The recent launch of Kimi's exploration version with AI autonomous search capabilities by Moon's Dark Side has reignited the competition in the AI search arena, a field that, while not new, is experiencing a significant resurgenceIn the latter half of this year, the buzz surrounding AI search has reached new heightsOutside of China, OpenAI has introduced an AI-powered search engine called SearchGPT, while Google has declared its intentions to launch AI-organized search result pages, further accelerating the transformation of search technology.

Amidst a flurry of activity, numerous AI startups are emerging, particularly those focusing on AI searchNotable players include Perplexity and Genspark, founded by former Baidu CEO Jing Kun, as well as firms like Glean and Hebbia that are focusing on enterprise search, along with Daydream, which is tailored for the e-commerce sectorAdditionally, in October, AI Pin's two executives announced a new venture called Infactory, an AI fact-checking search engine that has successfully secured $4 million in seed funding.

In China, virtually every significant player in the AI market, whether a large corporation, medium-sized firm, or startup, has introduced AI search products

Recently, Baidu rebranded its Wenxin Yiyan project to Wenxiaoyan, positioned as the new face of searchAccording to data released by "AI Product Rankings" in September, Chinese AI products like Zhihu Zhidao and 360 AI Search are showing impressive growth trendsIn the medium to long term, if AI search evolves further into intelligent assistants, it has the potential to redefine the internet as a new entry point, presenting both prospects and financial opportunitiesIn the short run, however, this high-frequency necessity remains a race between traditional and AI search, with no clear distinctions between the twoBoth act as transitional forms of intelligent assistance or supplementary features of AI-native products, representing considerable imaginative potential.

Despite the heightened enthusiasm in the industry, many insiders remain cautiously optimistic about this latest wave of AI search

• The Challenges Behind Hikvision's Peak Amid Layoff Turmoil
• Semiconductor Equipment Enters a Period of High Prosperity
• A-share Market Shifts into Rapid Rotation Phase
• Institutional Demand Fuels Dividend ETF Growth
• Robust U.S. Economic Data

Rather than indulging in romantic notions of 'search being everywhere,' practical concerns regarding costs, data availability, user experience, and viable business models have become increasingly visible.

The rekindling of interest in AI search among global tech giants and startups is not coincidentalA representative from a major tech corporation's search division explained, “Search has always been a battleground for AI industry players.” For one, it offers a chance for large corporations currently struggling with traffic to redefine and exploit new entry points, leveraging their strengths in the mobile ecosystemAdditionally, the multifaceted applications of search have highlighted significant inefficiencies in traditional search methods, thereby offering startups opportunities to progress in niche areas.

Since last year, three main categories of players have surged into the AI search field: first, those that create native AI search tools targeting both consumer and business ends, with companies like Mitata AI, Perplexity AI, Genspark, YOU, and Glean leading the charge, predominantly among startups

Second, established search engines that are integrating AI functionalities into traditional tools—companies such as Microsoft with New Bing and Google’s AI Overview, as well as 360’s AI Search—demonstrate a trend marked by anxietyLiang Zhihui, vice president of 360 Group, has pointed out that the impetus to launch AI search services earlier was largely driven by concerns: “When AI search fully matures, our traditional search engines and browsers could potentially be completely disrupted.” Finally, there are enterprises enriching existing AI products by incorporating search functionalities, such as Kimi and Baidu’s Wenxiaoyan.

However, unlike the previous surge of excitement, the current players are revitalizing 'old bottles' for several substantive reasonsThe first significant shift is the evolution in user search behaviorSearch is fundamentally a scenario of generalized demand

Liang Zhihui revealed in the AI Product Manager podcast that in 2019, only 42% of users were seeking addresses and 21% resources in 360 Search, whereas 37% were posing questionsBy 2024, user patterns had dramatically shifted; the proportion of searchers seeking addresses or resources had declined into double digits, while over 70% switched to inquiry-driven searchesThe explosive rise of generative AI technology has naturally fostered a greater focus on user questions.

The second notable change revolves around the easing concerns of leading tech companiesThey have gradually come to realize that traditional search and AI search are not in a zero-sum game of replacement; the habitual practices of users endureIn the face of new AI search tools, users might initially take them for a test drive due to novelty, but the significant switching costs in the long run—coupled with the inherent advantages of traditional search in terms of platforms, ecosystems, and content—lead many to revert back once comparable AI functionalities are introduced.

The third factor is that the previous wave of enthusiasm for AI search heavily benefitted from advancements in Retrieval-Augmented Generation (RAG) technology, which effectively merges the two steps of information retrieval and content generation

This integration has significantly reduced the hallucination issues prevalent in large models, leading to a balance between technological capabilities and practical requirements.

