I use Perplexity almost every day for research. I’m impressed with its capabilities. However, Google’s new AI assistant, Deep Research, has caught my attention. It’s available with the Gemini Advanced subscription. Deep Research gives you a hands-on way to build and change multi-step research plans. This raises a question: should I switch?
In this blog post, I’ll compare Google Deep Research and Perplexity. I’ll look at their features and performance. Then, I’ll decide which tool is best for different research needs. I’ll also offer my final thoughts on both tools and how they fit into the broader research landscape. Let’s dive in!
Round 1: Search Efficiency – Speed and Flexibility in Research
When I conduct research, speed and flexibility are paramount. I need answers fast, without a lot of back-and-forth. It’s also important to easily refine my research direction. I asked both tools to look up a simple topic: responsible AI. I asked both platforms to identify key trends in this field.
Google Deep Research:
-
Research Plan: Deep Research creates a thorough and thoughtful research plan. It includes valid questions, many of which I hadn’t thought of in my first prompt. The editing experience could be more intuitive, though. I’d prefer to directly edit the plan rather than submit another prompt.
-
Search Operators: Deep Research found it hard to use search operators. For example, they wanted to search only for PDF files. I had to explicitly state my requirements in the prompt.
-
Speed: The research process is slower, taking around 8-10 minutes to fully load a response.
-
Exporting: You can directly export to Google Docs. It comes with detailed citations, so it’s ready for more changes. This integrates seamlessly with other Google products like NotebookLM.
-
Direct Response: Perplexity skips the research plan and pulls resources directly based on the prompt.
-
Search Operators: Perplexity handles search operators effectively, immediately updating results upon refinement.
-
Speed: Response time is significantly faster.
-
Source Management: Perplexity allows you to uncheck unwanted sources, which Deep Research lacks. I like how Perplexity numbers sources in key points. It makes it easy to trace information.
-
Follow-up Questions: Perplexity provides follow-up questions to aid research, but they can be too generic at times.
Round 1 Verdict: Both tools show good research efficiency, catering to different preferences. Perplexity beats Deep Research in speed. It also has more flexibility to adjust search direction and refine scope.
Round 2: Source Reliability, Depth, and Insight Quality
In research, the sources must be reliable. The information should be deep, and the output needs to be of high quality. We must trust our sources and ensure they are current and reputable. The output should also be relevant and useful for our projects. For this round, I chose the trending topic of AI agents. I will focus on key players, their capabilities, implementation methods, and real-world applications.
Google Deep Research:
-
Source Reliability: Deep Research used trusted sources like Salesforce, IBM, and Zapier. I noticed a lack of diversity. There was limited representation from media outlets, discussion forums, academic sources, and YouTube. While most sources were up-to-date, the concentration on service providers felt limiting.
-
Deep Research gave a detailed response. They covered each question in the research plan. Their answer had a clear flow and included specific examples with data.
-
Output Quality: While mostly informative, some parts leaned too much on single sources. They sometimes used keywords from articles not directly linked to AI agents. This raises concerns about potential bias and accuracy.
Perplexity:
-
Source Reliability: Perplexity showed a wider range of sources. This included tech media like TechCrunch and Yahoo Finance. It also included social media sites like YouTube and LinkedIn, plus community forums. The sources were also timely and reliable.
-
Information Depth: Perplexity’s responses were more condensed and high-level compared to Deep Research.
-
Output Quality: Perplexity’s output felt more meaningful and less reliant on buzzwords. The information was clearer and more useful, especially regarding capabilities and implementation. It also consistently used multiple sources for each section, mitigating potential bias.
Round 2 Verdict: Deep Research had more depth, but Perplexity stood out. Its diverse sources and better output quality made it the more valuable tool in this case. Deep Research relied on single sources and made some mistakes. This affected its performance in this round.
Round 3: Context Retention and Cross-Referencing
Effective research needs you to keep context and link different sources meaningfully. To evaluate this, I posed follow-up questions related to the AI agent research topic.
First Follow-Up Question: I asked both tools how to measure AI agent performance. I wanted them to link these metrics to business outcomes based on the use cases provided.
-
Deep Research effectively maintained context and included specific examples from the original report. The suggested metrics were clear and connected to business impact. They included first call resolution and average speed to answer. The response time, however, remained slow, even for follow-up prompts.
-
Perplexity: The metrics from Perplexity weren’t as clear and didn’t link well to earlier use cases.
Second Follow-Up Question:
I asked both tools to compare their claimed abilities with the first results and early user feedback. This helped identify gaps and contradictions between marketed features and real-world performance.
-
Deep Research: Deep Research summarized overstated capabilities and organized the information clearly. However, the analysis was general, lacking specific examples. The user feedback sources were outdated, further limiting the analysis.
-
Perplexity: Perplexity provided a clear response. It pointed out gaps and contradictions using case study examples. I didn’t see user feedback, but I was impressed that it checked many sources for each follow-up prompt.
Round 3 Verdict: Deep Research kept context better, especially in the first follow-up question. Perplexity takes the win this time. Its strong cross-referencing and detailed analysis stood out in the second question.
Wrap Up and Final Thoughts
Deep Research and Perplexity both provide useful research tools. They provide trustworthy sources and answer questions effectively. However, it’s still key to fact-check.
Deep Research Strengths:
-
Detailed and comprehensive initial research plans.
-
Strong context retention throughout the chat.
-
Direct export to Google Docs with formatted citations.
Deep Research Weaknesses:
-
Lacks flexibility in refining the search plan and direction.
-
Slow response times, even for follow-up prompts.
-
Output quality can vary. It may depend too much on single sources or big brands, which can cause bias.
-
Limited source diversity, especially lacking in media, forum, and academic sources.
Perplexity Strengths:
-
Faster response times and greater search efficiency.
-
Easy to adjust search plans and narrow scope.
-
Diverse sources, including media, forums, and social media.
-
Strong cross-referencing ability.
-
Flexibility to switch between AI models, including specialized options like O1 and DeepSeek.
Perplexity Weaknesses:
-
Context retention could be improved.
-
Initial responses can be less comprehensive than Deep Research’s initial research plan.
My Overall Verdict: For my research needs, Perplexity stands out as the better choice. Its efficiency and varied sources make it a strong tool. Deep Research has great potential, especially for academic tasks. It’s useful for creating documents that need many citations. It needs to get faster, be more flexible, and include more sources to reach its full potential.
The Future of Deep Research: I think Deep Research would be more useful if added as a feature in Google Search, like Perplexity’s Pro Search. This would be a useful addition to Google’s tools, not just another search engine.
Both tools work well with other research tools, such as NotebookLM. Choosing between Deep Research and Perplexity depends on your research needs and preferences. I suggest trying both to see which one suits your workflow best.
Read Also: