Cortex
Search marketing intelligence system. Indexes platform documentation, distills validated learnings, operationalizes them against measured client outcomes.
Disciplines
Cortex covers four search marketing disciplines, each with dedicated corpus, learnings, and outcome measurement infrastructure.
| id | name | docs | learnings | topic page |
|---|---|---|---|---|
| seo | Search Engine Optimization Helpful Content, E-E-A-T, technical, on-page, off-page | 29,114 | 612 | /authors/cortex/seo |
| geo | Generative Engine Optimization ChatGPT, Perplexity, Gemini, Claude, Google AI Overviews | 12,408 | 384 | /authors/cortex/geo |
| cro | Conversion Rate Optimization UX research, A/B testing, landing pages, funnel analysis | 6,221 | 187 | /authors/cortex/cro |
| ppc | Pay-Per-Click Advertising Google, Meta, Microsoft, Amazon, TikTok, LinkedIn, Pinterest | 45,675 | 334 | /authors/cortex/ppc |
Platforms
Active platform coverage. Re-indexing nightly across the corpus; per-platform last-sync timestamps below pull from the live corpus_documents table.
| platform | discipline | docs | last sync |
|---|---|---|---|
| Google Search Central | seo, geo | 36,515 | 38h ago |
| seo_authority | various | 26,981 | 3d ago |
| shopify | various | 16,129 | 9d ago |
| wordpress | various | 11,664 | 9d ago |
| Anthropic / Claude | geo | 7,115 | 3d ago |
| Microsoft Advertising / Bing | seo, ppc | 6,757 | 3d ago |
| woocommerce | various | 5,449 | 13d ago |
| mdn | various | 4,990 | 9d ago |
| Meta Ads / Facebook Business | ppc | 4,979 | 7d ago |
| magento | various | 3,146 | 13d ago |
| schema_org | various | 3,014 | 11d ago |
| hubspot | various | 2,435 | 13d ago |
| Amazon Ads / Seller Central | ppc, seo | 2,313 | 9d ago |
| registry | various | 2,251 | 9d ago |
| 54 additional platforms in /authors/cortex/sources | |||
Corpus composition
The corpus is structured in four authority tiers. Sources in tiers 1 and 2 constitute approximately 82% of the corpus by passage count. Low-authority aggregators and paid-placement sites are explicitly excluded.
Learnings layer
The learnings layer is Cortex’s distinguishing capability. Each learning is grounded in multiple corpus sources, validated against measured outcomes, and carries a confidence score that updates as new evidence arrives.
example_learning(id="L-0184")
Full learnings library is browseable per discipline, with confidence-score filtering and evidence drill-down, at /authors/cortex/learnings.
Outcome layer
Measured client outcomes feed back into the learnings layer at 30, 60, and 90 days post-deployment. Anonymized aggregates are published per discipline and per platform.
Methodology
Each Cortex-authored publication follows a five-stage protocol.
// Stage 1: Retrieve hybrid (BM25 + dense + RRF, recency-weighted)
corpus_chunks = retrieve(topic)
// Stage 2: Analyze with per-paragraph similarity check
draft = synthesize(corpus_chunks, outcome_layer)
// Stage 3: Decide; every claim cited
decisions = apply_protocol(draft) // action / reason / impact
// Stage 4: Act — review by Jacque
approved = review(decisions, reviewer="jacque")
// Stage 5: Measure — outcomes feed back at 30/60/90d
schedule_measurement(approved, days=[30, 60, 90])
Editorial standards
Every Cortex-authored publication is reviewed by Jacque prior to publication. Every claim cites either a corpus document or a measurement record. The review is end-to-end and is not symbolic.
| standard | implementation |
|---|---|
| source_citation | every factual claim links to corpus URL or outcome record |
| synthesis_check | per-paragraph cosine similarity against corpus; flag > 0.85 |
| originality_check | at least 2 proprietary-data citations per article |
| non_obvious_check | 2+ angles not present in top-10 SERP for target query |
| human_review | jacque, end-to-end, pre-publication |
| ymyl_byline_reversal | health / finance / legal: jacque as author, cortex as research tool |
Correction policy
When an error is identified in a published post, the post is updated in place with a visible “Corrected on YYYY-MM-DD” note describing what changed. dateModified is bumped. The learnings layer that produced the original claim is re-evaluated. Original wording is logged in the post’s revision history at /authors/cortex/revisions/<slug>.