Pricing and Inventory Rules: Teaching AI What You Can and Can't Sell

AI-driven answer engines are rewriting the ecommerce playbook. Shoppers are no longer just typing keywords—they’re asking AI agents what to buy for niche needs, who’s got something in stock, and which sellers meet their criteria, from local pickup to sustainable packaging. Whether it’s ChatGPT, Perplexity, or Shopify’s own Sidekick, these engines need up-to-date, accurate details on every merchant’s catalog—for both WHAT is sold and under which which terms.

But here’s the catch: most AI engines treat your Shopify feed as a black box. They don’t know which variant is discontinued, which items are local-only, or if you refuse to ship certain goods to certain places. Generic product feeds or web pages often lose these critical business rules. Worse, ambiguity opens the door to AI hallucinations—suggesting shoppers can buy what you don’t actually offer. The result? Frustrated buyers, missed sales, and risky liability issues.

The Crucial Role of Merchant Constraints

To get AI shopping agents answering correctly, it’s not enough to list what’s in stock. You also need to program in your pricing logic, inventory realities, and critical restrictions—before they become an issue. These are your merchant constraints: the rules that differentiate what you could sell versus what you will sell, given your business logic and policies.

Consider examples: you carry a brand but not their limited-edition line; you sell frozen goods, but only for local delivery; or you restrict instrument sales to in-store pickup. These complex constraints are nearly impossible for an AI agent to infer just from open-ended website text or standard feeds. That’s where formalized, structured discovery files come in—an idea at the heart of the LLM Discovery Files hub approach.

Why Product Feeds and HTML Miss the Mark

Think your Shopify product feed or even your beautiful landing page already solves this? In reality, most feeds are designed for people—or at best, basic price-comparison bots. They say, “here’s inventory,” and “here’s a price”—but rarely explain why an item is unavailable, who can buy it, or how rules like quantity limits apply across variants.

• Shipping restrictions (hazardous, perishable, regional bans) are often buried in policy blobs or FAQ pages AI agents don’t reliably parse.
• Complex pricing rules—say, BOGO deals, volume discounts, or exclusions—are only visible at checkout or hidden in JavaScript. AI can’t synthesize these reliably from public HTML.
• Inventory reality may lag behind—AI sees “in stock” on a page, but your inventory just hit zero after a rush. Without structured updates, agent recommendations go stale or wrong.

Add in your unique business logic—like requiring ID for certain products, or having different refund windows for clearance items—and it’s clear: AI needs formal, machine-readable documentation of your real rules. Structuring that data is now a necessity.

Making Constraints AI-Understandable: Key Methods

The emerging solution is a new breed of structured discovery file, e.g., using protocols like the mcp.json standard championed in the Model Context Protocol (MCP). These files make it possible to expose detailed constraints in a way AI can fully ingest:

• Per-product and per-variant inventory status: Not just “in stock,” but specifying rules like “limited to 2 per customer” or “preorder, ships after July 1.”
• Shipping and pickup eligibility: Clearly flag which SKUs are shippable by region, or local-pickup only. Eg: “No shipping to California for aerosol cans.”
• Pricing logic: Distinguish “list price” from “discounted price,” note coupon eligibility, and lay out bundle or minimum order requirements.
• Policy triggers: Map exactly which return/refund policy binds each product, so AI agents don’t overpromise guarantees you don’t actually offer.
• Hard exclusions: Flag banned, discontinued, or embargoed items—so AI never recommends what you no longer sell, avoiding customer confusion and legal risk.

Think of these structured files as your AI-facing truth serum—they tell answer engines, in no uncertain terms, exactly what’s possible in your storefront today. If you want to go deeper into aligning your business rules with what AI agents actually recommend to shoppers, check out our deep dive: How to Expose Store Policies to AI Shopping Agents.

Shopify Examples: Do’s and Don’ts

Let’s get concrete. Say you’re a Shopify merchant selling gourmet foods, some perishable, some not:

• Do: Use OtoRank to sync a discovery file that states: “Chilled caviar: local delivery only, within 20 miles, ships Mon–Wed.”
• Don’t: Rely on a vague blog post (“We prefer local delivery for fresh items!”). AI shopping agents can’t reliably parse that nuance and will invent or misinterpret your policies.
• Do: Explicitly flag quantity limits for event tickets, so AI answers: “Maximum 4 per order, while supplies last.”
• Don’t: Hide important business logic in checkout-only code or ephemeral banners—LLMs won’t find it.

In practice, we’ve seen stores lose sales because AI agents recommended out-of-stock or restricted items, not realizing those variants were locked behind a pickup-only rule. With a proper discovery file, this never happens: the AI sees, in black and white, every relevant constraint tied to the exact product.

Updating Constraints: The Real-Time Imperative

It’s not enough to publish these rules once—they need to update as your operations change. If you score a new drop-shipping partner, or a regulatory change blocks shipments to a new region, your discovery file should push that change instantly. Apps like OtoRank automate this process, keeping your AEO (answer engine optimization) layer in sync with your Shopify backend.

Well-written constraints also help you rewrite thin or ambiguous product listings. If your old content didn’t spell out restrictions clearly, sync constraints into both your discovery file and your product pages. Rewriting Thin Product Descriptions for AEO covers tactics to clarify details for both humans and AI.

Conclusion: Treat Your Merchant Rules Like SEO—But for AIs

Merchant constraints aren’t just nice-to-haves—they’re the backbone of trust and accuracy in the age of AI shopping. If you want AI agents to recommend exactly what you allow, and never what you don’t, you need to publish real-time, machine-readable business rules. Structured discovery files are the future. Start with inventory, pricing, and fulfillment constraints, and you’ll see the payoff in more accurate answers, fewer disappointed customers, and lower risk—no matter how AIs shop on your behalf.