Dynamic AI Research Generator for Structured Data and CSV Export - n8n Workflow

Market Researchers: Users needing large, structured lists of companies, people, or products with specific data fields. Sales and Outreach Teams: Teams looking to generate highly enriched lead lists tailored to very specific criteria. Data Engineers: Anyone building sophisticated data pipelines that require dynamic, structured data ingestion from the web using an n8n workflow. Automation Specialists: Those looking for advanced n8n templates demonstrating complex AI orchestration and looping logic.

The modern challenge of web research lies not in finding data, but in structuring it reliably. This powerful n8n workflow solves this by using AI as the architect. Instead of relying on static configurations, the workflow first analyzes your plain text research request (e.g., 'Find 50 German software CEOs and list their LinkedIn, employee count, and recent funding').

An AI Planning n8n node dynamically generates the exact JSON schema and the necessary search queries. It acts as a thinking layer that decides precisely what columns the final CSV should have. The n8n workflow then systematically discovers the list of items and, in a critical looping phase, performs secondary, detailed AI-powered web searches (via Linkup) to enrich every single item based on the dynamically defined schema. This results in highly accurate, ready-to-use structured data, delivered instantly as a custom CSV file. This advanced n8n solution drastically reduces manual data cleanup.

This powerful n8n workflow executes research in three primary phases: Planning, Discovery, and Enrichment.

1. Trigger: The n8n workflow starts with the On form submission n8n trigger, capturing the user's plain text research description.


2. AI Planning: The input is sent to the Prepare prompts and schema n8n node (a Langchain LLM node leveraging GPT-5). This node acts as the 'brain,' defining the research strategy, including the discoveryQuery, the list schema (discoverySchema), the enrichment query (enrichmentQuery), and the final structured schema (enrichmentSchema). The OpenAI Chat Model n8n node ensures the output is a perfectly structured JSON object.


3. Initial Discovery: The n8n node Query Linkup to find the list executes the initial search using the generated discoveryQuery and discoverySchema to find the core items (e.g., company names).


4. Item Splitting: The Split Out n8n node separates the resulting list array into individual data items, preparing them for enrichment.


5. Enrichment Loop: The Loop Over Items n8n node begins the parallel enrichment process. For each item, the n8n workflow performs a second highly targeted HTTP Request (Query Linkup to find all properties for this item). This request uses the AI-generated enrichmentQuery and enrichmentSchema to fetch all necessary details (website, employee count, contacts, etc.) for that specific item.


6. Data Preparation: The Prepare final JSON for that item n8n node merges the initial identifying property with the newly enriched data into a single, clean JSON object.


7. Final Output: Once all items are processed and collected back into a single set, the Convert to CSV n8n node compiles the structured data into the final, ready-to-use CSV file.

To deploy and use this advanced n8n workflow, follow these steps:

1. Import the n8n workflow: Copy the provided JSON code and import it directly into your n8n instance using the 'New' menu > 'Import from JSON'. This template showcases advanced n8n automation.


2. OpenAI Credentials: The Prepare prompts and schema Langchain n8n node requires credentials for the OpenAI Chat Model. Ensure you connect your valid OpenAI API key to this n8n node.


3. Linkup Credentials: The workflow uses two HTTP Request n8n nodes for web searches (Query Linkup to find the list and Query Linkup to find all properties for this item). Both require a Bearer Token for the Linkup API. Configure the httpBearerAuth credential type in n8n with your Linkup API key.


4. Activate the n8n Trigger: Activate the On form submission n8n trigger and use the generated public form URL to submit your research requests.

On form submission (n8n trigger): The starting point of the n8n workflow. It provides a simple web form input field titled 'Describe your research' which captures the user's requirements.
Prepare prompts and schema (Langchain LLM Chain n8n node): This crucial n8n node orchestrates the research plan. It uses a detailed system prompt and GPT-5 to output five keys: ObjectName, discoveryQuery, discoverySchema, enrichmentQuery, and enrichmentSchema. It dictates the entire structure of the final output.
OpenAI Chat Model (LLM n8n node): Provides the processing power for the planner. It is configured to enforce a JSON object response format, ensuring reliable structured output for the downstream n8n nodes.
Query Linkup to find the list (HTTP Request n8n node): Performs the initial web search using the AI-generated discoveryQuery and discoverySchema. This n8n node is responsible for finding the list of items to be enriched.
Split Out (n8n node): Splits the array of discovered items (the List property) so that each item can be processed sequentially or in parallel during the enrichment loop.
Query Linkup to find all properties for this item (HTTP Request n8n node): Executed within the loop, this n8n node uses the AI-generated enrichmentSchema and enrichmentQuery to pull detailed, structured information for one single item.
Prepare final JSON for that item (Code n8n node): A custom JavaScript n8n node that merges the core identifying property (e.g., Company Name) with all the enriched properties from the Linkup search, preparing the clean JSON structure for the CSV conversion.
Convert to CSV (n8n node): The final action, converting the stream of structured JSON objects into a single, downloadable CSV file, completing the automation delivered by this n8n workflow.