Skip to main content
This document explains the AI Web Research node, which combines web search and structured data extraction capabilities into one powerful automation node. Built on advanced AI models, this node enables automated web research, data analysis, and information synthesis from multiple sources.

​
Getting Started

​
Quick Setup Video

​
Step-by-Step Guide

1

Add your research prompt

Write a clear prompt describing what you want to research
  • Use the format: “Given [input], find/analyze/research [output]”
  • Example: "Given a company name, find their latest funding and news"
2

Generate Inputs and Outputs

Click the button to create your schema
  • The AI analyzes your prompt and generates appropriate fields
  • Review the generated inputs and outputs
3

Connect inputs

Link data from previous nodes
  • The node shows expected input types (List or single value)
  • Match your data sources to the generated inputs
4

Select Research Type

Choose your processor
  • Use Auto-Select for intelligent optimization
  • Or manually select based on your needs
5

Run and review

Execute the research and check outputs
  • Citations and reasoning are always included
  • Connect outputs to downstream nodes

​
Schema Generation

​
Initial Generation

When you click “Generate Inputs and Outputs”, the system creates a custom schema based on your research prompt. AI Web Research Node

​
Schema Refinement

After generating your initial schema, you can refine it if needed by clicking “Regenerate Inputs and Outputs” again. This opens a dialog with two options: Schema Refinement Dialog
When to use: You want to adjust the existing fields without starting overHow it works:
  • Provide feedback on what to change
  • The AI modifies your current schema based on feedback
  • Preserves the overall structure while making adjustments
Example refinements:
  • “Add funding information and remove the website field”
  • “Include employee count and industry classification”
  • “Change company description to be more detailed”

​
Research Type Processors

Pro tip: Start with Auto-Select mode - it intelligently chooses between lite, base, and core processors to optimize for both cost and performance.

​
Processor Comparison

ProcessorCreditsTimeMax FieldsBest Use Cases
lite45-60s~2Quick lookups, simple facts
base815-100s~5Standard enrichment, basic research
core201-5m~10Business research, cross-validation
pro803-9m~20Exploratory research, deep analysis
ultra2005-25m~20Comprehensive reports, PDF analysis

​
Processor Selection Guide

Perfect for simple, fast lookupsUse cases:
  • Company addresses and phone numbers
  • Website URLs and social media handles
  • Basic business information (founded date, CEO name)
  • Simple yes/no verifications
Real examples:
"Given a company name, find their headquarters address"
"Given a website, extract the contact email"
"Given a business name, find if they have a mobile app"
Ideal for standard enrichment tasksUse cases:
  • Product offerings and service descriptions
  • Team size and office locations
  • Industry classification and business model
  • Recent announcements or updates
Real examples:
"Given a company website, extract their main products and pricing tiers"
"Given a startup name, find their target market and value proposition"
"Given a business domain, identify their key partnerships"
Recommended for most business researchUse cases:
  • Competitive positioning and market analysis
  • Financial metrics and growth indicators
  • Leadership team and board composition
  • Technology stack and integrations
Real examples:
"Given a company, research their funding history, investors, and valuation"
"Given a competitor list, analyze their pricing strategies and differentiators"
"Given an industry, identify top players and market dynamics"
Core processor includes confidence scores and detailed citations for each field.
For complex, exploratory researchUse cases:
  • Multi-dimensional company analysis
  • Deep competitive intelligence
  • Comprehensive market research
  • Investment due diligence
Real examples:
"Given a company, analyze their business model, revenue streams, competitive advantages, risks, and growth potential"
"Given a market segment, research all major players, their strategies, partnerships, and recent developments"
"Given an acquisition target, evaluate their technology, team, financials, and strategic fit"
Maximum depth for critical researchUse cases:
  • Analyzing lengthy PDFs and reports on the web
  • Comprehensive regulatory compliance research
  • Deep technical documentation analysis
  • Multi-source investigative research
Real examples:
"Given a company's SEC filings URL, extract all financial metrics, risks, and strategic initiatives"
"Given a technology, research all implementations, case studies, benchmarks, and limitations"
"Given an industry regulation, analyze compliance requirements, penalties, and implementation guidelines"

​
Output Structure

​
Standard Outputs

All research tasks include these base outputs:

Citations

Source URLs and references for all findings

Reasoning

Detailed explanation of research methodology
The language of the Reasoning output is automatically inferred from your research prompt and input data. If your prompt is written in French, the reasoning will be returned in French. To ensure consistent language across runs, write your research prompt in your preferred language.

​
Enhanced Outputs (Core/Pro/Ultra)

Advanced processors provide additional metadata for each field:

Field-Specific Reasoning

[field_name]_reasoning - How each value was determined

Field Citations

[field_name]_citations - Sources for specific data points

Confidence Scores

[field_name]_confidence - Reliability rating (High, Medium, or Low)

​
Practical Examples

​
Sales Intelligence Workflow

Research Prompt: "Given a company domain, find decision makers, 
                 recent news, and technology stack"
Processor: core
Inputs: company_domain
Outputs: 
  - executives (with LinkedIn URLs)
  - recent_developments
  - tech_stack
  - company_size
  - funding_status

​
Investment Research Pipeline

Research Prompt: "Given a startup, analyze their market position, 
                 team, traction, and competitive landscape"
Processor: pro
Inputs: startup_name, website
Outputs:
  - founding_team_background
  - market_size
  - key_competitors
  - unique_advantages
  - customer_traction
  - risk_factors

​
Best Practices

​
Writing Effective Prompts

  • Be specific about what information you need
  • Use clear input/output structure
  • Specify the context or use case
  • Include any special requirements
Good examples:
  • “Given a SaaS company website, extract pricing tiers, features, and integration partners”
  • “Given a company name and industry, find their main competitors and market share”

​
Optimization Strategies

1

Start with Auto-Select

Let the system optimize processor selection for you
2

Test with small batches

Validate your schema with 2-3 examples before scaling
3

Monitor credit usage

Track consumption and adjust processors as needed
4

Chain nodes strategically

Split complex research into multiple focused nodes

​
Advanced Techniques

​
Research Chaining

For comprehensive analysis, chain multiple nodes:

​
Troubleshooting

The Ultra processor can take up to 25 minutes. Consider using lower processors if speed is critical.
  • Verify your research prompt is clear and specific
  • Click “Regenerate Inputs and Outputs” to update schema
  • Ensure all required inputs are connected
  • Check that input data is in the correct format
  • Add more specific requirements to your prompt
  • Upgrade to core or higher processors
  • Review citations to understand data sources
  • Use confidence scores to filter results
The AI Web Research node represents the most advanced research capability on the Gumloop platform, combining automated web research with precise data extraction to deliver comprehensive, accurate results tailored to your business automation needs.