AI Revolutionizes Advertising: Higher Conversions & ROI – Experts Discuss Future Trends

The conversion rate of AI-generated advertising materials is 20% to 30% higher than that of manually created materials, according to recent data from Kuaishou Technology, a leading Chinese short-video platform. The cost of a single ad asset has plummeted to just a few cents, while a “generative reinforcement learning for bidding” technology is significantly improving return on investment for advertisers.

These developments signal a deeper transformation in advertising and marketing, as artificial intelligence moves beyond technological demonstration and into mainstream commercial application. To discuss the implications, 36Kr convened Jiang Peng, Vice President of Kuaishou Technology and Head of Commercial Algorithms, and Kazik, CEO of Virtual Reality Media and operator of the “Digital Life Kazik” platform.

Both executives highlighted the recent emergence of DeepSeek R1 as a pivotal moment. Kazik described it as demonstrating that Chinese models can now achieve performance levels comparable to OpenAI’s offerings, but at a substantially lower cost. “Every year there are recent releases in the AI field during the Spring Festival,” Kazik said. “Last year, DeepSeek R1 actually showed that Chinese models can reach the level of OpenAI o1 in performance while significantly reducing costs. This model’s popularity has sparked a nationwide debate.”

Jiang Peng echoed the sentiment, identifying the deep transformation of the content industry through AI as a key development. “The ability of AI to create content has made a leap recently and has had a lasting impact on the entire content ecosystem,” he stated. He too pointed to the increasing commercial appeal of AI, noting a colleague in Silicon Valley found the cost of using an AI assistant exceeded their own salary.

The discussion also centered on the surging popularity of OpenClaw, an open-source agent framework. Kazik explained its rise in terms of stages of public perception. “For most people, there are only three phases of AI perception: ChatGPT made them aware that ‘AI can chat,’ DeepSeek R1 made them aware that ‘AI can think,’ and OpenClaw made ordinary users aware of the concept of ‘agents’ for the first time.” He contrasted OpenClaw’s accessibility with earlier agent products like Manus and Claude Code, which he characterized as cloud-based services that felt like “renting” rather than owning.

Jiang Peng agreed, adding that OpenClaw allows ordinary users to experience the value of AI in solving practical problems. “Technically, OpenClaw is largely in line with the agent technologies of the past two years. Its unique value lies in enabling ordinary users to install AI locally and solve everyday problems,” he said. “I hope this popularity will bring new applications, new industries, and monetization opportunities for agents.”

Kuaishou’s data demonstrates the effectiveness of AI-generated advertising. Jiang Peng reported that the cost of producing ad materials on the platform has fallen to just a few cents per asset, with an overall conversion rate increase of 20% to 30% compared to materials created by customers themselves. He emphasized the platform’s use of a “personalized” strategy, tailoring materials to specific product features and user characteristics.

Kazik highlighted the cost savings in video production, stating that AI can produce TV commercials for a fraction of the cost of traditional methods – potentially 10,000 to 20,000 yuan per minute compared to 100,000 to 300,000 yuan for traditional production.

The “generative reinforcement learning for bidding” technology employed by Kuaishou is designed to optimize ad placement in a highly uncertain market. Jiang Peng explained that it addresses two core problems: environmental uncertainty and sequential decision-making. The technology has reportedly increased advertising revenue by 3% to 4%, improved ad placement capacity by 20% to 30%, and improved ROI for advertisers. Currently, approximately 90% of the ad placement process is automated.

Looking ahead, Jiang Peng anticipates continued competition around AI tools, but believes the focus will shift towards efficiency of investment and return. “Companies want users to actually use the AI tools and not just take advantage of the offers and then leave,” he said. Kazik echoed this sentiment, noting that cognitive development regarding AI happens in stages and cannot be simply driven by financial investment.

The commercial impact of AI on content creation was also discussed, with a focus on the potential for AI-generated comics to generate revenue. The conversation did not elaborate on specific revenue models or examples.