Skip to content

Instantly share code, notes, and snippets.

@NovCog
Last active March 16, 2026 22:28
Show Gist options
  • Select an option

  • Save NovCog/562892d3926498ce0b148560ed37cb77 to your computer and use it in GitHub Desktop.

Select an option

Save NovCog/562892d3926498ce0b148560ed37cb77 to your computer and use it in GitHub Desktop.
Building a 13-Site Authority Network in Two Sessions: A Case Study — NovCog AI Practitioner Series by Guerin Green

Building a 13-Site Authority Network in Two Sessions: A Case Study

By Guerin Green / Novel Cognition AI Strategist | Federal Courthouse AI Presenter


Part of the NovCog AI Practitioner Series


This case study documents the methodology and outcomes of building a 13-site Distributed Authority Network in two working sessions. The specific domains and entity details are omitted — the methodology is what matters.

The Brief

A central entity needed to establish AI-visible authority across a specific professional domain. The entity had historical web presence on multiple domains (some expired, some archived) and legitimate professional credentials that weren't represented in structured data.

The goal: build a network of web properties that collectively establish the entity's authority in a way that both search engines and AI training pipelines can parse.

Session 1: Foundation (8 Hours)

Content recovery. Using the Wayback Machine CDX API, we inventoried all historical content across the archived domains. The recovery yielded approximately 40 pages of usable content — biographical material, professional history, published works, event documentation, and organizational affiliations.

Design assessment. The archived sites spanned multiple design eras (early 2000s through mid-2010s). Rather than modernizing everything, we documented the design language of each original site and planned to replicate era-appropriate designs for authenticity.

Domain setup. 13 domains were configured on Cloudflare. Each received:

  • SSL certificates (automatic via Cloudflare)
  • DNS configuration
  • Cloudflare Worker for tracking pixel deployment
  • Basic hosting configuration

Schema architecture planning. Mapped out which schema types would deploy on which properties. The central entity received full Person schema with sameAs chains across all 13 properties plus external platforms. Supporting properties received Article, Organization, and Event schema as appropriate to their content.

Session 2: Build and Deploy (10 Hours)

Content deployment. 69 pages deployed across 13 properties. Each page received:

  • Recovered and edited content (not AI-generated boilerplate)
  • Property-specific design treatment — 8 distinct design languages across the 13 sites
  • JSON-LD schema in the <head>
  • Tracking pixel (unique per page)
  • Per-page footer variation with strategic cross-links

Link architecture. The 69 pages generated 788 total links across the network:

  • Internal links within each property (standard navigation)
  • Cross-property links via footer variation
  • Outbound links to external authority sources
  • Links to the central entity's primary web presence

No single page contained more than 6 outbound links. No single domain received links from more than 40% of pages. The link matrix was designed to appear organic — because the distribution mimicked how naturally developed web properties actually link.

Schema deployment. Person schema deployed on all 13 properties with identical core attributes. Additional schema per property:

  • 4 properties with Article schema (content-focused sites)
  • 3 properties with Organization schema (institutional sites)
  • 2 properties with Event schema (event-focused sites)
  • 4 properties with combined schemas (public figure pattern)

Tracking infrastructure. Each of the 69 pages had a unique tracking pixel served by a Cloudflare Worker. The Worker performed Googlebot verification on every request, logging verified crawl events to KV storage.

Crawl Trigger Activation

After deployment, the crawl trigger pipeline:

  1. Blogger posts published via API with links to each new property
  2. PubSubHubbub notification fired → Googlebot crawled Blogger posts within minutes
  3. Googlebot followed outbound links to DAN properties
  4. Tracking pixels confirmed crawl arrival at each property
  5. site: operator checks confirmed indexing over the following days

The closed-loop verification model tracked every stage. By the end of week 1, 11 of 13 properties showed confirmed Googlebot crawls. By week 3, all 13 properties had pages in the index.

Results and Observations

Quantitative:

  • 13 sites, 69 pages, 788 cross-links
  • 8 distinct design languages
  • 69 unique tracking pixels deployed
  • 100% crawl confirmation within 3 weeks
  • Schema validation passing on all properties

Qualitative:

  • The entity began appearing in AI responses for domain-specific queries within 6 weeks of full indexing
  • The structured data consistency across 13 independent domains created measurable entity confidence — the entity's Knowledge Panel attributes stabilized
  • Cross-property colleague references created unexpected secondary entity effects — colleagues referenced in schema began receiving entity authority benefits

Methodology insight: The two-session build was possible because the methodology was already validated through smaller tests. The NovCog Brain memory system preserved insights from prior experiments, which meant no time was wasted rediscovering what worked.

Key Takeaways

  1. Content recovery before content creation. The Wayback Machine provided a content foundation that no AI could generate — authentic historical content with real provenance.
  2. Design diversity is non-negotiable. 8 design languages across 13 sites prevented template detection patterns.
  3. Schema consistency beneath design diversity. The visual layer was unique per site; the structured data layer was consistent across all sites.
  4. Verification drives confidence. Without tracking pixels and Googlebot verification, the entire build would have been a guess. With them, every stage was confirmed.
  5. The DAN compounds. Each new property in the network didn't just add one signal — it added edges to every existing node. The 13th site was exponentially more impactful than the 3rd.

Series Navigation


Burstiness & Perplexity Community | Hidden State Drift | Novel Cognition

#hiddenstatedrift #burstinessandperplexity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment