What Happens When AI Becomes the “Buyer”: Survival Strategies for Small and Medium-sized E-commerce Revealed by the Moves of Instacart, ChatGPT, and Meta

Conclusion: The "Readers" of Product Pages Are Changing Instacart reads purchase intent using LLMs, ChatGPT integrates

By Kai

|

Related Articles

Conclusion: The “Readers” of Product Pages Are Changing

Instacart reads purchase intent using LLMs, ChatGPT integrates payment functions, and Meta automatically assesses code quality with AI.

When these three news items are lined up, one question emerges.

“Are you still writing your product pages solely for humans?”

This is not just a story from the R&D departments of large corporations. For local small and medium-sized e-commerce businesses, it directly relates to tomorrow’s sales. Let’s break it down step by step.

—

Instacart — AI Understands “What You Want to Buy” First

Instacart is a grocery delivery platform. What happens when it incorporates LLMs into its search and product recommendation?

Traditional search relied on “keyword matching.” When a user types “chicken breast,” chicken breasts are displayed. That’s obvious.

After the introduction of LLMs, things changed. When a user inputs “a healthy dinner that kids will enjoy tonight,” chicken breast, broccoli, and low-fat cheese are suggested together. In other words, products are now matched based on “intent” rather than keywords.

Instacart has announced that this AI search has improved the accuracy of related product displays, leading to an increase in cart value. In its 2024 announcement, the company stated that personalized suggestions utilizing AI contributed to improved purchase rates.

What small and medium-sized e-commerce businesses should consider is clear.

Product pages that only contain product names and prices will not allow AI to read “intent.”

If the structured text explaining “who,” “in what situation,” and “why” this product is used is not written on the product page, it will not appear on AI’s recommendation list. If it doesn’t appear, it’s as good as non-existent.

—

ChatGPT Holds the Wallet — “AI Buying on Your Behalf” Becomes Reality

OpenAI has begun integrating shopping features into ChatGPT. When a user says, “I need a foldable umbrella for a business trip, lightweight and automatic opening, under 3,000 yen,” ChatGPT suggests multiple products that meet the criteria, compares them, and guides the user to purchase.

Even more noteworthy is the movement towards payment integration. If ChatGPT is linked to payment methods, users can complete their purchases with just a simple confirmation of “This is good.” In other words, consumers may finish shopping without ever viewing the product page.

The implications of this are significant.

Traditional e-commerce flows involved “humans viewing pages, comparing, adding to cart, and checking out.” When AI becomes the buyer, this flow changes to “AI reading information, selecting based on criteria, and completing the payment.”

The readers of product pages are shifting from humans to AI.

What worked for humans — “stylish photos” and “emotionally appealing copy” — will not suffice for AI. What is needed includes:

  • Structured data (price, weight, size, material, and usage clearly stated in JSON-LD, etc.)
  • Comprehensive specifications (AI dislikes vague information; instead of “lightweight,” specify “128g”)
  • Clear use cases (keywords tied to intent such as “business trip,” “outdoor,” “commuting”)

This is actually an opportunity for small and medium-sized enterprises.

Large corporations have vast product numbers, making it prohibitively expensive to adapt all product pages to structured data. For small e-commerce businesses with 100 to 500 products, adding structured data can be done in about 30 minutes per product. Even for 500 products, that totals 250 hours. Outsourcing this would cost around 500,000 yen.

They can establish “product pages that AI can read” before large corporations do. This is the turning point for small businesses.

—

Meta’s JiTTests — The Implications of AI Judging Quality

Meta’s JiTTests (Just-in-Time Tests) generate automated tests for high-risk code changes using LLMs. According to Meta’s reports, this method has led to the early detection of 74 significant vulnerabilities related to privacy.

You might think, “This is about software development and has nothing to do with e-commerce.”

But it does relate. The essence is the same.

The structure of “AI making quality judgments” is spreading across all domains.

AI is used in Google’s search quality evaluation. AI is involved in assessing the reliability of product reviews. AI also plays a role in product rankings on e-commerce platforms.

In other words, the entity judging the “quality” of products is shifting from human buyers and consumers to AI. Meta’s JiTTests are a symbolic example of this.

The implication for small businesses is as follows.

Prepare product information that is not just “looks good to humans” but can be assessed as high quality by AI.

Specifically, the following types of information can influence AI’s quality judgments:

  • Clear disclosure of raw materials and ingredients
  • Transparency in manufacturing processes
  • Documentation of third-party certifications and test results
  • Structuring user reviews (not just star ratings, but also the quality of review text)

—

So, What Should You Do?

The underlying structure that connects these three examples is simple.

Traditional Future
Humans search, humans read, humans buy AI reads intent, AI compares, AI recommends (or purchases)
Product pages appeal to “human emotions” Product pages need a structure that withstands “AI analysis”
SEO = inserting keywords SEO = structured data + clear intent + comprehensive specifications

Here are three actions that small e-commerce businesses should focus on starting today.

1. Add Structured Data to Product Pages

Use Schema.org’s Product type to describe price, stock, reviews, and specifications in JSON-LD. If using WordPress, plugins can handle this easily. The cost is almost zero. If done in-house, it costs nothing; outsourcing would be around 500 to 1,000 yen per page.

2. Write About “Who, When, and Why” in Product Descriptions

AI pulls products based on “intent.” Simply stating “high-quality Japanese towels” is weak. Write, “A towel chosen for baby shower gifts, focusing on the texture from Imabari. Safe for newborn skin with additive-free finishing.” This also costs nothing; it only takes time.

3. Avoid Ambiguity in Specifications

“Light” → “128g,” “large capacity” → “500ml,” “durable” → “load capacity 30kg, compliant with JIS standards.” AI compares using numbers and standards. Vague adjectives are just noise to AI.

All three actions can be completed for under 100,000 yen in additional costs. Small businesses can be nimble because they have fewer products.

—

The Real Question Is Here

When you hear “AI becoming the buyer,” it might sound like a distant future.

However, AI is already selecting products at Instacart. More consumers are using ChatGPT to search for products. There are also cases where AI Overviews appear in Google search results, allowing purchasing decisions to be made without opening product pages.

“Product pages that are not chosen by AI” are the same as products that do not appear on shelves.

Just as SEO shifted from “something you should do” to “something you must do to survive,” the preparation of product information compatible with AI will follow the same path.

There is one difference. This time, small businesses can move first. With fewer products, they can structure all pages in a fraction of the time it takes large corporations. Decision-makers are closer to the ground, allowing them to say, “Let’s start next week.”

While large corporations are still going through internal approvals, small businesses can rewrite all their product pages. That is how small businesses can compete.

The question is simple. Is your product page ready to be read by AI?

POPULAR ARTICLES

Related Articles

POPULAR ARTICLES

JP JA US EN