Intelligent Multi-Stage Web Research and Expert Report Generator - n8n Workflow

• Market Researchers & Analysts: Users needing deep, data-driven reports on specific topics.


• Content Creators & Technical Writers: Those who require highly accurate, referenced content synthesized from multiple online sources.


• n8n Automation Specialists: Users looking for advanced examples of Retrieval-Augmented Generation (RAG) implementation within an n8n workflow, leveraging LangChain nodes.


• Businesses: Organizations that need to automate knowledge retrieval and competitive intelligence gathering.

This sophisticated automation transforms a simple research question into a comprehensive, expert-level report. The process uses multiple steps of AI-driven analysis, ensuring that the final output is based on the most relevant and current web data available. Unlike single-query solutions, this n8n workflow employs a two-stage search strategy (Query 1 and Query 2) to refine results iteratively, guaranteeing high relevance. By using the Brave Search API for queries and a dedicated Article Extractor API, the n8n solution overcomes common limitations of basic web scraping. This structure makes this one of the most powerful n8n templates for advanced knowledge workers seeking reliable AI-generated research, complete with APA7 citations. Every n8n node is carefully configured to maximize data quality and extraction efficiency.

The entire process is initiated by an n8n trigger, specifically a Webhook, which accepts the 'Research Question'.

1. Preparation: The initial query is captured alongside the current date.


2. Prompt Engineering: The Analyst Emulator n8n node uses Google Gemini (via LangChain) to convert the user's request into a high-level, structured system prompt for complex analysis.


3. Query Optimization (Stage 1): The Query Maker - 1 n8n node uses this system prompt to generate the optimal initial search query.


4. Web Search (Stage 1): The Query 1 n8n node executes the search using the Brave Search API.


5. Ranking and Refinement: The Query 1 Ranker & Query 2 Maker n8n node analyzes the search results, ranks the top three most relevant URLs, extracts initial snippets, and critically formulates a refined New_Query for a deeper dive.


6. Web Search (Stage 2): If necessary, Query 2 executes the refined search.


7. URL Consolidation: The n8n workflow merges the top-ranked URLs from both queries into a list of five authoritative sources.


8. Article Extraction: The workflow iterates over the five URLs using a Split In Batches node. For each URL, the Article Extractor1 n8n node fetches the full, clean text content via RapidAPI. A Delay n8n node is strategically placed to avoid hitting API rate limits.


9. Report Generation (RAG): All extracted content is aggregated. The final Research Reporter n8n node uses a massive prompt, leveraging all the gathered text and snippets, to synthesize a comprehensive, multi-paragraph report tailored for an expert audience, complete with APA7 citations. This is the ultimate output of the n8n workflow.


10. Delivery: The final report is returned via the Respond to Webhook node.

To deploy this expert n8n workflow, follow these steps:

1. Import the n8n Workflow: Copy the provided JSON data and import it into your n8n instance.


2. API Keys - Google Gemini/LLM: Configure credentials for the Google Gemini Chat Models (Google Gemini Chat Model, Google Gemini Chat Model14, Google Gemini Chat Model15, Google Gemini Chat Model16). If you are using another LLM, replace the existing LLM n8n node connections accordingly.


3. API Keys - Brave Search: Obtain a free API key from the Brave Search API website (api.search.brave.com). Paste this key into the X-Subscription-Token header parameter within both the Query 1 and Query 2 HTTP Request nodes. This is essential for the primary n8n trigger action.


4. API Keys - Article Extractor: Obtain an API key for a suitable Article Extraction service (e.g., RapidAPI's Article Extractor 2). Update the X-RapidAPI-Key and X-RapidAPI-Host headers in the Article Extractor1 n8n node.


5. Webhook Setup: Activate the main Webhook n8n trigger node. Copy the 'Production URL'.


6. Test Execution: Configure the Webhook Call n8n node (located in the documentation sidebar) with the copied Production URL and your 'Research Question'. Execute the test to see the complex n8n automation in action.

This n8n workflow relies heavily on LangChain LLM nodes for intelligent decision-making and text generation.

Webhook (n8n trigger): The starting point for the n8n workflow, accepting the user's 'Research Question' via a query parameter.
Analyst Emulator (chainLlm n8n node): Uses Gemini to generate an extremely sophisticated system prompt, defining the analysis methodology (meta-reasoning, three chains of thought) for the subsequent LLM calls.
Query Maker - 1 (chainLlm n8n node): Generates the first optimized, non-long-tail search query based on the user request and today's date, outputting a structured JSON object.
Query 1 & Query 2 (HTTP Request n8n node): Executes web searches using the Brave Search API. Requires the Brave API key in the header.
Query-1 Combined / Query-2 Combined (Code n8n node): Processes the raw JSON results from the Brave API, extracting the title, URL, and description snippets, and aggregating them into a single text block for LLM processing.
Query 1 Ranker & Query 2 Maker (chainLlm n8n node): A critical decision-making point. It ranks the top 3 URLs from Query 1, extracts initial findings, and dynamically generates a follow-up query (New_Query) if the initial results are insufficient. It ensures the n8n workflow is adaptable.
Code (n8n node - URL Consolidation): Combines the top 3 URLs from Ranker 1 and the top 2 URLs from Ranker 2, preparing five items for the batch processing loop.
Loop Over Items (Split In Batches n8n node): Manages the concurrent or sequential fetching of the top 5 articles.
Article Extractor1 (HTTP Request n8n node): Hits a third-party API (e.g., RapidAPI) to retrieve the full, clean article content for each ranked URL, effectively performing structured web scraping.
Delay-to-Avoid-Request-Per-Minute-Cap (Delay n8n node): Essential flow control to prevent throttling when calling external APIs during the loop iteration.

• Research Reporter (chainLlm n8n node): The final stage. This powerful n8n node performs Retrieval-Augmented Generation (RAG), synthesizing the original request, extracted snippets, and the full text of all articles into a polished, expert-level report that adheres to strict length and citation standards (APA7).