AI · Automation 2025 Completed
Google Deep Research — AI-Powered Research Assistant
A reusable n8n sub-workflow that generates search queries, crawls Google results, converts pages to Markdown, and returns AI-summarized, relevance-filtered research.
Built by Redoan Uzzaman, Full-Stack Developer & AI Engineer.
Overview
Google Deep Research is a building-block n8n sub-workflow. Any parent workflow can call it with a research context, and it returns a clean set of summarized, relevant findings.
How it works
- An LLM turns the provided context into several diverse search terms (the number is configurable).
- Each term is run through the Google Custom Search API, with 3–10 results per query and an option to restrict results to Reddit for community insight.
- Documents such as PDFs are filtered out, pages are downloaded, failed fetches are skipped, and the HTML body is extracted and converted to Markdown.
- GPT-4o-mini summarizes each page, then a second pass judges whether it actually matches the research context.
- Relevant summaries are aggregated and returned to the caller.
Engineering notes
Three LLM chains with structured output parsers keep every step machine-readable, and error filtering keeps one broken link from failing the run. The same sub-workflow powers the research stage of the AI Content Generation Engine.
Key Features
- AI-generated, diverse search queries
- Google Custom Search integration with optional Reddit focus
- Full-page extraction and HTML-to-Markdown conversion
- Per-page AI summarization
- AI relevance filtering against the research context
- Callable sub-workflow with configurable parameters
Tech Stack
- n8n
- OpenAI GPT-4o-mini
- Google Custom Search API
- Markdown
- JavaScript
Want something like this built?
Redoan Uzzaman takes on AI, automation and full-stack projects for clients worldwide.