The "online library" - 360doc Personal Library (hereinafter referred to as 360doc), which has been running in the digital wave for 20 years and carries the memories of netizens' youth, will cease services on May 1. On April 13, a reporter from "Daily Economic News" (hereinafter referred to as "Daily Reporter") discovered that its official website had stopped updating content, but the login, data backup, VIP refund application, and wallet withdrawal functions were all open normally.

On the same day, every reporter visited Beijing Liuzhi Information Technology Co., Ltd. (hereinafter referred to as Liuzhi Information), the parent company of 360doc. The door of the office where it was registered was closed, and the reporter did not see any staff through the window. When the reporter arrived at the office of Liuzhi Information at around 2 p.m., he saw that there were only two employees in the huge office area.

"Facing the wave of the AI ​​(artificial intelligence) era, my small team and I are no longer able to provide it with the resources it needs to move into the next twenty years." At about 10 pm on April 13, Cai Zhi, the founder of 360doc, said in an interview with reporters that he has been very busy recently and has been dealing with user follow-up matters. In her view, 360doc is a centralized product born in a specific period of the Internet, and its model will be difficult to "resurrect" as it is in the future.

The departure of 360doc is not only a farewell to an Internet product, but also a profound question about the "next stop" of personal knowledge base.

In the past two decades, users have been accustomed to collecting and saving good articles to their personal libraries. However, the AI ​​​​era has brought new possibilities to make the collected content "alive" and become an AI personal assistant that truly understands you. When AI personal assistants become the next main battlefield for global technology giants, who among giants such as Google, Microsoft, Tencent, Byte and Alibaba can take the lead in seizing the next high ground?

On-site visit: The venue was deserted and the transfer failed

The end of 360doc

When you open the official website of 360doc, the first thing you see is an "important notice": the service will be stopped on May 1st, and users are reminded to back up articles and process refunds and withdrawals before that date.

Screenshot of 360doc personal library official website

On April 13, every reporter visited the registered address and actual office location of Liuzhi Information.

At about 11 o'clock in the morning that day, the reporter first came to the registered office of Liuzhi Information in Haidian District, Beijing. The door of the site was closed tightly. Looking inside through the glass window, no staff were seen in the room. At around 2 p.m., the reporter arrived at the company's actual office address in Chaoyang District, Beijing, and saw that the overall area was slightly deserted.

During the more than an hour the reporter was at the scene, he saw only two employees. In response to reporters’ questions about the company’s closure and follow-up business arrangements, the other party did not respond.

In fact, as early as the end of January this year, 360doc announced that due to the company's business adjustment, it decided to transfer all platform assets (core technology, data and operation team) of 360doc Personal Library to partners with sincerity and ability for free.

But two months later, this free transfer ended in failure. In this regard, Cai Zhi admitted that she had had in-depth communications with a number of potential takeovers: among them there were traffic management teams who intended to dismantle assets and quickly realize cash; there were industry veterans who calculated the benefits of data; there were also Internet giants who had the possibility of strategic synergy but were silent because no one wanted to take responsibility.

"Some have to be dismantled and sold, some have to be 'raised' and then sold, and some have to be integrated with AI and sold." Cai Zhi lamented that he regards 360doc as his "daughter", but in this era, data is more valuable than people, training is more profitable than services, and AI is more important than users.

Questioning the times: 100 million users, 20 years of accumulation

Why is it difficult to resist the impact of the AI ​​era?

It is undeniable that the AI ​​wave has impacted 360doc to a certain extent.

"The departure of 360doc is a typical case of the change of times, but it cannot be simply attributed to AI. It has been established for more than 20 years and has nearly 100 million users. It is a typical knowledge container from the Internet 1.0 to 2.0 era." Liu Xingliang, a well-known digital economist and president of the DCCI Internet Research Institute, said frankly in an interview with Mijing reporters.

360doc was launched in 2005. Cai Zhi recalled in a letter to users that in 2005, he was at home on maternity leave and taking care of his newborn child while typing the first line of 360doc code on the computer.

Every reporter noticed that in the PC (personal computer) era, 60% to 70% of 360doc’s traffic came from search engines such as Baidu, and at its peak, advertising revenue accounted for more than 60%. Relying on traffic and advertising, Liuzhi Information's revenue grew rapidly from 2016 to 2017, reaching nearly 40 million yuan at one time, and successfully landed on the New Third Board.

But now, with the sweep of large models and agents such as ChatGPT, Kimi, Deepseek, and OpenClaw, the traffic of the "360doc Personal Library" website has dropped significantly, and the business continues to suffer losses.

Financial report data shows that from 2023 to 2024, Liuzhi Information's operating income was 10.039 million yuan and 8.088 million yuan respectively, and the net profits attributable to shareholders of the listed company were -2.447 million yuan and -1.963 million yuan respectively. In the first half of 2025, the company achieved revenue of approximately 2.518 million yuan, a year-on-year decrease of 43.84%; the net profit attributable to shareholders of the listed company was approximately -4.183 million yuan, a year-on-year decrease of 585.92%.

In Liu Xingliang's view, the dilemma faced by 360doc may be twofold: First, its core value is storage and organization, rather than understanding and use; second, its business model has not been established for a long time and is inherently valuable, but it cannot make money.

