With the launch of Deep Research Providing to even Free Subscribers, Perplexity AI changed the Game
The landscape of research and analysis is undergoing a significant transformation, driven by the rapid advancements in Artificial Intelligence (AI). Across diverse fields, from scientific discovery to market analysis, AI is increasingly being leveraged to augment human capabilities, offering new avenues for information gathering, processing, and interpretation. A key development in this evolution is the emergence of sophisticated AI models equipped with what are often referred to as “deep thinking” or “deep research” functionalities. These advanced systems promise to move beyond the limitations of traditional search engines and basic AI chatbots, offering the potential to conduct in-depth investigations and generate comprehensive insights with unprecedented depth and efficiency.
Among the prominent players in this evolving space are Perplexity AI with its Deep Research feature, Google with its Gemini family of deep thinking models, and OpenAI with its ChatGPT platform powered by advanced GPT models.
Methodology and Capabilities
Perplexity AI introduced its Deep Research feature in February 2025, positioning it as a tool designed to significantly reduce the time required for in-depth research and analysis. At its core, the functionality of Deep Research involves conducting numerous searches, reading and processing information from hundreds of sources, and employing reasoning to synthesize the material into a comprehensive report. This entire process is designed to be autonomous, typically taking between two to four minutes to complete, a timeframe that contrasts sharply with the hours a human expert might require for similar tasks.
The methodology employed by Deep Research involves an iterative research process coupled with reasoning capabilities. The AI dynamically refines its research plan as it uncovers new information, mirroring the way a human researcher might adjust their approach based on emerging insights. This feature is equipped with both search and coding capabilities, allowing it to not only find relevant information but also to process and analyze data programmatically when necessary.
The underlying technology driving this systematic exploration of complex topics for Perplexity AI is a proprietary framework known as “test time compute (TTC) expansion”. This framework enables the AI to dissect queries, perform numerous web searches, evaluate a multitude of sources, and synthesize findings using probabilistic reasoning models.
Once the AI has thoroughly evaluated the source materials, it proceeds to the report writing phase, where it synthesizes all the gathered research into a clear and comprehensive report. The final output offers users flexibility in terms of how they can utilize and share the information, with options to export the report as a PDF or a standard document. Additionally, the report can be converted into a Perplexity Page, facilitating easy sharing with colleagues or friends.
Initially launched on the web platform, Perplexity AI has announced plans to extend the availability of Deep Research to iOS, Android, and Mac platforms in the second quarter of 2025.
Use Cases and Applications
Perplexity AI designed Deep Research with the aim of empowering a broad range of users to conduct expert-level analysis across various complex subjects. The tool is particularly adept at generating work artifacts in domains such as finance, marketing, and technology. Beyond these professional applications, Deep Research is also intended to serve as a valuable personal consultant in areas like health, product research, and travel planning. The versatility of the tool extends across numerous industries, with potential applications in academic research, journalism, market research and competitor analysis, legal research, medical research, thesis writing, and even personal learning on a variety of topics.
For business users, the tool offers potential applications in competitive analysis, enabling the comparison of financial performance, market strategies, and analyst sentiments to develop comprehensive competitor profiles. It can also aid in identifying emerging market trends and innovations, providing companies with a strategic advantage in adapting to market shifts quickly.
Performance Benchmarks
Perplexity AI has highlighted the performance of its Deep Research feature on several industry benchmarks. Notably, it achieved a 21.1% accuracy score on Humanity’s Last Exam, a comprehensive benchmark consisting of over 3,000 questions across more than 100 subjects. This score reportedly outperforms Gemini Thinking and several other leading models, suggesting a strong capability in handling complex, expert-level knowledge. Additionally, Deep Research demonstrated a 93.9% accuracy on the SimpleQA benchmark, which tests for factuality, again exceeding the performance of many leading models.
Beyond accuracy, Perplexity AI also emphasizes the speed of its Deep Research tool, stating that it completes most research tasks in less than three minutes. In a direct comparison test, Deep Research on Perplexity was approximately nine times faster than ChatGPT Pro while also referencing 150% more sources. Aravind Srinivas, CEO of Perplexity, has noted that Deep Research’s performance on the Humanity Last Exam benchmark is quite close to that of OpenAI’s o3 model, despite being significantly faster and more cost-effective.
