All Episodes
17 min

230: A Shoe Company And A Cookie Company Now Say The Exact Same Thing To AI Agents

SpotifyApple
HOSTED BY
Slobodan "Sani" Manic

Slobodan "Sani" Manic

Website Optimisation Consultant, No Hacks Founder & Keynote Speaker

CXL-certified conversion specialist and WordPress Core Contributor helping companies optimise websites for both humans and AI agents.

I opened three unrelated Shopify stores, Allbirds, Brooklinen, and Partake Foods, and asked each what it could do for a machine. All three answered with the same roughly 800 words, identical character for character. A shoe company, a bedding company, and a cookie company now sound exactly alike to any AI agent that visits.

On August 5th, Shopify switched on WebMCP across every store using their templates. No installation was required, and no merchant was asked. The page now hands browsers a list of functions an agent can call directly: Search the Catalog, Go to Checkout. The descriptions read like stage directions for machines. One literally says "Follow it." Another tells agents not to ask shoppers about missing options when they only want to browse. Someone decided when customers get consulted, and it was neither the customer nor the merchant.

This is actually the right way to build it. The tools read the same database as the storefront humans see, so the two cannot drift apart. One good default beats a million bad implementations. But the room is still empty. Etsy reports agent traffic under one percent. Shopify's "AI-referred orders tripled" stat is people arriving from AI answers, then buying with their own hands. That is not agents completing transactions.

If agents do start buying, your photography, badges, reviews, the copy someone agonized over, none of it survives the call. Back comes a title, a price, a size run, and availability. What's left to compete on is whether your prices are right and your stock is accurate. The boring work becomes the entire game.

Shopify WebMCP RolloutIdentical Agent InstructionsTool Descriptions as Stage DirectionAgent Commerce Adoption RealityBrand Voice vs. Data Accuracy

KEY TAKEAWAYS

  • Shopify wrote what your store says to machines, and every other Shopify store says it too. The tool descriptions are version 0.1.0, identical across all merchants regardless of brand or category.
  • WebMCP tools reading the same database as the human storefront prevents drift between channels. This architectural decision means one good default protects a million stores from inconsistent implementations.
  • If agent commerce scales, competition shifts entirely to data quality: accurate prices, real-time stock availability, complete catalogs, and machine-parsable policies. Brand voice does not transmit through JSON responses.
  • Current AI commerce metrics are misleading. Shopify's 'tripled AI-referred orders' measures humans arriving from AI answers, not agents completing purchases. The infrastructure exists, the traffic does not.
  • Shopify's documentation at shopify.dev/docs/agents teaches people how to build buyer agents, not how to run stores. The platform is positioning to influence both sides of the marketplace.

SHOW NOTES

The Same 800 Words, Three Different Stores

Allbirds, Brooklinen, and Partake Foods all hand AI agents the same roughly 800 words, character for character, written entirely by Shopify. Allbirds spent years learning how to sound like Allbirds. Partake Foods has its own voice for allergy-friendly cookies. Brooklinen built a brand around premium bedding. None of that branding reaches an agent that asks what the store can do.

On August 5th, 2026, Shopify enabled WebMCP across every store running its liquid templates. There was no installation, no opt-in, and no merchant notification. The page now serves a list of callable functions to any browser-based agent: Gemini in Chrome, Perplexity Comet, whatever comes next. WebMCP is at version 0.1.0, and merchants have zero control over what it says.

Shopify's Tool Descriptions Tell Agents When to Skip the Shopper

Shopify's WebMCP tool descriptions are explicit instructions rather than marketing copy. The checkout tool says: "Verifies the cart is not empty and moves the browser to the checkout page. The tool result then confirms the destination and instructs you not to navigate again. Follow it." Another tool handling product variants tells agents not to ask users about missing options when their intent is only to browse.

Someone made decisions about when shoppers get consulted, and it was neither the shopper nor the merchant running the store. Shopify's documentation at shopify.dev/docs/agents addresses the people building buyer agents, not store owners. Shopify is teaching machines how to shop while hosting the stores they will shop at.

