AI News AI资讯 7h ago Updated 2h ago 更新于 2小时前 42

FIFA mistakenly issues '0-yuan tickets' for World Cup, urges fans to pay 误发世界杯“0元票” 国际足联催球迷补票款

After Douyin Doubao introduced paid subscriptions, its monthly active users decreased by 6.1 million. This number is like a splash of cold water on the dream of "commercialization implementation" that large model players are eager to discuss. The situation is simple: Doubao, an AI assistant owned by ByteDance, was previously free, and everyone enjoyed using it. Now, with the introduction of paid plans, over six million users have voted with their feet and quietly left. This is far from being a m 豆包付费后月活减少610万。这个数字像一盆冷水,浇在了大模型玩家们热衷于讨论的“商业化落地”美梦上。事情很简单:字节跳动旗下的AI助手豆包,之前免费用,大家用得挺欢,现在一收费,超过六百万用户用脚投票,默默离开了。这哪里是什么商业化里程碑,这分明是一场尴尬的“信任测试”失败。

60
Hot 热度
70
Quality 质量
50
Impact 影响力

Analysis 深度分析

After Douyin Doubao introduced paid subscriptions, its monthly active users decreased by 6.1 million. This number is like a splash of cold water on the dream of "commercialization implementation" that large model players are eager to discuss. The situation is simple: Doubao, an AI assistant owned by ByteDance, was previously free, and everyone enjoyed using it. Now, with the introduction of paid plans, over six million users have voted with their feet and quietly left. This is far from being a milestone in commercialization—it is clearly an embarrassing failure of a "trust test."

All manufacturers claim that large models represent the future, productivity, and disruption. But the premise of "future" and "disruption" is that users truly believe you are worth the money—or, even more crucially, that they cannot do without you. The loss of 6.1 million users brutally reveals a reality: for a significant portion of early adopters, current AI assistants are essentially "nice to have but not essential"—an entertainment or light-use tool. The core value they provide is nowhere near enough to convince users to continuously pay out of their own pockets. When it was free, people were "just playing"; once payment was required, they immediately "played out."

Behind this is the collective growing pains of the entire industry shifting from a frantic "land grab" phase to a "cost-accounting" phase. Previously, all companies were burning money and subsidizing users—even at a loss—to acquire users, gather data, and train models. Reports painted a prosperous picture of "tens of millions in monthly active users" and "astonishing growth rates." But how much of that traffic came from "bargain hunters," and how much of it genuinely formed usage habits and reliance? Once payment was introduced, it was like the tide going out to reveal who was swimming naked. Doubao’s attempt effectively served as a costly stress test for the entire industry.

Manufacturers may argue that this is about exploring reasonable business models and a necessary step in market education. However, my point of contention is: isn’t this "education" a bit too hasty? Are the model’s capabilities truly stable enough to seamlessly support various paid scenarios? In professional fields like document processing, code generation, and complex reasoning, the "hallucinations" and unreliability of large models remain a Damocles sword hanging over everyone’s heads. Users are willing to pay for tools that are stable, precise, and solve real pain points—but they will be reluctant to continuously pay for an "intelligent companion" that is unpredictable and requires repeated manual verification and correction.

Look at the competition: OpenAI’s ChatGPT Plus and Claude’s Pro plans also face growth bottlenecks. Globally, the paid conversion rate for AI applications is a universal challenge. This is not just a pricing issue—it’s a matter of product strength and user mindset. When free alternatives (including open-source models and free quotas from other providers) are still abundant, where exactly is the "moat" for paid services? Is it just early access and a slight speed advantage?

Doubao’s setback might offer all AI manufacturers a chance to step back and reflect: stop celebrating model parameter races and high funding rounds. Return to the users themselves and solve those niche—but real and payment-worthy—"must-have" scenarios. Is it more precise industry report generation? More reliable automated office workflows? Or irreplaceable personalized creative production? Excelling at one thing to the point where users feel they "cannot do without it" is far more promising than building an "all-purpose but mediocre" chatbot.

The departure of 6.1 million users is not a rejection of AI—it is a rejection of a version they deemed "not worth paying for." After the technological revelry, the logic of business has never changed: value determines price, not the other way around. It seems the coming-of-age for large models will require a few more cycles of "charging—user loss—further improvement" to truly mature.

豆包付费后月活减少610万。这个数字像一盆冷水,浇在了大模型玩家们热衷于讨论的“商业化落地”美梦上。事情很简单:字节跳动旗下的AI助手豆包,之前免费用,大家用得挺欢,现在一收费,超过六百万用户用脚投票,默默离开了。这哪里是什么商业化里程碑,这分明是一场尴尬的“信任测试”失败。

所有厂商都在说大模型是未来,是生产力,是颠覆性的。但“未来”和“颠覆”的前提是,用户真的觉得你值这个钱,或者说,离了你不行。610万的流失,冷酷地揭示了一个现实:对于相当大一部分尝鲜用户来说,目前的AI助手,本质上还是个“有了挺好,没有也行”的娱乐或轻度工具。它提供的核心价值,远远没到能让人掏出真金白银持续订阅的程度。免费时,大家是“玩玩”;一收费,立刻“玩完”。

这背后是整个行业从狂热的“跑马圈地”向“算账”阶段切换时的集体阵痛。之前,所有公司都在烧钱补贴,用免费甚至赔本的方式抢用户、做数据、练模型,报告里一片“月活千万”、“增速惊人”的繁荣景象。但这些数据有多少是“羊毛党”贡献的,有多少是真正形成了使用习惯和依赖的?一收费,就像退潮后看到谁在裸泳。豆包的这次尝试,相当于是替整个行业做了一次代价不菲的压力测试。

厂商们或许会辩解,这是探索合理的商业模式,是市场教育的必经之路。但我想吐槽的是,这个“教育”是不是太急了点?模型能力真的稳定到能无缝支撑各种付费场景了吗?在文档处理、代码生成、复杂推理等专业领域,大模型的“幻觉”和不可靠性依然是悬在头上的达摩克利斯之剑。用户愿意为稳定、精准、解决实际痛点的工具付费,但很难为一个时灵时不灵、需要自己反复核对修正的“智能伙伴”长期买单。

看看隔壁,OpenAI的ChatGPT Plus、Claude的Pro方案,它们也面临增长瓶颈。全球范围内,AI应用的付费转化率都是一道难题。这不仅仅是定价问题,更是产品力和用户心智的问题。当免费的替代品(包括开源模型、其他厂商的免费额度)依然大量存在时,付费的“护城河”究竟在哪?难道仅仅是抢先体验和一点点速度优势吗?

豆包的这次挫折,或许能给所有AI厂商一个冷静的机会:别再自嗨于模型参数的比拼和融资额的高低了。回到用户本身,去解决那些哪怕很小、但确实存在且愿意付费的“刚需”场景。是更精准的行业报告生成?是更可靠的自动化办公流程?还是无可替代的个性化创意生成?把一件事做到极致,让用户觉得“离不开”,比泛泛地做个“全能但平庸”的聊天机器人,要有前途得多。

610万人的离开,不是用户抛弃了AI,而是抛弃了一个让他们觉得“不值得付费”的版本。技术狂欢之后,商业的逻辑从未改变:价值决定价格,而非相反。大模型的成人礼,看来还得在“收费-流失-再改进”的循环里多摔打几次才行。

Disclaimer: The above content is generated by AI and is for reference only. 免责声明:以上内容由 AI 生成,仅供参考。

融资 融资 政策 政策
Share: 分享到: