Automated SEO Intelligence Gathering: Building an AI-Driven News Analysis and Keyword Extraction Workflow with n8n

An open-source project based on n8n demonstrates how to automatically convert multi-source news RSS feeds into structured SEO insight reports using machine learning clustering and LLM summarization, providing content marketing teams with real-time topic tracking and keyword recommendations. Targeting the Japanese market as a demo, this project integrates RSS subscription, clustering, and LLM analysis, outputting results directly usable for content planning without manual intervention.

For B2B enterprises targeting specific markets, continuously tracking local news to extract SEO opportunities is time-consuming and prone to omissions. Traditional manual monitoring is inefficient and struggles to identify hidden cross-topic connections. With the popularization of AI automation tools, it's possible to build intelligent content intelligence systems at lower costs.

Developer carsonjan open-sourced the n8n-demo project, showcasing a complete automated SEO intelligence gathering solution. Data sources use RSS/XML subscriptions (resistant to changes, easy to parse) with a mixed Japanese-English strategy: Google News JP (comprehensive), Toyokeizai (economic and business), Nikkei Asia (international perspective on Japan), and the English version of METI (Japan's Ministry of Economy, Trade and Industry, official releases), ensuring deep local coverage and cross-language verification.

The core innovation of the workflow is machine learning clustering for preprocessing news titles. The process: convert titles to vectors using gemini-embedding-001 → reduce dimensionality to 50 dimensions with sklearn PCA → divide into 6 topic clusters using KMeans → map back to original titles to form topic groups. This step allows LLMs to focus on in-depth analysis within topics, avoiding token waste.

Clustered titles are passed to LLMs to generate structured reports. Each topic cluster includes: topic positioning (core topic area), content summary (key developments), SEO keywords (3-5 relevant terms), and content recommendations (customized creation directions for B2B tech transformation companies). The output directly integrates into content teams' workflows without secondary processing.

This workflow addresses core pain points in SEO intelligence gathering: real-time performance (instant RSS acquisition), balance between breadth and depth (clustering avoids missing edge topics + LLM deep understanding), and operability (output directly links to content creation). For regional marketing teams, it reduces labor costs and enhances the data-driven nature of content strategies.

The project architecture is scalable: the number of clusters, data sources, and LLM models can be adjusted; replacing RSS sources allows quick migration to other regions; the framework can be extended to areas like competitor monitoring and public opinion analysis.

n8n-demo showcases a practical solution combining traditional RSS with modern AI. Through a hierarchical architecture of machine learning preprocessing + LLM deep analysis, it achieves efficient conversion from raw news streams to structured SEO insights, serving as a reference open-source implementation for improving content intelligence automation.