Perplexity launches Ask, its answer engine

Perplexity, a startup founded a few months earlier by Aravind Srinivas, Denis Yarats, Johnny Ho and Andy Konwinski — engineers with backgrounds at OpenAI, DeepMind, Meta and Quora among others — launched Ask, a public web tool combining search retrieval with a language model that answered questions in prose, with inline numbered citations back to its sources, rather than returning a ranked list of links. The company called it an “answer engine” rather than a search engine, a distinction it repeated in later interviews as its core positioning against both conventional search and the newly launched ChatGPT, which answered fluently but, unlike Ask, cited nothing and could not browse the live web.

The product’s mechanism — retrieving web pages at query time and grounding a language model’s answer in what it found, with the sources shown alongside the answer — was an early form of what the industry later termed retrieval-augmented generation, not yet a phrase in wide use. This distinguished it from ChatGPT at launch, which drew entirely on knowledge encoded during training.

Ask attracted little mainstream press coverage at the time, arriving in the same weeks that ChatGPT’s launch was dominating coverage of AI, but it established a product category and much of the interface convention — cited, synthesised answers rather than links — that Perplexity built on through repeated funding rounds over the following years, en route to a $9 billion valuation two years later and a series of copyright disputes with publishers, including Dow Jones and the New York Post, over how its answers used their content.