Microsoft Bing is getting a new “Deep Search” feature powered by OpenAI’s GPT-4. The feature is designed to give users more relevant and comprehensive answers to complex search queries. Microsoft notes that Deep Search is not a replacement for Bing’s existing web search, and is instead an enhancement that allows for deeper exploration of the web.

The new feature, according to Microsoft in a blog post, builds on Bing's current web index and ranking system and enhances them with GPT-4, which takes the search query and turns it into more of a comprehensive description of what the results should include.

For example, assume a user is researching the existence of loyalty programs in other countries and types in the query "how do points systems work in Japan." Deep Search would take the query and expand it into the following

Describe the various loyalty card schemes operating in Japan, including the advantages, requirements, and disadvantages of each. Explain the popular kinds of loyalty cards such as convenience stores, supermarkets, and restaurants. Compare and contrast the pros and cons of using loyalty cards versus other modes of payment in Japan, including current rewards and benefits. Highlight the most popular services and participating merchants.

With this longer description, you can describe your intent more accurately than you could with just a few words.

If your search query is slightly fuzzy, Deep Search will find all possible intents and builds a long description for each of them. Deep Search then shows the intents to you, so you can pick the correct one.

Once the expanded description has been created, Bing will pull relevant results that often don't show up in typical search results, Microsoft says. Deep Search finds pages that might match the expanded query, rewrites the query and then searches for those variations too.

Sticking with the same example above for loyalty points query, Deep Search may also search for things like "loyalty card programs Japan," "best loyalty cards for travelers in Japan," "comparison of loyalty programs by category in Japan," "redeeming loyalty cards in Japan" and "managing loyalty points with phone apps."

"By doing this, Deep Search can find results that cover different aspects of my query, even if they don't explicitly include the original keywords," Microsoft wrote in the blog post. "Regular searches on Bing already consider millions of web pages for each search and Deep Search does ten times that to find results that are more informative and specific than the ones that rank higher in normal search."

Once Deep Search collects web pages, it ranks them according to how well they fit the expanded description. When ranking, Deep Search observes how well the topic fits the query, whether appropriate details, how trustworthy the page is and how new and popular it is.

Microsoft notes that Deep Search is optional and can take up to 30 seconds to complete, which is why it's not meant for every query or user. For users who don't want more comprehensive answers, they can get regular Bing search results right away.