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
lite25-60s~2Quick lookups, simple facts
base415-100s~5Standard enrichment, basic research
core101-5m~10Business research, cross-validation
pro403-9m~20Exploratory research, deep analysis
ultra1005-25m~20Comprehensive reports, PDF analysis

Processor Selection Guide

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

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 - Accuracy confidence (0-100)

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.
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.