At the moment when the wave of AI is surging, countless products have sprung up, but many are dying quickly. This article takes stock of the recently discontinued AI products and analyzes the reasons for the suspension of these products for your reference.
Tell AI, you are about to stop operating, what kind of suicide note will AI leave?
DeepSeek will write a farewell letter to the user, without forgetting to add bullets; Yuanbao will cosplay himself as a koi, trying to splash a splash in the palm of the user’s hand to leave a thought; The most anthropomorphic bean bun presents a neat and standardized and emotionally restrained small composition, politely saying goodbye.
These imaginative scenes are in stark contrast to the cruelty of AI product shutdowns in reality, they are often born in the hustle and bustle, but say goodbye silently in silence, leaving only users on social media, “My favorite AI, why did you run away?” Shocked and regretful.
We took stock of the recently discontinued AI products, from the fist applications of star startups, to the testing software of Internet giants trying to lead the industry, to countless small team applications struggling in subdivisions such as AI companionship and AI tree holes.
They were either acquired and integrated due to strategic adjustments, or forced to shut down due to commercialization difficulties. These “new residents of the AI cemetery” have a common characteristic: “come quickly, go fast”.
01 New residents of the AI cemetery
If you ask on Xiaohongshu how to say goodbye to information anxiety, then someone will definitely recommend MiniMax’s “Everything Tracking”.
Positioned as an AI information aggregation and personalized content steward, Everything Tracking has received widespread praise on social media for its efficient information processing capabilities and intelligent recommendations during the public beta phase in early 2025.
What does a product manager need to do?
In the process of a product from scratch, it is not easy to do a good job in the role of product manager, in addition to the well-known writing requirements, writing requirements, writing requirements, there are many things to do. The product manager is not what you think, but will only ask you for trouble, make a request:
View details >
Users can use it to search and organize global Internet information 24 hours a day. However, despite the good reputation of users and the forward-looking product functions, with the increasing pressure of commercialization and the high consumption of computing power and other resources by the product itself, this product that had high hopes was soon discontinued.
Another similar product is “Bubbling Duck”, as a role-playing agent product, after a year of market competition, Bubble Duck has gradually lagged behind competing products in key data such as the number of C-end daily active users.
According to reports from Intelligent Emergence, as Leapfrog Xingchen shifts its strategic focus from C-end applications to more basic model research and development and ToB/ToC dialogue product Step AI from the end of 2024 to the beginning of 2025, the Bubbling Duck team was merged, and the product only retained a small number of employees for operation and maintenance, essentially entering a state of shutdown.
These leading AI startups are helpless about adjusting their products, while Internet companies are eliminating AI products more quickly.
Xiaohongshu, known as the “App Store” of AI software, has not only provided promotion channels for many AI applications, but also underwent frequent adjustments to its incubated AI functions and products. According to incomplete statistics, since 2023, Xiaohongshu has successively taken off at least three AI-related projects:
Trik Al (AI drawing application), removed from the shelves due to copyright issues. Xiaohongshu’s group AI assistant and AI voice chat are said to have stopped service due to factors such as content security and compliance.
In the AI wave, in addition to leading companies and Internet giants, there are more small and medium-sized entrepreneurial teams and independent developers. They often have limited resources and face greater pressure to survive.
In the crowded AI companion track, products that Character.Al benchmark are emerging one after another, but there are few successes. At the beginning of 25, “Phantom AI” and “Paper Airplane AI” were discontinued one after another.
Choose the niche track “momoAI Emotional Companion”, which is an AI tree hole app. The team had updated the version in April 2025, but suddenly stopped operating at the end of May, catching many users who were used to talking here by surprise and looking for alternatives on social platforms.
New residents of these AI cemeteries, regardless of their background, often exhibit some common characteristics: they come and go quickly.
“One month to build, three months to die” has become the norm for entrepreneurship in the AI industry.
A large number of AI products, especially at the application level, have a very short time window from launch to disappearance. Users often lament that “it is gone as soon as it is easy to use”, and even some products fail or stop operating shortly after public beta or small-scale promotion.
02 Success is AI, failure is AI
The specific reasons why entrepreneurial projects come and go quickly in the AI era are complex, but they are inseparable from one sentence: AI technology is inclusive, making entrepreneurship “easy”.
As the price war of model manufacturers enters a fever pitch, and the optimization of large model algorithms and model distillation technology become more and more mature, the once unattainable AI infrastructure has become as popular as hydropower.
In 2023, the ChatGPT call fee will be $20/million tokens, and by 2025, the cost of equivalent capability models such as Google Gemini-1.5-Flash will drop to $0.07/million tokens, a huge drop.
With the sharp decline in the cost of using large models is the evolution of development tools, UI design can be handed over to AI, code can be handed over to AI, copywriting can be handed over to AI, and technical barriers to software development have disappeared.
As long as you have an idea, even a technical novice, you can quickly iterate an AI product prototype that can run and chat.
However, the most important thing in entrepreneurship is not to make a product, but also to consider multiple factors such as technological development, market competition, business model, capital dynamics and user needs.
When basic technical capabilities are no longer the main bottleneck, the ability to truly discover market pain points, propose effective solutions, and commercialize them becomes a more scarce resource.
In the AI tool aggregation website “DANG! There is an AI Graveyard project that records AI software that has been suspended overseas. As of June 13, 2025, 1,267 AI software has been suspended, and the largest proportion of them are AI writing applications.
The characteristics of AI writing applications do not need to be elaborated too much, which is very in line with the technical characteristics of large models, but the convergence of the capabilities of the underlying large models, the product differentiation is not high, and it is easy to be imitated, and it is very easy to fall into low-level involution.
Looking back at the new residents of the AI cemetery mentioned above, most of them are AI companion applications, which are in the same predicament as AI writing applications: they rely on the underlying large model and are easy to imitate.
These products all have common problems: they are just simple imitations of existing successful models, lack of real innovation, and do not serve the needs of users well.
AI has greatly lowered the threshold for developing and releasing a product, but it has also lowered the threshold for developers to shut down these products. This is not destined to be a trend that users are looking forward to seeing.
