Search engines hadn’t changed much in a very long time — until AI search entered the scene. Perplexity is one of the artificial intelligence platforms trying to reshape how we find answers online by skipping the list of links and delivering direct, conversational results.
Unlike traditional Google Search, which sends you off to other sites, Perplexity tries to be the site. You can use it on the web or download its free app for Android, iOS, Mac and Windows.
Perplexity AI was founded in August 2022 by Aravind Srinivas (formerly a research scientist at OpenAI), Denis Yarats (with experience at Facebook AI Research), Johnny Ho (ex-Quora engineer) and Andy Konwinski (a Databricks co-founder).
In a short time, it has grown from a scrappy startup into a fast-evolving generative AI tool with pro features and an agentic AI shopping assistant that can make purchases on your behalf. It’s part search engine, part chatbot and it’s changing how people search for information.
Perplexity has become especially popular among students, researchers and tech-savvy users who want fast, reliable answers without having to dig through multiple sites. It has amassed 22 million active users across its website and app, while its mobile app has been downloaded 13.9 million times since its launch.
Read on to find out what exactly Perplexity is, how it works, what sets it apart from ChatGPT and other generative AI tools — and about some of the legal battles it’s been facing.
What is Perplexity?
Perplexity calls itself the world’s first “answer engine” that uses large language models to understand and give you direct, detailed responses to your questions. It searches the web in real time and, instead of showing you a list of blue links, it pulls info from what should be trusted sources (though Reddit often finds itself on that list) and summarizes the information into clear, up-to-date answers in natural language.
The company also encourages you to double-check the sources, which, considering how often we have seen AI tools hallucinate, should be your default practice by now.
Interestingly, Perplexity was the first generative AI tool whose search results came with citations. Now, ChatGPT, Google’s Gemini and even Claude (because it can access the internet) are adding citations with deep research features.
As OpenAI and Google push their premium research tools behind paywalls, Perplexity has joined the deep research race with a more accessible option. The feature breaks down complex questions into smaller subtasks, pulls from academic papers, technical sources and long-form articles and compiles it all into a detailed report, often in just a few minutes. So it’s slightly faster at generating results than most competitors.
Rob Howard, an AI consultant and founder of the education platform Innovating with AI, thinks highly of these advancements.
“Deep research is the first tool that really changes my job,” Howard told me. “The core idea of ‘go read 600 articles in 10 minutes and give me the highlights and then link me to them’ is just incredibly valuable for so many people. And it’s kind of like having an army of interns or an army of research assistants working with you, for a very low price.”
What can Perplexity do?
Perplexity answers questions like “What is quantum computing?” and “What’s the best time to visit Japan?”
You can follow up with additional questions or refine your query like you would in a chat. Similarly to Google’s “People also ask” section when you search, you can expand on the topic with an offered set of questions under the Related section at the bottom of the page.
CNET’s staff chose Perplexity as the best chatbot for research, which is what most people use it for, but it can also:
- Summarize long articles or documents.
- Explain complex topics in plain language.
- Offer comparisons and pros/cons lists.
- Perform internal knowledge search on uploaded files like PDFs, Word docs and Excel sheets (Pro and Enterprise users only).
- Look up real-time stock prices and financial data (data is sourced via Financial Modeling Prep API).
If you’re a Perplexity Pro subscriber in the US, the “Buy with Pro” feature handles your shopping request from start to finish. It selects a product based on your preferences, completes the purchase using your saved details and, for a limited time, even offers free shipping.
Perplexity also introduced a voice assistant for iOS, offering hands-free answers at a time when Apple was still holding off on its own AI features.
Another new release is Perplexity Labs, an agentic AI feature that lets you build full projects — think reports, dashboards and spreadsheets — just by typing out what you need in plain language. It handles everything from researching online to writing and running code, creating charts and organizing the results in one place.
Labs is available to Pro users on Perplexity’s mobile and web apps, with desktop support coming soon. While Deep Research is built for faster, well-cited answers, Labs is better suited for longer, more complex tasks where the goal is to produce something usable, not just informative. Most projects take about 10 minutes or more to complete.
How is Perplexity different from ChatGPT and Google Search?
Traditional Google Search crawls the web and ranks results. ChatGPT can generate content and answer questions based on its training data and real-time web results. Perplexity combines both approaches: It retrieves current information from the web and uses AI to summarize it. Similarly, Google’s AI Overviews now show AI-generated results above standard search results.
Perplexity, as previously mentioned, is known for a consistent and prominent display of source citations for every piece of information it provides. Other AI tools will give citations on demand.
Another distinguishable feature is that other gen AI applications prioritize open-ended conversation or creative content generation, like images and videos. Until recently, Perplexity was unable to do that, but since the end of April, it can now create images.
you can generate images on perplexity now. the UI is cute and fun. we have also added support for grok 3 and o4-mini for model selection options (which already supports gemini 2.5 pro, claude 3.7, perplexity sonar, gpt-4.1, deepseek r1 1776), and looking into supporting o3 as… pic.twitter.com/RX6L98pf2g
— Aravind Srinivas (@AravSrinivas) April 25, 2025
I tried creating something simple and fun to test it, and it turns out that images in Perplexity and ChatGPT are very similar, despite ChatGPT’s dog having five legs (ha). It comes as no surprise, given Perplexity uses OpenAI’s image generation models similar to those used by ChatGPT.
