Perplexity has introduced Deep Research, an AI-powered tool designed for conducting in-depth analysis across various fields, including finance, marketing, and technology. The system automates the research process by performing multiple searches, analyzing extensive sources, and synthesizing findings into structured reports within minutes.
Deep Research enhances Perplexity’s existing AI-powered question-answering capabilities by adopting a more methodical research approach:
- Iterative Research Process – the AI searches, reads, and refines its approach dynamically, mimicking a human researcher’s ability to adjust based on new findings.
- Comprehensive Report Writing – after analyzing the source material, the report displays synthesized information in a structured and coherent manner.
- Seamless Export & Sharing – users can export reports as PDFs or documents or convert them into a Perplexity Page for easy sharing.
Perplexity Deep Research demonstrates industry-leading accuracy, outperforming competing AI models on key benchmarks:
- 93.9% accuracy on SimpleQA – A factuality benchmark evaluating AI’s ability to provide correct and reliable answers.
Source: https://www.perplexity.ai/hub/blog/introducing-perplexity-deep-research
- 21.1% on Humanity’s Last Exam – A rigorous benchmark testing AI across 100+ subjects, from science and mathematics to history and literature.
Source: https://www.perplexity.ai/hub/blog/introducing-perplexity-deep-research
Aravind Srinivas, CEO of Perplexity, commenting on the benchmark results, noted:
Perplexity Deep Research is quite close to OpenAI o3 on the Humanity Last Exam Benchmark despite being an order of magnitude faster and cheaper. This is possible because DeepSeek is open source and cheap and fast.
In direct comparisons, Perplexity Deep Research has shown notable advantages in speed and source utilization. Alban Brooke, after testing the same prompt on Perplexity Pro and ChatGPT Pro, shared the results and insights:
- Perplexity was ~9x faster
- Perplexity referenced 150% more sources (50 vs. 20)
- ChatGPT provided a more in-depth but significantly longer response (31 pages vs. 6)
If I could only use one of these, it would probably be Perplexity Deep Research.
Initial users have given positive feedback on Deep Research, highlighting its accuracy and reliable sources. For instance, Richard Moore, a head of IT strategy and architecture at AJ Bel, remarked:
Wow, this is very very good! I've tried a couple of queries, and the results are significantly better than any other research tool I've tried. The sources are all high quality too, which is where open-source research tools (e.g., Storm) fail.
The tool is now available for free to all users, with Pro subscribers having unlimited access, while non-subscribers have a limited number of queries each day. It can be accessed through the web, iOS, Android, and Mac.
About the Author
