Names, versions and owners change fast. Checked October 2026.
In plain words
Perplexity is an app and website from a San Francisco company of the same name. It calls itself an answer engine. Instead of a list of blue links, you get a written answer, and every claim carries a small number you can click to see where it came from.
It is not a model of its own in the way GPT or Claude are. Perplexity searches the web, picks out useful passages, and hands them to a large language model to write the reply. Paying users have been able to choose which company's model does the writing, or use Perplexity's own.
This look-up-then-write method is called retrieval-augmented generation. It was launched in December 2022. The company has since added other products, such as a web browser with the assistant built in; the box above lists them.
Everyday example
A librarian who brings back marked pages
You ask a librarian: when is Hampi least crowded? They do not answer from memory. They pull three guidebooks, read the relevant pages, and come back with a short reply, with a slip of paper in each book marking exactly where each point came from. Perplexity does the same with web pages, in seconds.
Where the example breaks: A librarian chooses trusted books and reads them with care. Perplexity relies on what its search finds and how the model summarises it, so an answer can still misread a page or lean on a weak source. The numbers make that easy to check.
Latestas of October 2026
- Products
- Search, Deep Research, the Comet browser (free to download since October 2025) and Computer, which runs multi-step tasks
- Models
- More than 15 models from several companies, plus Perplexity's own Sonar models
These details change often. The rest of this page explains what it is, which stays true.
How it actually works
Perplexity has not published every detail of its system, but answer engines like it follow the same broad pipeline.
Turn the question into searches
The system turns your question into one or more search queries and runs them against the web, and in some modes against news, papers or files you upload.
Pick the best passages
From the pages it finds, it selects the passages most likely to answer the question. This is the retrieval step, which often uses embeddings to judge relevance.
Write with the sources in view
Those passages are placed into the language model's context window along with your question. The model writes an answer and tags each claim with the number of the source it used.
Follow-ups keep the thread
You can ask a follow-up, and it searches again with the earlier conversation in mind. A longer mode, similar to deep research, runs many searches and writes a report.
- Made by
- Perplexity AI, San Francisco, founded August 2022 by Aravind Srinivas, Denis Yarats, Johnny Ho and Andy Konwinski
- Launched
- 7 December 2022
- Models used
- Its own Sonar models plus models from other companies, which paying users can choose
- Other products
- Comet, an AI web browser (July 2025)
Several news publishers have sued Perplexity over its use of their articles, as part of the wider debate on AI and copyright.
Questions people ask
Is Perplexity the same as ChatGPT?
No. Both write answers with large language models, but Perplexity is built around searching the web for every question and showing numbered sources. ChatGPT is a general assistant that can also search when needed. They come from different companies.
Is Perplexity a search engine?
It is a search tool that writes the result for you. It searches the web like a search engine, then summarises what it found into one answer, with links to the pages used. It calls this an answer engine.
Can Perplexity be wrong?
Yes. It can misread a source, pick a poor one, or blend details incorrectly. Because every claim is numbered, you can click through and check. For anything important, read the original source.
Who founded Perplexity?
Aravind Srinivas, Denis Yarats, Johnny Ho and Andy Konwinski founded it in August 2022, with Srinivas as chief executive. The company is based in San Francisco.