WebMCP Tools Read the Same Database as the Storefront

The WebMCP tools read the same database as the human-facing storefront, so prices, inventory, and product details all come from one source. Reading from one source prevents the drift that plagues companies maintaining separate systems for different channels.

One good default beats a million bad custom implementations. If Shopify had asked each merchant to write their own tool descriptions, most would have done it poorly or not at all. Standardization at this layer protects the entire ecosystem from fragmentation. Knowing a default was set for you still changes things, especially when that default erases everything that made your store distinctive.

Etsy Reports Agent Traffic Under One Percent

Etsy reported AI agent platform traffic under one percent in its second-quarter 2026 earnings, so all this infrastructure serves an audience that barely exists. Shopify's headline that AI-referred orders tripled sounds impressive until you parse it: that metric tracks humans arriving from AI answers, then buying with their own hands. Referred traffic is not agentic commerce.

More than a million storefronts learned identical commands, and the room is still empty. Every major AI company is pouring money into browser agents, so twelve to eighteen months looks like a reasonable timeline for meaningful adoption.

An Agent Query Returns Title, Price, Size, and Availability

When I queried Allbirds for a product through WebMCP, back came a structured JSON file containing a title, a price, a size run, and availability. None of the photography, emotional targeting, badges, reviews, or carefully crafted copy came with it.

If agents become a real purchasing channel, what's left to compete on is whether your prices are right, your stock is accurate, your catalog is complete, and your policies are parsable. The boring operational work becomes the entire game. Operational accuracy pays off regardless of whether agents arrive at scale, because accurate data serves human visitors too. The stakes change when machines strip away everything else.

QUESTIONS ANSWERED

What is Shopify WebMCP and how does it work?

Shopify WebMCP is a machine-readable interface layer that Shopify enabled on August 5th, 2026 across all stores using their liquid templates. WebMCP provides AI agents with a list of callable functions like Search the Catalog and Go to Checkout, allowing browser-based assistants such as Gemini in Chrome or Perplexity Comet to interact with stores directly through structured commands rather than parsing HTML.

Do Shopify merchants control what their stores say to AI agents?

Shopify merchants currently have no control over WebMCP tool descriptions. Shopify wrote identical text for all stores, approximately 800 words that appear character-for-character the same across unrelated businesses like Allbirds, Brooklinen, and Partake Foods. The current implementation is version 0.1.0 with no merchant customization options.

How much AI agent traffic are e-commerce websites actually getting?

AI agent traffic remains minimal despite infrastructure investments. Etsy reported AI agent platform traffic under one percent in its second-quarter 2026 earnings. Shopify's claim that AI-referred orders tripled refers to humans arriving from AI-generated answers and buying manually, not autonomous agents completing purchases.

What is the difference between AI-referred orders and agentic commerce?

AI-referred orders involve humans who arrive at a store after receiving information from an AI system like ChatGPT or Perplexity, then complete purchases manually. Agentic commerce involves AI agents autonomously browsing, selecting products, and completing transactions without human intervention. Shopify's reported tripling of AI-referred orders measures the first category, not the second.

How does WebMCP change what e-commerce brands compete on?

WebMCP strips away visual branding, emotional targeting, photography, reviews, and copywriting when AI agents are the visitors. Agents receive only structured data: product titles, prices, sizes, and availability in JSON format. Competition shifts entirely to data quality metrics like price accuracy, real-time inventory, catalog completeness, and machine-parsable return policies.

Why did Shopify build WebMCP with identical descriptions for all stores?

Shopify's standardized approach makes WebMCP tools read from the same database as human-facing storefronts, which prevents data drift between channels. One consistent default across a million-plus stores prevents the fragmentation that would occur if each merchant wrote custom tool descriptions. The architecture prioritizes reliability and accuracy over brand differentiation at the agent interface layer.

ENJOYING THIS EPISODE?

Practical strategies for making your website work for AI agents and the humans using AI to find you. Once a week you get the new articles, the latest podcast episode, and a few links worth keeping.