news

Microsoft Bing enters the AI ​​search war again. Can the perennial runner-up make a comeback?

2024-07-28

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

New Intelligence Report

Editor: Yongyong Qiao Yang

【New Wisdom Introduction】Microsoft Bing, the second largest search engine market player, has introduced AI search functionality that combines traditional search results with AI search results. Can it shake up the dominance of big brother Google Search this time?

With the continuous development of AI technology, the search engine market has become a battleground.

Google released its own AI search function, AI Overview, at the I/O conference held in May.

OpenAI released SearchGPT in the middle of the night, aiming to compete with Google for market share in search engines.

Following last year’s New Bing, Bing announced in its technology blog that it would open a new AI-generated search feature to a small group of people.

As of June, Google held 91.05% of the global search engine market, while Microsoft Bing had only 3.7%.

As early as when Bing launched New Bing, Google was a little panicked for a while, but in the end it had no actual impact on Bing's overall market.

Perplexity and other newcomers seem to be unable to compete with the dominant player. After a series of operations, Google found that its market share has basically not changed.

What's even more cruel is that Bing is the default search engine for all new Windows computers, but about 92% of users will switch the default browser to Google Chrome immediately after purchasing a new computer.

It seems that Bing is planning to make a comeback. Although it only has a 3.7% share, it is still ranked second!

But this second place may not be stable yet. Under the existing situation, Google's "dominant" position with more than 90% market share is unlikely to be overturned.

However, the 2nd, 3rd, and 4th places may change. Whoever develops faster in AI may see an increase in market share.

Because, from the perspective of major technology giants, AI search does have a huge market.

For example, Google donated a lot of money to Wikipedia, and 20% of its traffic falls on Wikipedia. Baidu has Baidu Encyclopedia, Q&A, and so on.

The emergence of big models may completely subvert the content loop and fixed search guidance.

Wikipedia and Encyclopedia both require manual editing, and the need for manual editing means that the maintenance cost is very high.

One of the focuses of AI search is high-quality content generation, so large models have the opportunity to generate search results such as Wikipedia on their own.

From the user's perspective, future AI search products will not necessarily be judged by which one has higher technological content, but rather how to make search results more faithful to web page content and how to make more systematic arrangements while ensuring reliable sources.

There are only two requirements: one is that the source is reliable and there are fewer advertisements; the other is that the integration is clear to save time and effort.

Search new layouts

Bing’s blog discussed the new layout and features of search, which not only make it easy for users to find information, but also make it easy for users to view AI search results and browse related websites.

In desktop view, Bing displays three panels:

- Table of contents on the left

- AI answers in the middle area (with links to website resources)

- Traditional search results are on the right

It can be seen that after careful design, Bing's layout is very different from the layout styles in other search engines.

First, looking at the table of contents, it has the main topics at the top and directly related subtopics below.

This is much better than “People Also Asked” type navigation because it invites users to explore and click on related search results to continue to search for deeper content.

Most notably, Bing combines the power of generative AI and large language models with its search results pages.

It understands search queries, reviews millions of information sources, and preserves the content of the original search interface;

It can also dynamically match content and generate search results using new artificial intelligence technology to meet the multi-faceted needs of user queries.

Also, an interesting feature of BingAI search is that it displays the answer to the initial question first and also predicts related questions.

This is similar to a technique called “information gain,” in which an AI search assistant first ranks the relevance of an initial set of pages that answer a search query.

Then, the second, third, and fourth sets of search results of related additional information of interest to the user are also ranked.

What makes Bing different from the information gain technique is that Bing combines the two steps into one on a single page.

Based on the directory on the left, users can easily click and go directly to other AI answers and view traditional search results.

For example, a user searches Bing for "What are Italian Westerns?"

As mentioned above, Bing displays the AI-generated answer in the middle section, which explores the film genre in detail, including its history, origins, and well-known film examples, with source citations and links to further reading.

On the right side, regular search results continue to be highlighted on the page as before.

This is an important step for Bing to improve the search experience. It significantly improves the quality of search results while allowing users to strike a balance between the search methods they are accustomed to and the new AI search technology.

Bing is slowly rolling out this feature and will take time to gather feedback, test, and learn before making it more widely available.

Bing, a win-win for users and publishers

Not only on the user side, Bing's layout also takes into account the demands of the supply side.

Publishers are concerned that AI search tools will provide complete answers based on the original news content, eliminating the need for users to read the original article, resulting in a sharp drop in original clicks.

In other words, it is common that "secondary creation" content is more popular than the original, but it is just an AI-created version.

The most typical example is that last month, Perplexity used a Forbes report in its search results without mentioning the source in a small line of text until the bottom of the page.

As a result, Perplexity’s CEO received a letter directly from Forbes and took legal action to protect his rights against this infringement.

Bing said it will closely monitor how AI-generated searches affect publishers’ clicks and traffic.

The resulting search experience is designed with this in mind, including retaining traditional search results and increasing the number of clickable links, such as clear link references in search results, etc.

Early data suggests that the new interface and new features launched by Bing can maintain the website's traffic and meet the requirements for building a healthy online ecosystem.

Microsoft stressed that Generative Search is developed on the basis of a healthy web ecosystem.

It also said that although preliminary data showed that website clicks would remain unchanged, it remained cautious as to whether these results could be confirmed in a wider promotion.

after all, there is such a precedent, and Microsoft hopes to avoid this situation.

On the other hand, Microsoft will also strictly control search results.

Especially controversial answers and search results, especially since Microsoft has already generated enough negative headlines for its quick introduction of Copilot, such as political issues.