Research Agents

Introduction

A research agent autonomously searches sources, reads, synthesises and cites. OpenAI Deep Research, Perplexity Pro, and Elicit are leading examples.

Beginner analogy: a journalist who knows how to chase leads, cross-check sources and write a balanced piece.

Understanding the topic

Core concepts:

• Search tool + browser tool + summariser.
• Track sources for citations.
• Cross-check claims across ≥ 2 sources.
• Use a critic to evaluate completeness.
• Output: structured report with citations.

Syntax reference

Visual workflow / architecture:

Question → search → top-N URLs → fetch → extract → cluster
                                                       │
                                                       ▼
                                            synthesise + cite

Real-world use

Perplexity, OpenAI Deep Research, Elicit (academic), Storm (Stanford).

Best practices

• Always require citations.
• Use a deduper to avoid quoting the same source twice.

Common mistakes

• Synthesising without citations — looks like hallucination.

Hands-on exercise

Interview preparation — practice these questions:

• Q1. Five components of a research agent.
• Q2. Why cross-check sources?
• Q3. Scenario: agent cites a 404 page. Where do you intervene?