Perplexity Deep Research
Deep Research is Perplexity's multi-step autonomous research agent, launched in January 2025. Instead of answering a question in a single search, it breaks complex research queries into sub-tasks, iteratively searches, reads, and synthesises across dozens of sources, and produces a structured report - similar to what a research assistant might produce in a few hours, in minutes.
How It Works
Query decomposition
Perplexity breaks the research question into 5–20 sub-questions
Iterative search
Searches for each sub-question; uses findings to generate follow-up searches (up to 30+ source pages)
Cross-source synthesis
Reads and cross-references multiple sources; identifies agreement and contradictions
Structured report
Produces a formatted report with sections, key findings, and numbered citations
The full process takes 2–5 minutes - visible to the user as a live progress display showing what it is currently searching. You can export the report as a PDF.
What It Excels At
- Market research: Competitive landscape, pricing, feature comparison across many players
- Investment due diligence: Background on companies, founders, funding history, news
- Academic overview: State-of-field summaries for unfamiliar domains
- Travel / logistics planning: Multi-source practical research
- Technical comparison: Evaluating tools, frameworks, or vendors across many criteria
Limitations
- • Cannot access paywalled content (scientific journals, Bloomberg, WSJ)
- • No access to private company data, internal databases, or recent filings behind login
- • Synthesis quality depends on what is publicly available - thin coverage = thin report
- • Not appropriate for primary source research requiring original data or interviews
- • Reports can include hallucinated connections between cited sources - always verify key claims
vs ChatGPT Deep Research & Gemini Advanced
| Feature | Perplexity | ChatGPT | Gemini |
|---|---|---|---|
| Speed | 2–5 min | 5–30 min | 3–10 min |
| Citations | Inline, numbered | Inline | Inline |
| PDF export | Yes | No (copy only) | Yes (Docs) |
| Sources count | 20–50 | 50–100+ | 20–40 |
| Price | Pro ($20/mo) | Plus/Pro ($20–200/mo) | Advanced ($20/mo) |
Tips for Better Reports
- Be specific: "Compare Pinecone vs Qdrant for a 50M vector production RAG system" beats "vector databases"
- Specify the output format you want: "Give me a table, then a recommendation"
- Ask it to focus on recency: "Focus on developments since January 2025"
- Always verify specific statistics, dates, and quotes against the cited source