Turning Content Into a Search System

Overview

This work treats website content as an interconnected system of topics, products, pages, search intent, internal links, and evidence, rather than as a pile of individual articles. It runs on a working content-intelligence system, not a one-time audit: performance data pulls in automatically, consolidation and internal-linking opportunities get scored, and every recommendation goes through human review before anything changes.

Problem

A large content library accumulates overlap: duplicate topics, outdated articles, inconsistent taxonomy, weak internal linking, irrelevant outbound links, orphaned pages, and gaps between content and the products it should support. Each post gets optimized in isolation, which does not fix the structure, and without real performance data behind a decision, “this looks redundant” is just a guess.

My role

I designed and built the system: the data layer that merges organic performance signals, the topic taxonomy and product mappings, the consolidation and internal-linking logic, and the review workflow that governs every recommendation before it ships. I evaluate posts for consolidation, refresh, redirect, internal-link updates, product alignment, and canonical ownership, and I own the reporting layer that ties internal links, orphans, canonical targets, and performance together in one place.

Approach

Map the library as a system first. Define a taxonomy and the umbrella relationships between topics. Then merge real organic performance, Google Analytics, Search Console, and third-party keyword data, into that structure automatically, channel-segmented so paid or email traffic never gets mistaken for organic authority. Decide consolidation, redirect, and canonical ownership at the cluster level, using that performance data over word count whenever it’s available. Connect informational content to the relevant product pages. Audit outbound links for relevance. Route every recommendation through a review step, approved, rejected, or held, and feed rejected reasons back in so the next pass improves instead of repeating the same miss.

What I built and did

• An automated data layer that merges Google Analytics, Search Console, and third-party keyword data into one organic-performance signal per page, refreshed on a regular sync instead of rebuilt by hand
• A topic taxonomy with defined umbrella and supporting relationships, plus topic clusters and product / content mappings
• Consolidation, refresh, redirect, and canonical-ownership recommendations, backed by real organic traffic and visibility rather than word count alone
• Internal-linking recommendations connecting content to product pages
• A structured review workflow: every recommendation carries a priority, a confidence score, and a decision state, with rejected reasons captured so the system learns instead of repeating a miss
• Orphan-content triage that separates pages with no organic signal from pages quietly earning non-organic traffic, so nothing gets retired by mistake
• An audit of suspicious and irrelevant outbound links
• Reporting on internal links, orphans, canonical targets, and content performance

Selected artifacts

Sample rows from the content-cleanup review queue, generic candidates

Candidate	Type	Action	Priority	Confidence	Performance tier	Decision
Topic cluster, Set A (2 posts)	Overlap cluster	Merge or differentiate	High	85	High	Approved
Legacy guide, pre-2023	Blog cleanup	Refresh or consolidate	Medium	70	Moderate	Approved
Topic cluster, Set B (3 posts)	Overlap cluster	Merge or differentiate	High	90	High	Pending
How-to post, low internal links	Blog cleanup	Exclude or refresh	Medium	55	Low	Rejected
Definitional post, evergreen	Blog cleanup	Refresh or consolidate	Medium	70	Moderate	Approved
Topic cluster, Set C (2 posts)	Overlap cluster	Merge or differentiate	Medium	80	Moderate	Pending

Outcome

The content library now has a defined taxonomy, cluster-level consolidation and canonical decisions, and an internal-linking model tied to product mapping, informed by an automated organic-performance layer rather than judgment alone. It is being used to guide consolidation, internal linking, product alignment, and content prioritization, with every recommendation routed through review, and every rejection captured, before anything changes. No traffic, ranking, or conversion result is claimed here; the system change is in how the decisions get made.

Tools

Google Search Console, GA4, SEMrush, SurferSEO, WordPress, Obsidian, and AI-assisted analysis with custom workflows and scripts.

Reflection / key takeaway

Optimizing pages one at a time does not fix a content library. Structure does: a taxonomy, cluster-level decisions, and internal links that reflect how the topics and products actually relate.