Narrated by Aravind Srinivas, CEO of Perplexity edge Many people use chatbots like regular search engines. It makes sense to provide information about the things they are looking for (e.g. restaurants) directly from the source. So when people ask for restaurant recommendations, it integrates Yelp’s maps, reviews and other details into their responses.
“Our fundamental motivation is that we started as this tool to attract people to fact-check and do research. But research can be for anything, like if you’re a coffee addict, you want to know where are the good coffee shops ,” Srinivas said.
as edgeDavid Pierce recently noted that how Perplexity mixes text and links when answering questions is already a selling point compared to other chatbots like Copilot, Gemini or ChatGPT. Srinivas said that consideration was part of the reason the company ultimately chose Yelp. Srinivas said Yelp’s data allows Perplexity to offer multiple ways to provide answers, such as adding maps, photos, and even citation reviews.
These responses include links to the business’s Yelp page so users can view more reviews or other details. Perplexity did not disclose financial details of the deal and said it would not use Yelp’s data to train any models, as the company uses existing models such as GPT and Claude 2 for its chatbots.
The connection is part of Yelp’s Fusion program to license data to third parties, including BMW. Chad Richard, Yelp’s senior vice president of business and corporate development, wrote in a letter to edge.
Perplexity is the latest AI company to strike a deal with a constantly refreshing source of fresh information, though Srinivas noted that the company is not working with Yelp but rather licensing its data. Google will reportedly pay Reddit $60 million per year to access real-time data and use Reddit data to train its artificial intelligence models. OpenAI has signed agreements with news publishers such as Axel Springer to obtain training data.
Other chatbots can recommend cafes and restaurants, but most of this data comes from information models learned from the internet, almost as if it were searching for users. Some chatbots benefit from having access to services that already aggregate enterprise information. For example, Google’s Gemini can provide a list of restaurants when prompted, with details like opening hours, menus, and maps because it can be pulled from Google Maps.
Perplexity plans more such integrations in the future, and it already works with answer engine WolframAlpha for mathematical calculations. Srinivas offered some potential examples, such as job providers of shopping data or financial data.