Highlights:

  • Bing Generative Search directly responds to the AI Overviews feature introduced in Google Search in the U.S. in May.
  • As per Microsoft, Bing Generative Search leverages AI to match content and design search results dynamically. This new layout aims to understand user intent better and deliver more relevant information.

To respond to Google LLC’s experimental AI-powered search experience, Microsoft has introduced Bing Generative Search in a limited preview.

The company stated that the new feature integrates Bing’s foundational search results with the capabilities of large and small language models (LLMs and SLMs). This allows it to comprehend user queries and analyze millions of web sources to deliver immediate summaries and relevant information on the topic.

According to Microsoft, Bing Generative Search “dynamically matches content and generates search results in a new, AI-generated layout to more effectively fulfill the user’s query intent.”

The launch was announced by Jordi Ribas, Microsoft’s Head of Engineering and Product for Copilot and Bing, who described it as a “meaningful step forward” in the advancement of AI-driven search.

Bing Generative Search is a direct answer to the AI Overviews feature introduced in Google Search in the U.S. in May. Google explained that AI Overviews aimed to deliver quick answers for users who lack the time to sift through numerous articles for the information they need. It seeks to summarize the information the user is searching for by gathering the most relevant insights from various websites.

Microsoft’s offering functions similarly. In his blog post, Ribas provided examples to demonstrate how it works. For instance, if a user searches “What is a spaghetti western?” Bing will give information on the genre’s history and origin and highlight some of the most well-known films in that category. Additionally, it will display a list of links and sources from which the information was gathered.

Similar to Google’s AI Overviews, users can opt to dismiss the generative AI responses and summaries and view a traditional search results page instead.

Microsoft has undoubtedly taken extra precautions to avoid the controversy that accompanied the launch of Google’s AI Overviews, which initially made headlines for unfavorable reasons.

The new service was reported to generate a series of nonsensical, inaccurate, and sometimes potentially dangerous responses to user queries. In one instance, AI Overviews suggested a user “mix non-toxic glue with the cheese” to prevent it from sliding off a homemade pizza. In another instance, when someone inquired about the best way to clean a washing machine, AI Overviews provided a recipe for deadly mustard gas, along with instructions on how to make it.

Google isn’t the only AI search provider to have faced embarrassing moments. For example, an alternative service called Arc Search made headlines for confidently telling a news reporter that cut-off toes will eventually grow back. At the same time, Genspark cheerfully recommended a list of weapons for effectively killing someone.

Another risk Microsoft will want to avoid is the issue of lack of attribution, which troubled a competitor search engine called Perplexity. According to Forbes, Perplexity plagiarized articles from sites like Bloomberg, CNBC, and Forbes in its responses, failing to provide proper credit or attribution links.

Microsoft will also need to ensure that its new search experience doesn’t negatively impact smaller websites. A major concern with generative AI search is that it might divert traffic away from the websites from which it sources information. A study by Raptive Inc. earlier this year found that Google’s AI Overviews feature could potentially reduce traffic to publishers by about 25%, as its summaries often eliminate the need for users to visit the source articles.

Ribas assured that Bing Generative Search would not have the same effect, stating that the company is carefully monitoring how generative AI search impacts website traffic. He claimed that the new experience has managed to sustain website clicks. However, he did not provide any specific statistics or studies to support his assertions, instead referencing “early data” that has not yet been made public.