Methodology

How We Win SEO in the AI-Native Era

We run the same stack we sell. A 12-stage content pipeline where data comes before LLM, not after. Three BigQuery cross-channel views that answer business questions no tool dashboard can formulate. Five quality gates that block every deliverable until it passes. This page describes exactly how it works — nothing held back.

67
Skills in production
12
Agents in OpenClaw
100%
Citation rate, SEO model
3
AI platforms tracked per client

The Problem We're Solving

Three gaps every SEO agency has. We closed them.

3 wks

Data lag

Reports assembled monthly from 5 tools that don't talk to each other.

BigQuery joins GSC + GA4 + Ads + CRM. Answer in one query.

10×

Wrong AI layer

Generic LLM produces content indistinct from the 10 pages already ranking.

12-stage pipeline. Data first, model second. Every claim cites a source.

$0

Attribution gap

Most agencies can't trace an organic visit to the deal that closed Tuesday.

CRM ingestion pipeline joins organic sessions to pipeline and revenue.

The Operating System

Five components that run our agency daily.

The stack is not theoretical. It handles 15 active client engagements across B2B SaaS, real estate, travel, HR tech, and AI verticals. Every component listed here is in daily production use.

67

Skills in production

Every task has a Skill. Every Skill loads the client's context, voice rules, and strategy before a word is written.

80MB

SEO knowledge corpus

Fine-tuned model trained on Ahrefs, Backlinko, Moz, SEJ, Google Search Central. 100% citation rate. Zero fabrication.

12

Agents in OpenClaw

Guardian runs every 60 seconds. SEO Sentinel every 6 hours. Competitor Watcher daily. Each on a fixed schedule with health checks.

1

CLAUDE.md per client

Keywords, positions, brand rules, banned terms, active strategy — loaded automatically before the first keystroke of every session.

120

Automated workflows

GSC weekly, GA4 nightly, Ads daily, rank tracking, crawl triggers, Slack reporting. Mechanical work is gone before analysis starts.

3

AI platforms tracked

Google AI Overview, ChatGPT, Perplexity. Citation presence per commercial query, per client. Reportable, not anecdotal.

The Research Protocol

12 stages. Data first, model second. No shortcuts.

Every piece of content runs this sequence in full. The order is fixed because each stage invalidates assumptions the previous one could not see. Skipping Stage 5 means you write for what ranks generally, not what converts for this specific client.

01

Keyword resolve

BigQuery first. If no data history, DataForSEO. Never gut feel.

02

Live SERP

Top 10 via DataForSEO API — titles, lengths, SERP features, AI Overview presence.

03

Scrape top 10

Scrapling extracts heading hierarchy, entities, schema types. Structured JSON, not raw HTML.

04

Content gap analysis

What the top-10 pages have that ours doesn't. Content gap, not keyword gap.

05

BigQuery historical context

Client's own GSC + paid conversion data. This overrides external assumptions.

06

SEO KB grounding

seo_query_kb returns citation-grounded best practice. CONFIDENCE: ESCALATE → switch to live data.

07

Evidence brief

Stages 1–6 synthesised into a single brief. Writer drafts from this — never a vague prompt.

08–09

Write + anti-AI edit

Client writer skill drafts. 8-dimension validator strips 50+ AI-tell words and patterns.

10

Brand voice edit

Client editor enforces tone, banned terminology, competitor rules.

11

Validation gate

85/100 minimum. Pipeline halts below threshold and lists specific fixes.

12

Schema generation

JSON-LD for the page type. 4-layer validation before deploy.

BOFU Phase Gate

Before any content enters Stage 1, we confirm a phase gate: does the client have live pages for at least two competitor alternatives articles, one head-to-head comparison, and one “best [category] for [niche]” listicle? If not, the requested content must be one of those types — or it is deferred. At any moment, only ~5% of your target market is actively in-market. BOFU content captures that 5%, which converts at 10x the rate of cold informational traffic. Informational content before BOFU is live is traffic without pipeline.

How an Engagement Flows

5 phases. Fixed sequence. No exceptions.

