expected URLs (the test asserts presence; production crawl will confirm)

Files changed

• src/useCases.ts — new file, 30 deeply rewritten use-case entries
• src/pages.ts — extracted USE_CASES, rewrote methodology + how-it-works, upgraded blog and comparison renderers
• src/distribution.ts — rewrote 6 text-detection pages, 6 alternatives pages, all integrations, all media-specific API pages
• src/comparisons.ts — extended schema + rewrote all 4 vs pages
• src/blog.ts — extended schema + 4 new posts + author backfill on legacy posts
• src/y2k.ts — small CSS additions for new EEAT components
• test/benchmarkComparisonPages.test.ts — relax blog-post count assertion

🤖 Generated with Claude Code