AI Chatbots & Shopify: Why Your Products Might Show 'Out of Stock' (When They're Not!)

View original discussion by beckyfmartin on Shopify Community →

Hey there, fellow store owners! Let's talk about something that's been buzzing in the Shopify community and can be a real head-scratcher: AI chatbots incorrectly telling customers your products are out of stock when they're actually sitting pretty in your warehouse. It's frustrating, right? You know your inventory is good, but these clever Large Language Models (LLMs) seem to be getting the wrong idea.

Recently, a store owner named Becky (beckyfmartin) brought this exact issue to the community forum, asking for help because LLMs kept reporting her products as unavailable. The discussion that followed was a fantastic deep dive into how these AI agents 'read' our stores, and it uncovered some crucial insights that every Shopify merchant needs to know.

The Mystery of the 'Out of Stock' Ghost

Becky's initial thought was that the LLMs were simply scanning her theme's code, finding the words "out of stock" in a template string, and mistakenly assuming the product was unavailable. And honestly, that's a very logical first guess! Many themes *do* have these strings ready to display when a product actually sells out. But as several experts in the thread, like Maximus3 and Ploqo, pointed out, this is rarely the culprit.

Bristan_FARRE, who took a deep dive into Becky's actual product page, confirmed this. They found the `window.variantStrings` dictionary containing "Out of stock" was indeed present but was just a UI label dictionary, not an actual status. So, if an AI is smart enough to parse a web page, it's usually smart enough to know the difference between a UI label and a live inventory status.

Where LLMs *Actually* Get Their Info

So, if it's not the harmless theme text, what is it? The community discussion highlighted a few key areas:

1. Structured Data (Schema.org) – The AI's Rosetta Stone

This was the most emphasized point by many, including Jarid, Maximus3, and Appify_Commerce. LLMs and search engines heavily rely on structured data, specifically the `schema.org/Product` and `schema.org/Offer` blocks embedded in your product pages. This is machine-readable information that explicitly states things like product name, price, and, crucially, availability.

Bristan_FARRE's analysis of Becky's page showed that her primary Product block correctly reported `https://schema.org/InStock`. Good news there!

2. Duplicate or Broken Structured Data Blocks

Here's where things get tricky. Bristan_FARRE found a critical issue on Becky's page: two Product blocks. Even worse, the second one was invalid JSON due to an unescaped newline character inside the description. When you have conflicting or broken structured data, an AI crawler might pick the wrong one, or a parser error could lead to incorrect assumptions.

Vestfoldhall and Robert_Dainowski also flagged this as a common problem, often caused by SEO apps, review apps, or even theme customizations adding extra, sometimes conflicting, structured data.

3. Shopify's Agentic Sales Channels and Dedicated AI Feeds

This was a big revelation in the thread, brought up by v.marychenka, Robert_Dainowski, and Ploqo. Shopify now has built-in mechanisms for AI agents. Your store isn't just serving its regular product page to bots; it's also publishing inventory to dedicated "agentic storefronts" or feeds.

  • Sales channels > Agentic: In your Shopify admin, there's a setting under 'Sales channels' for 'Agentic'. If "Allow Shopify to manage for me" is active, Shopify Catalog publishes your inventory to these AI channels.
  • The `/llms.txt` and `/robots.txt` files: Your store has a special `robots.txt` file that tells crawlers what they can access. Crucially, as Ploqo and Bristan_FARRE pointed out, it often points to an `llms.txt` file (e.g., `graes.co/llms.txt`). This file declares things like a UCP profile and an MCP endpoint with a `search_catalog` tool. This means AI agents are specifically told to query *that path* for accurate pricing and availability, often passing buyer context (country, currency). If an agent queries without that context, it might get a stale or incomplete answer.

4. Caching and Retrieval Issues

Sometimes, the data itself is correct, but the AI is simply reading an old, cached version of your page. Icey.Lane and Appify_Commerce suggested asking the specific LLM for the exact URL and retrieval date it used, which can help diagnose caching problems.

Your Action Plan: How to Diagnose and Fix 'Out of Stock' Misinformation

Based on the community's collective wisdom, here's a step-by-step guide to ensure AI agents correctly report your product availability:

Step 1: Inspect Your Product Page's Raw HTML (Structured Data Check)

This is your first port of call. You want to see what the AI crawler sees, *before* JavaScript runs.

  1. Go to your product page in your browser (e.g., `yourstore.com/products/your-product`).
  2. View the page source: Press Ctrl+U (Windows/Linux) or Cmd+Option+U (Mac).
  3. Search for structured data: Use Ctrl+F or Cmd+F and search for