As demand for technological services grows, industry competition is transitioning toward engineering and product developmentThis involves optimizing model capabilities, including prompts, along with the utilization of search resources and proprietary technologies to unearth new forms and experiences in AI searchOne insider aptly noted, “Outwardly, this wave of AI search seems hot, but we remain level-headed about it.”

As the imagination surrounding search integrates into the reality of technology, phrases like “forget about the search bar; search will be omnipresent” are often referencedThe essence of search, after all, is the user's quest for specific information—not the act of searching itself, which is merely a pathway to answers

Each query represents a unique fusion of personal need and intention, and when paired with the generative capabilities of AI models, it continually meets diverse individual demands.

This perspective elucidates why Aravind Srinivas, the founder of Perplexity, does not perceive his company as a competitor to Google: “Our most significant adversary is not Google, but the fact that people inherently struggle to formulate precise questions.” Currently, the focal undertaking for AI search is to leverage new technologies, products, and interactions to accurately grasp user inquiries and search purposes.

In essence, a shared understanding among industry players is that AI search is moving away from an idealistic technology-driven romance towards a more pragmatic approachThis shift is underscored by three prominent changesFirstly, the entry barrier for AI search is progressively rising.

For instance, DayDream completed a $50 million seed round in June while Genspark followed suit with $60 million in July, achieving a valuation of $260 million

Moreover, Exa AI—an AI search engine company backed by Nvidia—has also announced a new round of funding totaling $17 millionThe rising entry barriers stem primarily from increasing costs associated with market entry.

Many AI search products today are built by integrating traditional search engine APIs with generative models, rather than establishing a fundamentally new AI architectureThe core reason lies in the prohibitive costs associated with AI search development.

From a cost perspective, the expenses surrounding AI search fall into model costs (including large model interface fees, infrastructure, training, and operational expenses), API fees, private data storage, and service costs related to building an index library, along with labor and customer acquisition costsA key barrier for AI search hinges on the quality and quantity of data, which directly affects the quality of output generated

Traditional search giants, for instance, incur consistent fixed costs associated with their dataset indexing libraries—a considerable financial undertaking involving hundreds or even thousands of servers for global web indexingOpenAI, too, has previously sought to bridge its data gaps through acquisitions like the database firm Rockset.

This awareness underscores why AI search firms are increasingly being valued higherCost remains one of the most pressing challenges for startups as they seek to navigate this burgeoning market.

The second noteworthy shift is that focusing solely on search is insufficientFu Sheng, chairman and CEO of Cheetah Mobile and Orion Star, remarked, “While AI search is currently in harmony, nuanced distinctions reflect different product strategies.” Whether bolstered by tech giants or leveraging unique ecosystems, players with existing search operations are beginning to integrate diverse AI capabilities and examine new channels to ensure that search remains pervasive.

One salient distinction lies in the evolution of AI capabilities: Baidu’s vice president, Xue Su, articulated that AI search diverges from the purely tool-centric “search” to a more human-centric synthesis of “search + create + chat,” incorporating numerous familiar features such as multi-dimensional summaries and long-form text synthesis.

In terms of tapping into various access points, the strategy includes unifying multiple hardware and multi-modal entries

For instance, Quark recently launched a new PC version, while Google partnered with Samsung to roll out the “search as you go” function—all aimed at discovering new entry points for enhancing search utility.

Furthermore, outside the major players, vertical-focused AI search startups are refining their approaches to deliver more customized user experiencesGenspark has focused on niche industries like travel and product searches, utilizing an AI copilot for custom page creation named Sparkpage, allowing users to edit and personalize content akin to the style seen on platforms like Xiaohongshu.

Likewise, Mitata Search, popular among researchers and students, recently unveiled podcast, document library, and image analysis features, further broadening its information sources and core user serviceAddressing the needs of its target user group remains paramount.

The third change is that more players are acknowledging the commercial aspect

As noted, commercialization remains a prevalent dialogue in AI searchFollowing the launch of their exploration version, Moon's Dark Side hinted at plans for future commercialization attemptsThe primary revenue models for AI search could include subscriptions or advertisement placementsRecently, Google announced a merging of its AI search capabilities within Google Lens, enhancing its shopping feature while incorporating advertising placements into the shopper's experienceLikewise, Perplexity AI is exploring advertising revenue models and reported handling 340 million queries in September, attracting interest from several major companies keen on placing ads within its platform.

In summary, artificial intelligence-powered search has not reached its conclusionAravind Srinivas previously remarked that two industries exemplify a robust “technology-product-user” cycle: autonomous driving and search