AIGC Brings New Business Opportunities
Over the past year, competition among large language model providers has intensified. Major models like Alibaba's Tongyi, ByteDance's Doubao, and Tencent's Hunyuan all offer AI copywriting functions for ordinary users. A large number of software based on these models have also emerged.
An AI industry technical expert told Jiemian News that the barrier to entry for AI copywriting software is relatively low, requiring only access to the API of large language models. Based on Jiemian News' experience, while these software have slight differences in creative rights, their pricing models are mostly similar.
Taking the web version of one software as an example, after logging in, users can use the AI copywriting function based on a point consumption system. After the free trial, the software's pricing changes to a membership fee of over 100 yuan. Membership benefits include a "package" of creative functions such as AI copywriting, AI video generation, AI digital humans, and AI face swapping.
There are also more vertical AI copywriting software that focus more on text creation but use the same business model as the above software. A software experienced by Jiemian News set a lifetime membership price of 168 yuan, offering AI copywriting as well as intelligent continuation and rewriting functions.
Most of these software are suitable for ordinary users. SaaS service providers targeting the enterprise sector have also successively launched AI products in the past two years, focusing on AI copywriting needs. Last year, Youzan and Weimob both launched AI tools including AI copywriting functions, with Weimob introducing a viral post imitation feature specifically for the Xiaohongshu platform.
Service providers also integrate large language models at the base, but because brand clients have higher and more complex requirements, they need to incorporate marketing business experience. Chen Chengze, Algorithm Director and WAI Product Consultant at Weimob, told Jiemian News that without specialized debugging, the effects of many software on the market are quite homogeneous. However, for brand clients and professional marketers, these copies are not up to standard and differ greatly from manually written ones.
For example, some maternal and infant clients hope the posts' style will be more formal and official, with characteristics of the account itself. Some clients want to imitate certain high-ROI posts, which involve more requirements.
In this process, Weimob uses core optimization methods including fine-tuning and RAG (Retrieval-Augmented Generation) technical solutions to incorporate the company's accumulated marketing cases and brands' personalized needs into the model, generating AI copy that "reads like human-written". This improvement method is not technically complex; the focus is on content accumulation. Regularly testing whether the large language model's generation ability is "keeping up with the times" is also an important part of optimizing AI copywriting functions.
Paying for AI Still Needs Time
After a period of "practice", AI can indeed become a capable assistant.
For individual creators or studios, using AI to save time and costs is the most urgent need. A Xiaohongshu blogger told Jiemian News that for writing business or product introduction-type copy, AI tools can already achieve relatively intelligent expansion and keyword coverage. For example, when part-time interns learn how to write account copy, giving AI a Xiaohongshu post they want to imitate can produce a stylistically similar post, saving a lot of time.
Letting AI do the basic "text building" to eliminate the most basic work is what advertising companies are most pleased to see.
A marketing professional at an advertising company serving fast-moving consumer goods brands told Jiemian News that she currently rewrites copy directly based on AI-generated content, which can save her about 20% of time.
The tool she frequently uses is Kimi. In fact, Moonshot AI, the company behind Kimi, excels at promoting on young platforms like Bilibili and Xiaohongshu, and has received investment support from Xiaohongshu. Many writers say they prefer to use such free software.
For brands, the effectiveness of AI depends on the industry and specific needs. According to data provided by Weimob, on the Xiaohongshu platform, content generated by AI creative models for a maternal and infant brand has a usability rate of 90%, basically ensuring consistency in platform style, brand tone, content language style, and even title style. However, for industries like tourism, where products have higher differentiation and marketing promotions require more detailed personalized adjustments, the time saved on text creation is currently only about 50%.
The effectiveness also directly affects users' willingness to pay.
The aforementioned blogger and marketing professional currently do not use paid software, partly due to cost constraints and partly due to the homogenization problem of basic AI tools. The technical expert mentioned above said that in cases leaning towards the user end, large language model providers often do not optimize for specific scenarios, which requires higher costs. Some small and medium-sized companies that do AI creative training in professional fields can actually produce better results, but the current situation is that products are mixed in quality.
QuestMobile data shows that at present, domestic AIGC applications mainly focus on revenue models such as charging for output content, subscriptions, or membership fees. Additionally, after rapid growth, user-end AIGC products often face challenges in user retention and activity rates. In comparison, enterprise clients have direct business needs and stronger willingness to pay.
However, for many brands hoping to achieve cost reduction and efficiency improvement through AI, paying for AI also requires significant decision-making costs.
Chen Chengze believes that the current growth of products needs more brands to use them. In addition to establishing reputation, there need to be more good articles for products to iterate and optimize, resulting in better effects and thus enhancing product competitiveness. This is also a consensus reached by many AI practitioners - more people using AI can promote AI learning, ultimately achieving harmonious human-machine collaboration.