Store Building Costs Have Collapsed. But Consumers Aren’t Keeping Up.

Shopify’s newly announced “Canvas” is a platform that allows users to build e-commerce sites simply by conversing with the AI agent “Sidekick.” Tasks such as product registration, design adjustments, and copywriting, which previously required outsourcing at costs ranging from 500,000 to 1,500,000 yen or took weeks to accomplish independently, can now be compressed into a few hours through chat-based interactions. The monthly usage fee is expected to be included in existing Shopify plans, resulting in an effective additional cost of just a few thousand yen.

The cost of building a store has dropped to less than one-tenth of what it used to be.

This is undoubtedly good news for small and medium-sized e-commerce businesses. However, there are also some inconvenient statistics emerging.

According to a survey in the United States, only 2% of consumers have actually purchased products using AI. While more people are using ChatGPT and Perplexity to research products, the action of “buying because it was recommended by AI” has not yet materialized.

The cost of creating has collapsed, but consumer behavior has not changed.

How this gap is interpreted will determine the next move for small and medium-sized e-commerce businesses.

“Being Able to Build a Store” and “A Store That Sells” Are Completely Different Matters

First, let’s clarify. What Canvas has solved is the issue of “building costs,” not the problems of “attracting customers” or “purchasing.”

Let’s reflect on the reality of e-commerce construction up to now.

From a cost perspective, this is revolutionary. Small local food manufacturers and craft workshops can now have e-commerce sites comparable to large companies in just a few hours. This is a fact.

However, the question of “A beautiful store has been built. But who will come?” remains unresolved.

What is even more dangerous is the misunderstanding that “since I built a store with Canvas, it should sell.” As the construction costs have decreased, the barriers to entry have also lowered, leading to an explosive increase in the number of e-commerce stores. In other words, while it has become easier to create, it has also become easier to get buried.

The Reason Only 2% of Consumers Shop with AI Is Not “Trust” but “Lack of Pathways”

The draft suggested that “consumers do not trust AI, which is why the purchase rate is low,” but this is somewhat off the mark.

The reality is much simpler. There is simply no established pathway for shopping with AI.

Consider the current consumer purchasing behavior:

1. Search on Google → Jump to Amazon, Rakuten, or official websites → Purchase
2. See something on social media → Click the link → Purchase
3. See it in a physical store → Buy on the spot or later online

There is still almost no room for the step of “asking AI” in this flow. While some people ask ChatGPT, “Please recommend a Japanese sake,” there is no mechanism for completing the purchase directly from there. Perplexity has started providing product links, but it is still in the experimental stage.

In short, the infrastructure for AI to function as an “entry point for purchases” is still not in place. Before the issue of trust arises, the pipes are simply not connected.

Therefore, this 2% figure should be interpreted not as “consumers rejecting AI” but as “the path simply not being established yet.” The moment the path is created, this number could jump significantly.

What Small and Medium-Sized E-Commerce Should Do Now Is Prepare to Be Found by AI Agents

So, what should small and medium-sized e-commerce businesses do?

To conclude, they should prepare to ensure that AI agents correctly understand their products.

This is somewhat different from traditional SEO. It is not about strategies to rank high in Google searches, but rather about creating a state where AI can read product information and recommend their products in response to user inquiries.

There are three specific actions to take:

1. Organize Structured Data

Product names, prices, ingredients, origins, review ratings, and stock status. These should be correctly marked up as structured data (schema.org). AI agents prioritize reading structured data over natural language. Shopify has compatible apps, and support for Canvas settings is expected to improve in the future.

2. Articulate “Why This Product”

When an AI agent is asked about “spicy Japanese sake for women in their 30s,” it must be clearly stated why the product fits that category for it to be considered. Not just specifications, but also writing down “who it is suitable for, in what scenarios, and why” as text on the product page.

This is actually a domain where small businesses excel. Large companies have thousands of SKUs and lack the capacity to write stories for each product. A local brewery with only 50 products can write detailed stories for all of them. Having fewer products can be a weapon in the age of AI agents.

3. Increase Third-Party Mentions

AI agents reference not only information from their own sites but also blog articles, review sites, and social media posts. Getting coverage from local media, increasing reviews on sites like Tabelog and Google Maps, and having local bloggers introduce products can all enhance the accuracy of AI agent recommendations.

In the battle to buy top positions in Google searches through advertising, small businesses cannot compete with large corporations. However, AI agent recommendations are determined by “the quality and quantity of information,” not by “advertising space.” This is where small businesses have the potential for a turnaround.

The Real Question Is Not “For Whom Are AI Stores?” But “When Will Consumer Behavior Change?”

“For whom do AI stores exist?”—this question is actually becoming outdated.

To be precise, AI stores already exist for businesses. They dramatically lower construction costs, eliminate the personalization of operations, and enable e-commerce that can be managed by a single person. This is a confirmed value at this point.

The remaining question is, “When will consumer purchasing behavior switch to AI agents?”

A useful reference is the curve of smartphone adoption. When the iPhone was released in 2007, it was said that “hardly anyone shops using smartphones” until around 2010. By 2015, more than half of e-commerce purchases were made via smartphones. The reversal happened in just five years.

Purchasing via AI agents could follow a similar curve. OpenAI’s GPT-4o has shopping features, Google’s search results now include AI Overview as standard, and Amazon has begun rolling out its own AI shopping assistant. The infrastructure is steadily being put in place.

It may take less than five years for 2% to become 20%.

Checklist for What Small and Medium-Sized E-Commerce Should Do Now

Finally, here is a summary of what local small and medium-sized e-commerce businesses can start doing tomorrow.

• Try out Canvas. If you haven’t used Shopify yet, test it with a free trial. Get a feel for the construction costs with your own hands.
• Double the amount of information on your product pages. Write not only specifications but also stories, usage scenarios, and comparative information.
• Implement structured data. This can be done with Shopify apps. If you are not tech-savvy, ask AI for the code.
• Reach out to local media and bloggers. Third-party mentions are cheaper than advertising and are more effective for AI agents.
• Search for your products with AI agents. Ask ChatGPT or Perplexity for recommendations and check if your products come up. If they don’t, analyze why not.

The cost of building a store is no longer the issue. The real question is whether you are prepared to be chosen when AI selects stores on behalf of consumers.

This preparation can be done faster by small businesses than by large corporations. With fewer products, quicker decision-making, and the ability to genuinely engage with each product.

Those who act now will be the first to benefit when 2% becomes 20%.