6. Cost Comparison: Perplexity AI Deep Research vs. Google Gemini vs. OpenAI ChatGPT
6.1 Perplexity AI Deep Research Pricing Structure
Perplexity AI stands out as it offers a freemium model for its Deep Research feature. It is available for free to all users, albeit with a limitation of five queries per day for non-subscribers. For users who require more extensive use, Perplexity AI offers a Pro subscription priced at $20 per month. This subscription provides unlimited access to Deep Research queries. Additionally, the Pro subscription includes other benefits such as access to advanced language models like GPT-4 and Claude 3.5, API access, the ability to search within internal files, and higher daily limits for regular search queries.
Comparative Table: Features and Costs Summary
| Feature/Aspect | Perplexity AI Deep Research | Google Gemini (Deep Research/Ultra) | OpenAI ChatGPT (GPT-4/Advanced Data Analysis) |
| Functionality | Autonomous research, reasoning, and report generation | Agentic research assistant, web browsing, report generation | Text generation, data analysis, web browsing via plugins/built-in |
| Key Capabilities | Iterative research, fast report generation, broad topic coverage | Native multimodality, long context window, audio overviews | Multimodality (GPT-4o), advanced data analysis, plugin ecosystem |
| Intended Use Cases | Finance, marketing, tech research; personal consulting | Competitive analysis, due diligence, topic understanding | Content creation, data analysis, specialized research tasks |
| Information Retrieval | Dozens of searches, hundreds of sources, speed emphasis | Browses hundreds of websites, emphasizes relevance and recency | Web browsing (paid tiers), specialized searches via plugins |
| Depth of Analysis | Mimics human research, strong on factuality benchmarks | Shows reasoning, critical evaluation, complex reasoning models | Advanced reasoning (GPT-4), in-depth data analysis (Advanced Data Analysis) |
| Output Format | Clear reports, PDF/document/Perplexity Page export | Detailed reports, audio overviews | Primarily text, visualizations and downloadable files (data analysis) |
| Ease of Use | Web-based, mobile apps planned, free tier available | Web and mobile apps, requires Google One AI Premium | Web and mobile apps, advanced features require paid subscription |
| Free Tier | Limited to 5 queries per day | 1-month free trial of AI Premium | Free plan with limited access to advanced features |
| Subscription Cost (Individual) | $20/month (Pro, unlimited Deep Research) | $19.99/month (Google One AI Premium) | $20/month (Plus), $200/month (Pro) |
| API Access | Yes (Pro subscription) | Yes (usage-based, varies by model) | Yes (usage-based, varies by model) |
Export to Sheets
Choosing the Right AI Research Tool
Perplexity AI Deep Research, Google Gemini’s deep thinking models, and OpenAI ChatGPT’s deep thinking models each offer unique strengths and cater to slightly different user needs in the realm of AI-powered research. Perplexity AI Deep Research distinguishes itself through its speed and its freemium model, providing a relatively accessible entry point for users seeking quick insights across a broad range of topics. Its performance on benchmarks like Humanity’s Last Exam suggests a strong capability in handling complex queries, and its emphasis on the number of sources consulted could be appealing to users prioritizing breadth of information.
Ultimately, the choice of which AI research tool is most suitable depends on the individual user’s priorities. Users who prioritize speed and a free entry point might find Perplexity AI Deep Research compelling. Those who value multi-modality, integration within the Google ecosystem, and innovative features like Audio Overviews could lean towards Google Gemini. Researchers who require a flexible platform with a strong focus on data analysis and access to a wide range of specialized tools through plugins might prefer OpenAI ChatGPT.
As the field of AI-powered research continues to evolve rapidly, staying informed about the latest advancements and evaluating these tools based on specific research needs will be crucial for maximizing their benefits.
You can access the deep research here: https://www.perplexity.ai/
For more information on deep research: https://www.perplexity.ai/hub/blog/introducing-perplexity-deep-research