"In the past, you had to collect, sort and search, but now you can just ask AI directly. But it does not mean that AI has defeated 360doc, but that the knowledge service has moved from 'save' to 'use' and to AI assistant." Liu Xingliang said.

Zhang Yi, founder of iiMedia Consulting Group, believes that this is a microcosm of the transformation of traditional knowledge services into the intelligent era. "The traditional static document library model itself has long-term structural problems such as copyright and monetization, and AI has only accelerated its elimination. The current AI search dialogue tools have actually greatly diverted user needs. Users have shifted from looking for documents to directly asking for answers. The low-interaction centralized model of simple storage lacks core barriers."

Discussion on the way out: "Resurrection" as it is is impossible

Decentralization is the way to break the situation

Facing the wave of AI, Cai Zhi confessed to every reporter that she had also thought about and imagined the future of 360doc.

In Cai Zhi’s view, the core asset of 360doc is not the 1.1 billion articles, but the relationship between users and content. The data behind these relationships were formed before AI generated content on a large scale, and are “native human data” that are not contaminated by AI.

When discussing the characteristics of "personal knowledge base" products that can truly prosper in the future, Cai Zhi said that user data sovereignty must be respected and a sustainable business closed loop must be formed.

"AI, localization, and social collaboration are just means, not ends. Users need to be sure that their knowledge will not be abused or used to train unknown models; the platform also needs to find a survival model that does not rely on advertising and data sales. Otherwise, no matter how advanced the technology is, it will not go far without the trust of users," Cai Zhi said.

What countless fans are most concerned about is, with the help of AI, can 360doc be "resurrected" in some form in the future?

"The 360doc model is a centralized product born in a specific period of the Internet. I think it is difficult to 'resurrect' it as it is." Cai Zhi said.

She bluntly told reporters that the environment of the times, user habits and business models have completely changed, and it is impossible for people to step into the same river twice. "In the future, there may be new knowledge base forms that are decentralized, controlled by users, and based on privacy computing, but that is already a brand new thing and is no longer the original 360doc."

Cai Zhi said bluntly that AI can help you organize, associate, and even generate new content, but it cannot replace your own process of screening, thinking, and internalization. “The core value of a personal knowledge base is the ‘private cognitive system’, which is the unique imprint left after you interact with information.”

Industry Outlook: Giants' "stuck" AI knowledge base

Super personal assistant can be expected

At the time of the shutdown of 360doc, the personal knowledge base track is facing the most intense "stuck war" in history - but the main battlefield of this war is no longer "storage", but "understanding".

The outside world is familiar with Google Gemini, but Google's real bet on personal knowledge base products is NotebookLM, which has been promoted in a low-key manner. This product has recently been launched simultaneously on the mobile terminal and is connected with the Gemini application. After users import PDF, web pages, audio and video and other materials, the system can convert them into "living knowledge" that can be spoken, summarized, and output as podcasts and slides. What deserves more attention is its "anti-illusion" mechanism - the AI ​​only answers based on the data uploaded by the user. Each answer comes with original text traceability, which points directly to the core pain point of the older generation of knowledge platforms "existing but not using". Microsoft's approach is geared toward enterprise scenarios. Copilot Notebooks integrate notes, documents, and meeting recordings into OneNote. On this basis, AI continues to refine insights and can even generate two-person voice overviews, extending knowledge management from individuals to team collaboration.

The logic on the domestic battlefield is different. Tencent's IMA Intelligent Workbench, Byte Doubao, and Alibaba Quark all embed knowledge base functions into larger AI portal products, competing for user stickiness and data accumulation rather than pure knowledge management experience.

An IEEE survey shows that 52% of technology leaders believe that "personal assistant/schedule manager" will be the AI ​​application most needed by consumers in 2026.

"The AI ​​era does not eliminate the need for knowledge bases, but rather reconstructs knowledge bases." Liu Xingliang further pointed out to every reporter that when the form of knowledge evolves from libraries to intelligent agents, "what we need is no longer a static library, but a 'library' that can talk."

Zhang Yi also told reporters that the next generation of knowledge services will become a personal "second brain" that can help users achieve a complete closed loop from passive collection to active understanding, assisted decision-making, and creation. This is a direction worth looking forward to.

At present, with the continuous iteration of technologies such as large language models, multi-modal interaction, and breakthroughs in terminal-side computing power, AI personal assistants have completely broken away from the limitations of early "voice command tools" and evolved into "intelligent partners" that are deeply integrated into all scenarios of life and work.

Liu Xingliang also observed that the core essence of the competition among giants is the competition for access, data, and execution capabilities. It is worth noting that with the proliferation of AI information and content, users’ trust costs are also rising.

"Products that can survive in the future need to have at least these keywords: first, AI native; second, 'privatization + controllability', because users are paying more and more attention to data security; third, long-term memory, which is no longer just about generating instant answers, but can continue to understand your data for years or even decades; fourth, executability, which no longer simply outputs results, but can perform tasks." Liu Xingliang said.

He predicts that a super AI assistant may appear in the next three years. This AI assistant must not be a simple chat robot, but a digital clone that can "understand me, remember me, and do things for me."