Perplexity Deep Research

Deep Research is Perplexity’s autonomous research mode, launched in February 2025. Where a normal Perplexity query runs one search-and-summarize pass, Deep Research plans an approach, issues dozens of searches, reads hundreds of sources, revises its plan as it learns, and assembles a structured report with inline citations for every claim. A typical run finishes in two to four minutes — noticeably faster than OpenAI’s Deep Research (which can take up to 30 minutes) or Google Gemini’s equivalent.

The other headline difference is access. OpenAI initially gated its version behind the $200/month Pro plan; Perplexity shipped Deep Research with a limited number of free queries per day for everyone, and effectively unlimited use on the $20/month Pro subscription. On research benchmarks like Humanity’s Last Exam, Perplexity’s version scored above standard chatbot models though below OpenAI’s agent — a fair summary of its position: faster and cheaper, somewhat less thorough.

Reports render with headings, tables, and numbered citations, and can be exported to PDF or published as a Perplexity Page. Follow-up questions stay in context, so you can drill into one section without restarting the run.

Key Features

• Autonomous multi-step search — dozens of queries and hundreds of sources per run
• Reports typically completed in 2-4 minutes
• Inline numbered citations linking every claim to its source
• Structured output with sections and tables rather than one wall of text
• Export to PDF or publish as a shareable Perplexity Page
• Conversational follow-ups that refine or extend the report
• Available on web, iOS, Android, and Mac apps

Pricing

• Free: A limited number of Deep Research queries per day
• Pro ($20/month): Effectively unlimited Deep Research, plus model choice, file uploads, and Perplexity Labs
• Enterprise Pro (~$40/user/month): Team management, SSO, and data-retention controls

Best For

Analysts, consultants, journalists, and students who need a decent cited overview of an unfamiliar topic in minutes — market landscapes, competitor scans, policy backgrounders — and who will verify the sources that matter before relying on them.

Limitations

Speed comes at a cost in depth: on genuinely hard analytical questions, OpenAI’s Deep Research and Gemini’s version usually go deeper. Perplexity also weights sources unevenly — affiliate listicles can appear next to peer-reviewed work — so treat the citations as leads to check, not verified facts. For academic literature specifically, Elicit or Consensus are better fits.