How to build per-location internal link recommendation engine

The 4-skill bundle on the multi-location-seo-architecture agent

One agent. Four coordinated skills. The Recommend + Personalize + Gate + Audit bundle consumes the canonical per-URL link-equity record from #575 + topical-embedding substrate + vector-similarity + CMS substrate and writes one canonical per-recommendation record.

Compliance overlay

Five anchors run per-recommendation before any insertion commits. The first anchor is operationally distinctive: Google PageRank patent + per-API license intersect CFAA-scraping doctrine + Lanham trademark surface when anchor text uses competitor brand.

FAQ

What is per-location internal link recommendation engine — and what is the AI-ML-recommendation times PageRank-patent times CFAA-license problem distinctive to this skill?

A multi-location retail operator with 80-300 stores generates a continuous stream of new + edited + republished pages. Internal links need to be recommended per page based on link-equity-distribution (from #575 sibling skill on same agent) + topical relevance + revenue impact + per-vertical regulatory constraints. The four-skill bundle on the multi-location-seo-architecture agent — Recommend, Personalize, Gate, Audit — consumes the canonical per-URL link-equity-record from #575 and writes a per-recommendation canonical record. The operationally distinctive anchor: Google PageRank patent + Topical PageRank + Reasonable Surfer Model + Hilltop algorithm + per-API license terms intersect CFAA-scraping doctrine + crawl-budget honoring + Lanham Act trademark surface (when recommendation anchor text uses competitor brand) + EU AI Act Article 22 + Annex III when AI-ML recommendation drives content prioritization. Per-vertical (HIPAA + FINRA + state bar + ABA Model Rule 7.1-7.5) gates per-recommendation. WCAG 2.2 AA + ADA Title III + Section 508 + EAA when recommendation rendering must remain accessible.

Why do Sitebulb + Screaming Frog + OnCrawl + Ahrefs + SEMrush + Moz Pro + CMS-native plugins break at multi-location AI-ML-recommendation scale?

Each crawler exposes per-account flat link-graph extraction. Each SEO platform exposes flat ranking signals. Each CMS-native plugin offers rule-based or shallow ML recommendations without per-vertical regulatory gating. None consumes a per-URL link-equity canonical from #575 sibling skill (Internal-PageRank + Topical-PageRank + Reasonable-Surfer + Hilltop + Block Level PR + revenue-attribution). None personalizes per-portfolio per-banner per-location per-page topical embedding via sentence-transformers + Cohere Embed + OpenAI Embeddings + Anthropic Voyage + BGE + GTE + vector-similarity (Pinecone + Weaviate + Chroma + Qdrant + Milvus + Vespa + Marqo). None gates per-recommendation against per-vertical regulatory anchors (HIPAA when MedicalBusiness + FINRA when FinancialService + state bar when LegalService + ABA Model Rule 7.1-7.5 + state medical board) + Lanham trademark surface (USPTO 15 USC 1125(a) when anchor text repeats competitor brand) + EU AI Act Article 22 + Annex III. None writes a per-recommendation audit trail with regulatory-defense retention. The four-skill bundle Recommend + Personalize + Gate + Audit sits above the crawler + SEO-platform + CMS-plugin substrate — it does not replace it.

How does Recommend + Personalize work?

Recommend runs per-portfolio per-page recommendation generation via per-URL link-equity canonical from #575 + topical-embedding (sentence-transformers + Cohere Embed + OpenAI Embeddings + Anthropic Voyage + BGE + GTE) + vector-similarity retrieval (Pinecone + Weaviate + Chroma + Qdrant + Milvus + Vespa + Marqo) + BM25 + Lucene + Elasticsearch + OpenSearch + Vespa retrieval + per-page revenue-attribution from GA4 + Adobe + Mixpanel + Amplitude + Heap. Per-page candidate-set: top-50 internal-link candidates ranked by composite score (Internal-PageRank + Topical-PageRank + Reasonable-Surfer + Hilltop + topical similarity + revenue-uplift estimate). Personalize runs per-portfolio per-banner per-location per-page customization: per-location service area + per-location vertical + per-location regulatory applicability + per-location language + per-location locale + per-location AI Overview presence + per-location SERP-feature presence (from #571 sibling). Per-recommendation anchor-text generation via per-vendor LLM (OpenAI + Anthropic + Mistral) under zero-retention with brand-voice canonical from voice-attribute extraction (#549 sibling). Per-recommendation per-platform (web + mobile + AMP + AI Overview snippet) format.

What does Gate + Audit do?

