AI agents, don't you think it's someone else's business? How to prepare for a future where AI uses and buys its own services

The footsteps of "Agentic Commerce," in which AI autonomously completes purchasing behavior, are approaching. In this article, we will explain in an easy-to-understand manner the key points of the seminar jointly held by GMO TECH and GMO-PG, from the new era of information design methods to ensure that AI agents can "discover and quote" their services to the response required for future infrastructure of payments and advanced fraud countermeasures.

*This article is an archive of the webinar held in April 2026. The content is current as of April 2026.

Where is Agentic Commerce

"Agentic Commerce" in which AI searches and compares on behalf of the user, and finally autonomously performs up to payment. The expected level varies, from the level at which AI supports comparison to the level where AI agents negotiate purchases with each other.

As a precedent, OpenAI announced "Instant Checkout" and the payment protocol "ACP" in September 2025, which allow purchases to be completed within ChatGPT, and was one of the first to collaborate with overseas platforms. However, in March 2026, it was reported that Instant Checkout would be discontinued, and while there is currently a movement to reduce due to user experience issues and strategic reviews, the future infrastructure development that will lead to Agenric Commerce is steadily progressing, such as Google's announcement of the new standard "UCP".

In Japan, ZOZO has begun to provide new purchasing experiences using AI, such as responding to interactive coordination suggestions and product searches using natural language using Apps In ChatGPT.

The clear difference between companies that are "discovered" by AI and those that are not

Although the domestic expansion of Agentic Commerce is a little further away, the AI-based "discovery" phase is already in full swing. There is also data* that shows that the click-through rate when the company is cited in AI searches increases, while the click-through rate of uncited sites halves, so it can be said that it is no longer a matter of life and death. *Ahrefs survey results

From the perspective of "LLMO (Large Language Model Optimization)", which creates a structure that is easy to cite by AI, the following three conditions must be met.

• Retrievable: To make it easier for AI to cut out and cite information, it's important to break down a single topic into the smallest units of about 200 to 400 characters. If the product information is only in an image or trapped in a PDF, the AI will judge the information as "non-existent."
• Verifiable:Avoid unverifiable vague expressions like "Actual many" or "high quality" and present specific numbers or fact-based information.
• Corroborated: The most important thing is that the company name, service name, and public figures are consistent not only on your own website, but also in multiple external sources such as press releases, industry media, and Wikipedia. This is because AI collates multiple sources to check consistency before citing.

3 steps to take from tomorrow

These measures are permanent investments that should be started now, regardless of industry or company size. Specifically, it is recommended to proceed with the following steps:

1. This week: Select one important page from your site to check if you're answering questions that might be derived from users.
2. This month (breakdown into smallest units):Present information buried deep in FAQs and in images and PDFs, and organize them so that AI can read them.
3. This quarter (external consistency check): In cooperation with the public relations department and others, we will correct and unify discrepancies in the company's notation and figures in external media.

infrastructure of payments in the AI Age and Advanced Fraud Protection

After AI discovers and proposes its own services, it will also need to be handled by the infrastructure of payments. In the future, it will be necessary to support payment protocols that are highly compatible with AI agents and adopt a mechanism that automatically executes payment from the outside.

What is even more important to pay attention to is the reality that cyberattacks and fraudulent purchases have already been AI-enabled. To prevent attacks from high-volume, optimized AI bots and ensure legitimate purchases, defenders also need to implement advanced AI-centric detection systems.

Conclusion: Selected by AI to complete transactions safely

The future of AI agents completing purchases is not a distant story, but a real business challenge that is already underway behind the scenes.Creating a state of "AI discoveries" through information design optimization (LLMO) is the most important initial setting for e-commerce service providers, and it is also something that can be addressed immediately.At the same time, it is essential to update infrastructure of payments to prevent increasingly sophisticated fraud attacks and support new protocols.

Let's start with the "decomposition of information into the smallest units" and "external information integrity check" introduced in this article, and build a solid business foundation that will be chosen by both users and AI in the next-generation commerce environment.