Build an AI research workflow
A repeatable process for collecting sources, extracting evidence, organizing findings and checking important claims.
Workflow map
Question → sources → extraction → evidence table → fact-check → output
Core tools: Perplexity or another research tool for source discovery, ChatGPT or Claude for analysis, and a document or database for evidence tracking.
Define the research task
Write one focused question, define the date range and list the decisions the research should support. Set inclusion and exclusion rules before collecting sources.
Collect source material
Prioritize primary documents, official announcements, papers and reputable reporting. Save the source URL, title, publisher and publication date.
Extract claims and evidence
Give the AI assistant the source material and ask for claims, supporting evidence, dates, numbers and uncertainties. Keep citations attached to each finding.
Build an evidence table
Store each claim beside its source, date, supporting passage and confidence note. Separate sourced facts from AI-generated interpretation.
Fact-check important claims
Open the original source for material claims, confirm dates and figures, and resolve conflicting evidence before publishing or making a decision.
Suggested output
- Executive summary
- Key findings with citations
- Evidence and source table
- Conflicting or uncertain points
- Open questions and next checks
AI output can contain errors. Treat source documents as the authority for important factual claims.