Gate runs 5 anchors per-recommendation before any insertion commits. (1) Google PageRank patent + per-API license terms (Ahrefs + SEMrush + Moz Mozscape + Sistrix + Screaming Frog + Botify + OnCrawl + Sitebulb) + CFAA 18 USC 1030 + hiQ Labs/Van Buren/Meta v Bright Data + robots.txt + Google Crawl Budget + Sitemap protocol. (2) USPTO trademark monitoring + Lanham Act 15 USC 1051 + 15 USC 1125(a) false-designation when recommendation anchor text uses competitor brand + FTC Endorsement Guides 16 CFR Part 255 + FTC Act Section 5 + Pfizer 1972 + FTC MARS + Robinson-Patman + FDD Item 12 + 15-state franchise registration + state UDTPA + Lanham. (3) HIPAA 45 CFR 164.502 + 504 + 514 when MedicalBusiness recommendation includes patient-relationship signal + state medical board + ABA Model Rule 7.1-7.5 when LegalService + state bar 50-state + FINRA Rule 2210 when FinancialService + SEC Regulation FD + FDA OPDP when prescription + DEA Schedule + alcohol TABC + CalABC + state-board. (4) EU AI Act Article 22 + Article 26 + Article 50 + Article 13 + 14 + 15 + Annex III high-risk when AI-ML recommendation drives content prioritization + Article 6 + 27 FRIA + Digital Services Act + DMA + GDPR Article 22 + Article 6 + 7 + 28 + 30 + LGPD + DPDP + PIPEDA + Quebec Law 25 + CCPA + CPRA + COPPA + 18-state privacy. (5) WCAG 2.2 AA + ADA Title III + Section 508 + EAA EN 301 549 + SEC Reg S-K material disclosure + FASB ASC 350 intangible-asset. Audit writes a per-recommendation WORM canonical record: per-URL link-equity input snapshot + topical-embedding + vector-similarity score + revenue-uplift estimate + per-location personalization + per-anchor gate-pass + Lanham trademark surface + per-vertical applicability + accessibility WCAG 2.2 AA + per-vendor license + CFAA attestation + AI-ML provenance + EU AI Act transparency label. Storage: AWS S3 Object Lock + Azure Blob immutable + Google Cloud Storage Bucket Lock + Wasabi WORM. Retention: 7-year FTC + 7-year IRS + 7-year HIPAA + 7-year state bar + 6-year SEC + 3-year FINRA + GDPR Article 30 + EU AI Act Article 12 + SOC 2 CC7/CC8.

What does this skill connect to on the multi-location-seo-architecture agent and across the swarm?

On the multi-location-seo-architecture agent: multi-location-seo-architecture + internal link equity distribution analysis (#575 UPSTREAM canonical link-equity record) + url-hierarchy-authoring + canonical-tag-management + orphan-page-detection + title-rewrite-tracking + ai-overview-tracking + seo-alerts. Across the swarm: schema-audit-remediation agent (rich-result eligibility #563 + JSON-LD #549 + continuous schema audit) + local-pack-rank-tracking (#559 + #567 + #571 cluster) + voice-attribute extraction (#549 brand-voice canonical) + governance-decision-router five-destination routing + master-record + per-jurisdiction compliance multi-state franchise. Build-pillar siblings: tiered pre-filter deterministic gates for AI content compliance + marketing AI autonomy profile configuration + per-location per-cohort two-sigma anomaly detection + per-location dynamic content for multi-location communications. Commercial-pillar parent: multi-location-seo-architecture.

What does the 6-workstream pre-engagement-baseline reporting cycle look like for this skill?

Every two weeks during the Tier 3 Fractional CMO with AI Swarm engagement, six workstreams report against the pre-engagement baseline. Workstream 1: per-portfolio recommendation coverage — pages with recommendations enabled + per-page recommendation volume + per-page link-equity canonical from #575 freshness. Workstream 2: Recommend candidate-set distribution — per-page top-50 candidate distribution + composite-score distribution + topical-embedding model coverage. Workstream 3: Personalize per-location customization flow — per-location service area applied + per-location vertical applied + per-location regulatory applicability + per-location AI Overview signal from #571. Workstream 4: Gate-pass/gate-fail distribution — per-anchor gate-fail rate + Lanham trademark surface coverage + per-vertical applicability + accessibility WCAG 2.2 AA. Workstream 5: Regulatory-defense audit coverage — PageRank patent + per-API license + CFAA + Lanham + per-vertical HIPAA/FINRA/state bar + EU AI Act Article 22 + Annex III + DSA + WCAG 2.2 AA. Workstream 6: FBC feedback-loop pattern-learning — per-page realized-vs-predicted link-equity uplift + per-recommendation acceptance rate + per-vertical near-miss + per-Google-update impact retrospective.

Related reading

• multi-location SEO architecture
• Sibling build-pillar: internal link equity distribution analysis (#575 UPSTREAM canonical record)
• Sibling build-pillar: per-location SERP feature-presence (#571 SERP-feature signal)
• Sibling build-pillar: voice-attribute extraction (#549 brand-voice canonical)
• Fractional CMO with AI Swarm
• AI Readiness Assessment