Image source: generated by Unbounded AI

The iteration speed of large models is amazing.

From the emergence of technology to the full spread of commercialization, AI search is undoubtedly one of the best starting points for large models to land on the C side: it not only carries the pioneering role of large language models after they come out of the industry, but also responds to users There is an urgent need for efficient information acquisition.

Especially in China, where the number of Internet users reaches 1.092 billion and search engine users account for 827 million in the huge market (CNNIC data, 2023), AI search undoubtedly stands at the center of the traffic trend.

But this is just the beginning.

In the world of traditional search engines, search methods based on keyword matching are relied on, by indexing web content and ranking search results according to the relevance of search terms.

The main disadvantages of this kind of search are: information overload, insufficient relevance, advertising interference, varying quality, insufficient timeliness, lack of personalization, poor interactivity, difficulty in balancing depth and breadth, and inaccurate semantic understanding. Poor visualization and insufficient structuring.

As a result, the process of obtaining information is time-consuming and laborious for users.

Large language model (LLM) Chatbot is good at generating answers based on existing knowledge base. If the question involves the latest developments in the real world, LLM may not be able to provide the correct answer.

For example, GPT3, although it has shown great potential in natural language understanding and generation, is trained based on offline data and lacks real-time performance. It cannot extract the latest information from the Internet or answer questions in a timely manner. event issues.

It can be said that the emergence of AI search engines has almost perfectly solved these problems.

The core of the AI ​​search method lies in three steps: asking questions, filtering, and integrating answers.

In other words: AI search combines the retrieval capabilities of traditional engines and the intelligent question and answer capabilities of large AI models, setting off a revolution in real-time search and quick question and answer. In fact, it is not only an improvement in efficiency, but also a reshaping of traditional interaction logic.

Although currently AI search may still face challenges such as "illusion problems" (generating wrong information) and insufficient real-time performance. But the efficient and personalized experience it provides is enough for users to begin to re-experience the meaning of "search".

The market game of AI search: the blooming of a hundred flowers and the rise of the ecosystem

The product form of AI search is rapidly expanding with the needs of different scenarios.

From professional fields to general platforms to embedded functions, AI search has become a new application entrance to the Internet. It can be roughly divided into the following four categories:

Category 1: Vertical field AI search

Provide deeply customized professional search services for specific industries or disciplines. For example, Delifasou and MetaLaw in the legal field; Phind and De in the IT fieldvv; Consensus in the scientific research field; Reportify in the investment research field; and Xiaotian AI for agricultural applications. These tools leverage domain expertise to greatly improve information quality and delivery efficiency.

The second category: independent AI search platform

The most popular ones are independent platforms such as Perplexity, Felo, 360AI, Mita AI, Tiangong AI, Bocha AI, ThinkAny, and You. , with AI search as the core, exploring multi-functionality. The highlight of these platforms is the combination of technology and user experience, constantly attracting users through innovative features, and trying to challenge the status of traditional search engines.

The third category: Comprehensive AI question and answer products

Most of the products in this category have gradually added AI search modules from LLM chatbot, such as ChatGPT4’s networking mode, Zhipu AI, Kimi , Wenxinyiyan, iFlytek Spark, Doubao, etc., embed AI search as a functional module into existing large language model applications, gradually enriching user interaction experience.

Category 4: Embedded AI search function

Traditional Internet and social platforms such as Baidu, WeChat, Zhihu, Douyin, Quark, Bing, etc. are also quickly using AI search as a New means of retaining users. For example, WeChat's built-in smart assistant and Zhihu's AI question and answer mode are all optimizing content distribution and user retention through AI search.

Even grass-growing platforms such as Xiaohongshu have begun to be gradually penetrated by AI search. It can only be said that no matter how slow social platform companies are today, they cannot escape the express train of big models, and AI search has become the best landing portal to further gather and retain users, and reduce jump steps and exit probability.

This extensive application layout not only proves the adaptability of AI search, but also means that the competition in the future market will be more brutal.

How to make money with AI search?

The market is going up and down, and the survival of the fittest is becoming more and more serious. Everyone is worried. In addition to "drawing cakes" to obtain financing, what else can AI search do? make money?

