The predicament and solution of the AI toy industry: From the technological frenzy to the return of product-oriented thinking

       From 2024 to 2026, AI toys rapidly transformed from a niche topic to a hot industry term. Whether in tech media, the investment circle, or consumer electronics exhibitions, they were frequently mentioned. We witnessed a large number of new products emerging - talking plush toys, interactive robots with large models, so-called "intelligent companions" that claim to accompany children's growth, and digital pets that can interact with family members... Financing news was coming one after another, new product launches were continuous, and major exhibitions couldn't do without them.

       However, beneath this hustle and bustle, an unsettling reality gradually emerged: Despite the continuous emergence of products, consumers' reactions were far less enthusiastic than expected. For many, these AI toys brought more immediate experiences of "novelty", "strangeness", and "fun", but it was difficult to turn them into long-term purchasing motivations. Was it because the technology was not yet mature enough to support practical value? Or was the underlying model's ability insufficient to touch people's hearts? Or was it due to insufficient market education, preventing consumers from understanding their significance? These explanations have their own merits, but if we delve deeper, we will find a less pleasing but more to-the-point conclusion - the core bottleneck of the current AI toy industry is not funds, algorithms, or market perception, but the lack of a truly "product manager".

       Here, "product manager" does not merely refer to a position in an enterprise's organizational structure, but rather a responsibility role, a way of thinking that questions and guards the existence reasons of the product. We need someone who can firmly answer: Why do we make this product? What specific and real needs does it solve? Is this need truly there, or is it just our assumption? At the same time, someone is needed to make wise choices and restrain themselves within the boundaries of AI's capabilities, rather than unrestrainedly adding functions. If these questions are not seriously considered and answered, then no matter how advanced the model is, how attractive the promotion is, such a product is essentially just a "technical experiment", rather than a product that can be truly accepted by users.

       During the observation of the industry, we found a common phenomenon: The development paths of most AI toys start from "what we can do", rather than "what the users need". For example, seeing that AI can have natural language conversations, they immediately think of chat toys; when the model has memory functions, they promptly launch "growth companions"; when emotion recognition technology is introduced, it is quickly packaged as "emotional companions". Once technology emerges, the product is hastily pieced together, as if the existence of technology itself is the entire reason for the product. But few people ask in reverse: Without the aid of AI, does this need still exist? If the AI's capabilities can only reach the passing grade, does this product still have independent value? Under this mindset, there will naturally be a large number of "logically sound but difficult to implement in reality" products on the market.

       More fatal is that many AI toys cannot even clearly define "who the users are". In product introductions, "users" are often described as an extremely broad group: children, parents, young people, even the entire family. But the true product manager must make a difficult choice - who is the first user? Who has the final decision-making power and veto right for purchase? In what emotional state and scenario will this toy be used? Does it solve frequent practical problems, or is it merely an imagined need? If these basic questions are not seriously considered and clearly answered, then the failure of the product is almost inevitable.

       Another seriously overlooked misconception: Many practitioners believe that AI toys should accumulate as many functions as possible, the "smarter" the better. But in reality, AI toys need more restraint. Not all the things that AI can do are suitable to be incorporated into toys. For instance, the ability to have continuous conversations does not necessarily mean that they should keep talking indefinitely; the ability to learn users' behaviors does not necessarily mean that they should record everything; the ability to remember past interactions does not necessarily mean that they should accumulate data indefinitely. More importantly, very few people assess which functions might cause parents' concerns, which interaction methods might lead children to become dependent, and which "intelligent" functions should be actively turned off in specific situations. These issues are not merely technical challenges; they are manifestations of product judgment.

       Therefore, when we look back from the 2025 milestone, we can't help but ask: Who are the true decision-makers of AI toys? It is not the CEO of the company, not the underlying large models, and not even entirely the end users. The true decision-makers should be those who are willing to be responsible for the needs themselves - they need to be responsible for "what not to do", for "why this product should not be made now", and for "what range of 'intelligence' capabilities should be limited". These issues are not merely technical problems; they are manifestations of product judgment.

       Over the past two years, the development speed of the AI toy industry has far exceeded many people's expectations. The technology has become increasingly mature, costs have gradually decreased, and product forms have become increasingly diverse. When an industry reaches this stage, what is most urgently needed is no longer to discuss "can it still be done", but to seriously think about "how to make it a real product". In this sense, when AI toys move towards 2026, what they urgently need is not just more advanced algorithms and richer functions, but to return to product thinking, so that each product can clearly answer: For whom is it made, what problem does it solve, and why is its existence irreplaceable.