# What AI Shopping Agents Do When the Product Is Out of Stock

Source: https://aicommercewatch.com/blog/2026-10-11-ai-agents-out-of-stock-behavior.html

October 11, 2026

*An agent that cannot buy is an agent that improvises. How shopping agents handle stockouts, and what merchants should do about it.*

AI shopping agents are starting to complete real purchases, which means they are also starting to hit the most common ecommerce edge case: the product is out of stock. What an agent does next reveals a lot about how these systems reason, and it creates new problems for merchants who assumed a stockout was a dead end. Here is what actually happens.

## The agent improvises

When the exact product is unavailable, agents do what a determined shopper might do: they look for alternatives. That can mean a different size, a different color, a refurbished unit, or a competing product. The quality of the substitution varies wildly. Some agents ask the user first. Others just pick the closest match and check out.

This is where merchant intent and agent behavior diverge. A shopper who wanted a specific model in black does not necessarily want a different model in white. Agents optimize for task completion, and 'bought something similar' counts as completion in a way the user may not accept.

## The gray areas get uncomfortable

Out-of-stock situations push agents toward third-party sellers and marketplaces with looser controls. An agent told to buy a specific item may end up purchasing from a seller the merchant would never authorize, at a markup, with different return terms. The user sees a confirmation. The details are buried.

There is also the backorder trap. Some agents will place backorders or preorders without clearly communicating the wait. A shopper who needed the item this week discovers the delivery date after the charge. Merchants should assume agents will take any purchasable path and design accordingly.

## What merchants should do about it

First, make stock status machine-readable and unambiguous. Clear structured data about availability, restock dates, and authorized sellers helps well-behaved agents make better decisions. Ambiguity is where bad substitutions happen.

Second, decide your substitution policy and encode it. Which alternatives are acceptable? Are third-party sellers ever okay? What is the price tolerance? These are business rules that need to exist somewhere an agent can read them, not just in a human-readable returns page.

## The bigger picture

Stockouts are a preview of a larger shift. Agents do not browse the way humans do; they execute against goals. Every edge case in your catalog, every ambiguity in your policies, becomes a decision the agent makes on the user's behalf, sometimes badly.

Merchants who treat agent traffic as a first-class channel, with clear structured data and explicit policies, will get better outcomes than those who treat it as weird bot traffic. The agents are not going away. The question is whether they work with your intent or around it.

## The authorized-seller problem

Stockouts push agents toward third-party sellers, and that creates a brand-control problem. An agent buying your product from an unauthorized reseller at a markup, with different return terms, is technically completing the task while violating everything the brand stands for.

The fix is an agent-readable authorized seller list: structured data naming the sellers you stand behind, your minimum advertised price policy in machine-readable form, and clear return terms per channel. Agents cannot respect policies they cannot parse. Publish the rules where the machines look, not just where the lawyers look.

## Frequently asked questions

### Can merchants block AI agents from buying?

Technically yes, but blocking is a blunt tool that also blocks legitimate automation and frustrates users who delegate shopping. Most merchants are better served by guiding agent behavior than blocking it.

### Do agents respect robots.txt?

The well-behaved ones claim to. But purchase-capable agents often operate through browser automation that does not consult robots.txt the way a crawler does. Do not rely on it as your control mechanism.

### Should we build an agent-specific API?

If agent traffic is material for your store, yes. A structured buying API with clear inventory, pricing, and policy data beats screen-scraping on every dimension: reliability, cost, and control.

## Frequently asked questions

### What do AI shopping agents do when an item is out of stock?

Most move on quickly. An agent assembling options will usually substitute a comparable in stock product or drop the store entirely, because its job is to complete a purchase, not to wait. Clear stock status and good alternatives are what keep the store in the running.

### Should out of stock pages suggest alternatives?

Yes, and the alternatives should be real, in stock, and genuinely comparable. Agents follow those suggestions when ranking options. A page that dead ends teaches agents your store is a dead end.

### Does hiding out of stock products help or hurt with agents?

Clear labeling beats hiding. An agent that discovers a stock problem late in the journey counts it against the store. Accurate, machine readable stock status lets the agent route around the gap and still buy from you.

[Home](../index.md) / [Blog](index.md)