01

Baseline + audit

Weeks 1–4

Screaming Frog crawl → CSVGSC + GA4 → BigQueryP0 / P1 / P2 issue list6-month roadmap
02

Technical fixes + data infra

Weeks 2–8

P0 → P1 → P2 sequenceGA4 events fixed Week 1Schema 4-layer validatedInternal linking reviewed
03

BOFU content first

Weeks 4–ongoing

Competitor alternativesHead-to-head comparisonsCategory listicles12-stage pipeline starts
04

Link building: 2 tracks

Weeks 6–ongoing

Track 2: external listicles (priority)DA50+ links + LLM citationsTrack 1: DR40+ outreachAhrefs competitor gap
05

Measurement + reporting

Weekly cadence

GSC weekly → BigQueryGA4 events nightlyCRM-organic join (if access)Fixed weekly report

How We Use BigQuery

3 views. 5 sources joined. Answers no dashboard gives.

v_keyword_unified

Paid keyword → organic rank. One query.

PAID_ORGANIC_OVERLAPSEO_CONTENT_GAPSEO_GROWTH_OPPORTUNITYSEO_PUSH_TO_PAGE1

Shows where paid budget is redundant and where content investment closes a proven gap.

v_url_performance

Every URL: paid spend + organic traffic + GA4 conversions.

Paid spend redundant (strong organic rank)Organic traffic, zero GA4 conversionsConversions but no organic traffic15–30% paid overlap found on typical client

Moves budget decisions from opinion to data. The 15–30% overlap usually funds the SEO retainer.

v_search_term_opportunities

Converting paid terms with no organic page. Build those.

Converting terms with zero organic targetingHigh conversion count + high CPAMissing-intent: what buyers search vs what tools suggest

The content roadmap. Every page idea traces to a term already proven to produce customers.

GSC

Rankings, CTR — weekly

GA4

Events, conversions — nightly

Google Ads

Spend, search terms

CRM

Pipeline + revenue

Ahrefs

Backlinks, rank tracking

Quality Gates

5 checks. /deliver blocks automatically.

01

>75/ 100

Brand validation

Voice, banned terms, CTA rules, ICP fit. Per-client validator — not a shared template.

02

>80/ 100

SEO compliance

Title, meta, heading hierarchy, keyword density, schema type, OG tags. No critical failures.

03

>85/ 100

Anti-AI validation

50+ banned words, 8 dimensions. "leverage", "seamless", "meticulously" each trigger a block.

04

100%pass

4-layer schema

JSON-LD parse → Rich Results rules → HTML embed → Search Console confirms. One fail = no deploy.

05

Everynumber

Data traceability

Every stat traces to a Python script in the project directory. AI cannot estimate a measured number.

Human Work vs Agent Work

We are not replacing strategists with agents.

What humans handle

  • Client strategy and priority calls
  • Interpreting what data means for a specific client
  • Brand voice decisions and editorial judgment
  • Recognizing when a keyword cluster does not match buyer intent
  • Competitive positioning recommendations
  • Client relationship and communication
  • Deciding what not to build

What agents handle

  • Keyword research and clustering at scale
  • Content brief generation with SEO scoring
  • Schema markup generation and validation
  • GSC and GA4 data ingestion and warehousing
  • Weekly rank tracking and reporting
  • Backlink gap analysis and prospect research
  • Anti-AI content validation (8-dimension scoring)
  • Internal linking opportunity identification
  • Technical audit crawl data processing

The ratio shifts over time. As we build deeper context for a client, more judgment work gets encoded into Skills and the fine-tuned model. But the loop always has a human reviewing the output before anything leaves the agency.

Why the Model Works

4 structural advantages. Each one measured.

15–30%

Paid overlap on typical onboard

v_keyword_unified flags every keyword you pay for where organic rank ≤ 3 already earns the click free.

0

Guesses on the content roadmap

Every page idea traces to a converting paid term in v_search_term_opportunities. Proven demand, not prediction.

5

Gates before anything reaches you

/deliver blocks automatically on brand, SEO, anti-AI, schema, and data traceability.

10×

BOFU conversion rate vs informational

Competitor alternatives are live before blog posts start. At any moment, ~5% of your ICP is actively buying.

See it in practice

Eight active client engagements running on this methodology, spanning five industries.

View case studies

Ready to Work Together?

See this methodology applied to your site.

We start with a technical audit and data infrastructure setup before making any promises. If the foundation is not in place, we build it first.

  • Free 30-minute audit call
  • Technical baseline review included
  • Fixed-scope proposal within 48 hours