Mercari enhances product listings and better supports sellers with GPT-4o mini

Mercari, one of Japan’s leading online marketplaces, saw downloads soar after expanding into the U.S. in 2014. By integrating GPT‑4o mini and other models, Mercari is redefining how users buy and sell secondhand goods, making the process more seamless and accessible worldwide.
To simplify listings and boost sales, Mercari incorporated OpenAI’s API into two key features. These enhancements reduce friction, helping users create high-quality product pages that drive more successful transactions.
Launched in October 2023, the Mercari AI Assistant helps users optimize their listings. If an item remains unsold for a certain period, the AI evaluates the listing against best practices and provides tailored suggestions.

This feature initially relied on a hybrid model approach using the OpenAI models:
- Offline analysis: GPT‑4 analyzed top listings in each category to identify key attributes.
- Online suggestions: GPT‑3.5 Turbo evaluated active listings of users and provided real-time improvement tips.
The result? Mercari observed a statistically significant increase in average sales per user, demonstrating the feature’s effectiveness in improving listing quality.
“Today, these features are seamlessly integrated into the listing experience, allowing users to benefit from AI without even being aware of it,” says Yasuo Hishii, Supply Team Director at Mercari. This approach ensures the technology works behind the scenes, enhancing the process without adding complexity.
Taking the AI-driven selling experience further, Mercari introduced AI Listing Support in September 2024.
“The success of the Mercari AI Assistant gave us a lot of confidence in this direction,” says Max Frenzel, Head of AI/LLM at Mercari. “We believed we could make significantly larger improvements to the listing experience itself. So we decided to try to massively simplify it and reduce the friction across the entire listing process.”
With AI Listing Support, users can list items effortlessly by simply uploading photos. Powered by GPT‑4o mini, the feature suggests relevant categories, then generates complete titles and descriptions tailored to the product and its category.

Originally developed with GPT‑4, the feature later transitioned to GPT‑4o mini for superior cost-efficiency and reduced latency. This switch also enabled Mercari to handle multiple images per listing, which significantly improved the quality of AI-generated content.
“It took about two months from the initial concept to the final release,” says Hishii. Even after launch, Mercari continues to refine AI Listing Support by comparing AI-generated descriptions to actual listings, further enhancing the user experience.
Positive results and user feedback
- Over a few hundred AI-assisted listings are created per minute: Listing conversion rates received a substantial boost.
- Positive user feedback: Users appreciated reduced friction and time savings. Less experienced sellers reported increased confidence in creating professional-quality listings.
“This feature has delivered clear, measurable results,” Hishii notes. By simplifying the listing process, these AI-powered tools help more users achieve their goals, resulting in a smoother and more satisfying experience.
“Other competitors would occasionally catch up or even briefly surpass OpenAI’s models, but OpenAI always swiftly released a new version that put them back on top.”
Strong leadership and grassroots innovation were both critical in embracing AI initiatives at Mercari.
The rise of ChatGPT and rapid advancements in AI prompted the company to reassess its strategy. “We had very strong top-down support to make these initiatives happen,” Frenzel explains. “Our leadership, from the Group CEO to the C2C Marketplace CEO, emphasized the importance of not missing this AI wave.”
This top-down approach was complemented by grassroots innovation. Developers across the company began experimenting with AI tools, developing Slack bots and quick web applications. “We also saw it converge from the top down and from the bottom up,” says Frenzel. “We're now creating a centralized hub where everyone across the company can submit their tools and bots, fostering a collaborative environment for AI integration.”
This dynamic combination of top-down leadership and bottom-up innovation has been instrumental in Mercari's successful integration of AI.

Looking ahead, Mercari envisions AI playing a pivotal role in shaping the company’s future products and services. “We believe AI can have a transformative impact across the entire Mercari ecosystem,” says Frenzel. “We plan to explore these possibilities much more deeply.”
The AI/LLM team focuses on three parallel goals:
- Developing flagship features: Improving and expanding key offerings like AI Listing Support.
- Creating AI building blocks: Providing foundational AI components for all product teams, enabling them to deliver AI solutions to customers even without internal AI specialists.
- Innovating internal tools and services: Leveraging AI to build new tools and deliver productivity boosts across the entire Mercari Group.
“Search and recommendation features powered by more advanced AI models are already in the works,” Frenzel explains. “We’re also focusing on fraud detection, customer support, and beyond. We want to give every product team the ‘Lego blocks’ of AI so they can build their own solutions.”
The Supply team is looking to improve the post-sale experience for users. “One unique aspect of Mercari is the number of offline tasks that users need to manage,” Hishii explains. “There are additional tasks after a sale as well such as packaging and shipping. Choosing the right packaging, shipping method, and shipping origin can still be pain points for many users. We want to work with OpenAI to provide support in these experiences as well.”
By integrating OpenAI's advanced models, Mercari aims to create a cohesive AI ecosystem where each interaction enhances the platform's capabilities. “OpenAI is our key partner in this process,” Frenzel explains. “The key features we've launched so far were all developed in partnership using OpenAI's models. We're being bold and committing to future developments of AI we expect will come from OpenAI.”