Open-Source Playbook

How I Used AI to Build a Content Engine That Hit 11M Impressions in 90 Days

11.08M impressions and 44.8K clicks in 90 days โ€” with monthly clicks up 28x since May. The complete methodology, the diagnostic scripts, and the safety guards. Not theory. The actual operating system.

Monthly Impressions โ€” May โ†’ Sep
428K6.16M
11,081,935
Impressions
44,832
Clicks
1,232
Peak Daily
7.7
Avg Position
GSC Feedback LoopsMulti-Model AI AgentsQuality GatesTemplate Phrase DetectionNews SitemapsWebSub PingsKeyword AnticipationCannibalization ClustersInternal Link GraphBuild Cost ControlLiving PagesStructured DataGSC Feedback LoopsMulti-Model AI AgentsQuality GatesTemplate Phrase DetectionNews SitemapsWebSub PingsKeyword AnticipationCannibalization ClustersInternal Link GraphBuild Cost ControlLiving PagesStructured Data

What this is

In May 2026, ValueAddVC.com was a VC blog with 428K monthly impressions, 480 posts full of AI template phrases, and no system for deciding what to write, what to fix, or what to kill.

Four months later: 11.08M impressions and 44.8K clicks in 90 days, 1,090+ posts, and a self-improving feedback loop where Google Search Console data drives every editorial decision.

This playbook is the complete methodology. The guide below walks through every system we built. The open-source toolkit lets you run the same diagnostics on your own site.

The Guide

18 chapters across 4 acts
I

Build the Engine

Architecture, voice, formats, structured data

4 chapters
01

The Starting Point

What the site looked like at 428K monthly impressions โ€” the content debt, the AI fingerprints, the zero-strategy publishing.

02

Voice Training

How to make AI-written content sound like a specific human wrote it โ€” not generic, not robotic, not obviously AI.

03

The 5-Format System

Deep Explainers, News Analysis, Ranked Lists, Question-Led, and Contrarian Takes โ€” each with specific word counts, chart requirements, and selection weights.

04

Structured Data as Infrastructure

Article, FAQPage, ItemList, and NewsArticle schemas โ€” the markup that turns blog posts into rich results.

II

The GSC Reckoning

The data that changed everything

4 chapters
05

The AI-Overview Discovery

The moment we realized 80% of our impressions were AI citations โ€” queries that would never produce clicks. How that changed everything.

06

Title Rewrites at Scale

92 title rewrites targeting position 4โ€“20 pages with high impressions and low CTR. The methodology, the patterns, the results.

07

Cannibalization Clusters

21 clusters of pages competing for the same queries โ€” splitting authority and ranking worse than one consolidated page.

08

Template Phrase Purge

500+ AI fingerprints found and killed. 25 regex patterns for detecting the phrases that signal scaled-content-abuse to Google.

III

The Iteration Loop

Compounding improvements, week over week

6 chapters
09

Keyword Anticipation

Publishing before demand spikes โ€” IPOs, earnings, funding rounds. Being ranked when the world searches.

10

Living Pages

The refresh drip strategy: updating existing pages instead of only creating new ones. Staleness thresholds by content type.

11

The Internal Link Graph

191 orphan pages found and linked. How to build the link graph that Google's crawler actually follows.

12

Technical SEO Bugs

Broken OG images, leaked template variables, injected ad links, dead pages โ€” the health checks that run twice daily.

13

The News Wire

News sitemaps, WebSub pings, accurate lastmod โ€” getting crawled in minutes, not hours.

14

Build Cost Control

[nobuild] tags, deploy-tick safety nets, content-only skip logic, ISR caps โ€” how to run 960+ pages without spending $500/month on builds.

IV

The System

Quality gates, reviews, orchestration

4 chapters
15

Quality Gates

9 deterministic checks every piece of content must pass before publishing. Cannibalization, source verification, template phrases, typecheck.

16

The Weekly Review

The Sunday ritual: trending queries, dropping pages, CTR triage, query monopolies. How GSC data drives the next week's work.

17

Agent Orchestration

Multi-model AI pipelines: Opus/Fable for planning, Sonnet for writing, Haiku for mechanical tasks. Max 3 concurrent agents. Safety guards.

18

What I'd Do Differently

The mistakes that cost weeks โ€” and the principles that emerged from them.

The Toolkit

MIT-licensed, works with any site
26 Scriptsnode scripts/<name>.mjs
human-query-splitHuman vs AI-agent impressions per page
weekly-reportWeek-over-week movers and CTR triage
striking-distancePosition 5โ€“20 opportunities
ctr-auditClick gap vs your own CTR curve
gsc-rewrite-candidatesTitle rewrite candidates
cannibalization-detectorPages competing for one query
query-gap-minerQueries with no targeted page
refresh-trackerAging pages losing traffic
matched-control-readoutDid the change beat untouched pages?
rewrite-measurerOne change date vs a control band
traffic-guardDaily traffic-cliff alarm
sitemap-healthSitemap URLs that aren't indexable
redirect-checkerRedirect chains in sitemaps
broken-link-checkerBroken internal and external links
schema-validatorInvalid or incomplete JSON-LD
meta-length-checkerTruncated titles and descriptions
evidence-verifierFact quotes really on the source page
ai-citation-trackerWhich domains AI engines cite
factual-density-scorerExtractable facts per page
rewrite-detectorSame news story under a new URL
content-auditTraffic + quality scoring with labels
thin-content-detectorThin pages, with safety guards
template-detectorAI template-phrase scanner
orphan-finderPages nothing links to
websub-pingNotify the feed hub on publish
indexing-submitterIndexing API, job & livestream pages only
15 Configs
quality-gates.json
Publish quality gates
title-engineering.json
Title and snippet rules
aeo-rules.json
Answer-engine extractability rules
schema-rules.json
Structured data mirrors visible content
ctr-curve.json
Position โ†’ expected CTR
bot-traffic-rules.json
Human vs machine query patterns
noindex-strategy.json
When (not) to noindex
refresh-rules.json
Refresh drip by content type
anti-ai-rules.json
Template phrase blocklist
format-rotation.json
5-format content system
keyword-anticipation.json
Event calendar methodology
vertical-expansion.json
Adding new content verticals
health-checks.json
Live-site health monitors
content-pipeline-guards.json
Safety guards & repo locks
agent-orchestration.json
Multi-model AI pipeline
4
JSON-LD Schemas
Article, FAQ, ItemList, News
5
Code Examples
Sitemaps, components, Vercel
19
Sample Outputs
Generated from test fixtures
1
GitHub Action
Weekly automated report

Results

MetricMay โ€™26Sep โ€™26Change
Monthly Impressions428K6.16M14.4x
Monthly Clicks82623.4K28x
Click-Through Rate0.19%0.38%2x
Posts Audited480960+
Title Rewrites092
Cannibal Clusters Fixed021
Template Phrases500+0
Orphan Pages1910

Google Search Console, trailing 28 days ending May 24 and Sep 22, 2026. Every figure and window is in RESULTS.md.

Run it on your site

MIT-licensed. Works with any site that has Google Search Console access. Clone the repo, authenticate with GSC, and run your first diagnostic in under 5 minutes.

$ git clone https://github.com/TraceCohenTech/ai-seo-playbook.git
$ cd ai-seo-playbook && npm install
$ npm run rewrite-candidates -- --site sc-domain:yoursite.com
TC

Trace Cohen

Founder, Value Add Holdings LLC

3x founder, 65+ angel investments, 15 years in New York tech. Built ValueAddVC.com as a research platform for private-company data โ€” then used AI to scale it from a side project to 4.6M monthly impressions.