Like traditional search engines, advertising will be one of the main sources of revenue for AI search in the foreseeable future. For example, the currently popular Perplexity, according to US media, has planned to introduce an advertising model on the search platform, is negotiating cooperation with brands such as Nike and Marriott, and plans to launch a "sponsored" question (brand-related search answer) model.

Ads will be sold for more than $50 per thousand impressions, aiming to tap into the $300 billion digital ad market dominated by Google.

As users’ usage habits become more and more biased toward AI, advertising services are expected to become the business path for many domestic AI search providers, completely replacing traditional engines.

The second is the subscription service launched by many AI search engines, providing advanced Pro functions.It is also a relatively common way to make money.

Perplexity.ai currently offers a $20 per month Pro service subscription, which gives users access to more advanced model and image generation capabilities. As of August 2024, the company's annualized revenue (full-year revenue calculated based on sales in recent months) has grown from $5 million in January to $35 million.

The third is to provide customized enterprise solutions, this type of business model has also been put into operation. Perplexity launched the Perplexity Publisher Program, partnering with publishers such as TIME, Der Spiegel and Fortune to share revenue from content interaction.

Perplexity also plans to help media and content creators build AI assistants on its platform and provide customized services.

Bing, as part of Microsoft, also provides Bing Search API for enterprise customers, allowing enterprises to integrate Bing's search functions into their applications and charge based on usage.

The fourth is model authorization and sales. Through API authorization or direct sales of models, the AI ​​search platform can help other companies quickly build AI assistants. This "technology export" model is not only a profit point, but also an important means to expand ecological influence.

From information to transactions, AI search may revolutionize Xiaohongshu’s life?

The business model is about to be completed, and the various functions of AI search are gradually diversified and enriched, becoming more comprehensive.

Looking at perplexity’s Discover, space, Libray and other functions, how do they look more and more like Xiaohongshu’s planting grass? We need to add a little live broadcast function to bring goods, but you can’t tell the difference?

Just the day before yesterday (November 19th), Perplexity Launched a new AI shopping assistant called "Pro Buy".

Perplexity CEO Aravind Srinivas said on social platform

They also provide users with the innovative “Snap to Shop” AI search tool, which allows users to ask relevant questions by taking photos of products.

Similar to Google Lens, Taobao Photo Search, and Xiaohongshu Photo Search, users can search by uploading pictures and find matching purchase options even if they don’t know the name of the product.

In addition to “Snap to Shop”, Perplexity alsoThe "Buy with Pro" function has been launched, allowing users to quickly obtain information about the products they want to buy using AI search on the Perplexity website or app, and seamlessly view specific products from specific merchants.

All orders purchased through "Buy with Pro" receive free shipping.

When users search for products, Perplexity will automatically aggregate and analyze product reviews from multiple e-commerce platforms to better meet users' query needs.

In addition, Perplexity has launched a "Merchant Program" that allows large retailers to directly integrate their product data, thereby increasing their chances of appearing as "recommended products" and having the opportunity to participate in one-click checkout systems.

Is this going to completely transform from a platform that provides information and answers to a platform that supports native business transactions? So has AI search decided to get involved in e-commerce sales?

Further imagine, when AI search is combined with e-commerce live broadcasts, and when "sponsorship issues" gradually replace traditional advertising forms, will content platforms such as Xiaohongshu and Douyin feel huge competitive pressure?

In the future, will AI search become the new overlord of content distribution and e-commerce transactions?

These questions are worth pondering.

As AI search functions continue to diversify, its boundaries have extended from simple information retrieval to direct business transactions.

The new product form has subverted it again - content is the transaction, and what you see is a closed loop.

Looking back, they have completely subverted the positioning of traditional search engines; looking forward, they are still trying to break into the market of "selling goods" platforms, and AI search has gradually evolved into a collection of information and recommendations. A comprehensive platform integrated with transactions...

Finally

The rise of AI search is the result of two-way drivers of technological progress and user needs.

From subverting traditional search logic to exploring new business models, its potential is limitless. However, the premise of all this is the continuous improvement of technology and the continuous optimization of user experience.

It is foreseeable that in the next few years, AI search will continue to expand in depth and breadth, eventually forming an intelligent search ecosystem covering all scenarios.

The endgame of AI search may not be a simple tool, but a new paradigm for the interaction between humans and knowledge.

This revolution has entered its half-time.