(Yicai) July 12 -- As more tech firms roll out artificial intelligence-powered search engines, there are still a number of hurdles, such as gaining the trust of users and finding a profitable business model, before they can fully replace traditional ones, market insiders said.

Alibaba Group Holding’s search engine Quark launched a one-stop AI search service on July 10 and transformed into a ‘super search box.’ And search engine giants, such as Baidu, 360 and Sogou, already have AI search functions available.

But it is unlikely that AI search engines will completely replace conventional ones, the experts say.

There is still a long way to go before AI searches gain the trust of users, said Shen Yang, a professor at Tsinghua University’s School of Journalism and Communication as well as its College of AI. Users need to provide better prompts in order to generate more in-depth content from AI search engines. And they also need to be able to judge the accuracy of the feedback.

There is also still a big time gap, said Huang Chaoqing, a senior AI product manager and consultant to two companies in the US and China. It takes AI search engines three seconds to generate results, while traditional search engines take milliseconds.

Besides, how to introduce a profitable business model without affecting the core user experience is a long-term challenge for AI search engine developers.

It is not easy to get users to pay for services because Baidu’s search engine and other large language models are mostly free of charge, Huang said. Also, AI search engines' ad-free feature may lead to drops in the ad revenue of conventional search engines such as Baidu and 360.

More people are using AI searches, though. Kunlun Tech’s Skywork AI search engine has more than one million daily active users, the company said in May. And Meta Sota’s Metaso AI search engine has more than 7.2 million monthly visits, the Shanghai-based company said in March.

“AI searches make it a lot cheaper for people to ask questions, reducing the costs by one-thousandth of those in the past, which makes people ask questions more proactively,” said Wu Hanqing, founder of KMind, a developer of personal AI computers and former chief security scientist at Alibaba Cloud, earlier this month.

“Previously, when people asked a question on online question-and-answer communities, it could take two to three days before someone would answer it. But it only takes kFind, a smart search engine, three seconds to offer quite a good answer by aggregating information available on the web,” he added.

Editor: Kim Taylor