Skip to Main Content

AI: Artificial Intelligence Resources

Information and resources for Artificial Intelligence applied to academic environments

What is Perplexity?

Perplexity is a tool created to help streamline the research process. When given a prompt perplexity will use AI Technology to search the internet for real time information to analyze and summarize into a concise natural language text. By doing this, Perplexity helps users find relevant scholarly articles, generate insights, and navigate complex research topics with ease. 

 Its core features include a powerful search engine that retrieves academic literature and a context-aware citation tool that clarifies the impact and relevance of sources. Perplexity also provides a list of sources it has used to create its response 

Additionally, Perplexity AI offers a summarization feature that distills large volumes of information into concise, comprehensible summaries, making it easier to grasp key concepts and findings. Ideal for both initial explorations and in-depth research. 

Perplexity Deep Research

In February 2025, Perplexity added a new search mode: Deep Research.  It is based on a similar chain-of-thought based agent as other tools offered by Google's Gemini, OpenAI's Deep Research, and DeepSeek Deep Research.  For researchers, it differs from these products in two important ways:

  • First, one can access it with a free-tier account (with a limited number of searches per day).
  • Second, one can use Perplexity's "Focus" feature to prioritize academic sources (taken from Semantic Scholar and PubMed, primarily).

Searches take about five minutes, and go through a long chain of intermediate searches before an LLM integrates the results into one three-page report.  The report usually draws on text from fewer than a dozen sources, but the final output will return all the websites and documents Perplexity Deep Research discovered along the way, which can sometimes be the most relevant to the user's needs.

The report will be broken down into numerous sections and subsections, and clicking on the title of such sections and subsections will start a new Deep Research search on that particular topic.  One can also ask follow-up questions, and otherwise use it like a chatbot.

Like most such tools, one should expect the report to prioritize the "consensus" rather than innovative, possibly untested new research that might overturn the consensus.  These reports are probably most useful at the outset of researching a particular topic, and discovering seed papers to read and discover subsequent work from.