FORM PR-02 · PROJECT PR-0004 · AI Workflow Engineering · SEO Analytics

AI SEO Strategist

Built an AI-assisted SEO workflow in n8n that brings together search performance, website analytics, keyword research, and technical crawl findings. The system turns that evidence into a focused Top 5 action plan, with supporting findings, business rationale, and confidence for each recommendation. My work focused on making recommendations traceable, keeping incomplete data visible, and helping someone decide what to work on next.

PR-0004.1

The problem

Scattered evidence, unclear next steps

Inputs: Four complementary evidence sources
Scope: Up to five selected pages plus research topics
Output: Up to five prioritized, reviewable actions

Why it matters

Search visibility, website engagement, keyword opportunities, and technical health sit in different tools. Comparing them manually takes time, and a collection of metrics does not explain which action deserves attention. I brought these sources into one workflow so recommendations can be assessed alongside the findings that support them.

PR-0004.2

How it works

From four sources to a focused action plan

Stage	System behavior
01	Select a repeatable set of relevant pages and an explicit reporting window.
02	Request page and query performance from Google Search Console, alongside organic landing-page analytics from GA4.
03	Import OpenSEO keyword research and Screaming Frog crawl findings from CSV or JSON files.
04	Wait for the evidence branches, match records by normalized URL, and retain source dates, coverage, and limitations.
05	Analyze each page and carry unmatched keyword research forward as proposed content opportunities.
06	Rank the strongest candidates across the selected scope, validate the response, and remove duplicate actions.
07	Produce one HTML and structured JSON action plan for human review, with no more than five distinct recommendations.

Four evidence sources
Top 5 action plan
Human review

PR-0004.3

Problems solved

Making the workflow dependable beyond one page

Added synchronization points so a fast source cannot trigger a report before the other branches finish.
Corrected processing modes and preserved item links so every selected page reaches analysis.
Replaced positional joins with URL matching and mapped actual crawl-export headers to the evidence model.
Kept failed and missing measurements visible instead of converting unavailable evidence into zero.
Preserved unmatched research topics without attaching performance figures from unrelated pages.
Validated recommendation fields, evidence references, priority, and confidence before the report is assembled.
Used automated fixtures to check multiple pages, source failures, stale inputs, invalid model responses, and one combined report.

PR-0004.4

Delivered work and strengths

Integration · Evidence handling · AI controls · Reporting

Integration

n8n
Google Search Console
GA4 Data API
OpenSEO Imports
Screaming Frog Imports
CSV / JSON

Evidence handling

Repeatable Page Selection
URL Matching
Nested Query Evidence
Source Dates
Coverage Checks
Research Opportunities

AI controls

Page-Level Analysis
Cross-Page Prioritization
Structured Responses
Evidence Checks
Duplicate Removal
Human Review

Failure handling

Branch Synchronization
Item Linking
Missing-Data Handling
Source Failure Isolation
Stale-Data Flags
Fixture Verification

Delivered artifacts

Importable Workflow
Evidence Model
Top 5 HTML / JSON Report
Import Templates
Setup Guide
Automated Checks