Agents In Use
What’s new todayLast update
47 use cases added today

ChatGPT agent assembles an Instacart grocery cart

TechRadar’s Eric Hal Schwartz tested whether AI agents could turn a short grocery request into a usable delivery cart. He gave ChatGPT and Claude the same Instacart shopping task, including infant formula, vanilla ice cream and fruit. The comparison focused on how much of the online shopping process each assistant could complete, rather than on a general recommendation list.

Source published techradar.comAgent: ChatGPT agent
OUR ADAPTATION

Suggested prompt

Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.

Use the agent to complete this specific task, show the sources or items it used, and stop for my review before sending, buying or changing anything.

Evidence at a glance

Sourced account
Source poster / publisher
Eric Hal Schwartz
Original account
Read the source ↗
Published
Source link checked
Reported result
ChatGPT assembled an Instacart cart for the test list. The article does not report a completed purchase or delivery.
Human involvement
The shopper can verify product variants, quantities, delivery details and payment before completing the order.

A linked account supports attribution. It does not mean we independently reproduced its result.

Report an issue with this caseOpen correction email

Opens your email app with the case link included. Nothing is sent automatically.

How the agent did it

  1. Schwartz gave ChatGPT and Claude the same short grocery list.
  2. ChatGPT searched Instacart products and populated a cart with requested groceries.
  3. The shopper could review the selections and delivery details before checkout.

TechRadar’s Eric Hal Schwartz tested whether AI agents could turn a short grocery request into a usable delivery cart. He gave ChatGPT and Claude the same Instacart shopping task, including infant formula, vanilla ice cream and fruit. The comparison focused on how much of the online shopping process each assistant could complete, rather than on a general recommendation list.

ChatGPT used the Instacart connection to search for products and assemble a cart. Schwartz could inspect the selected items and review the delivery details before checkout. Claude also produced a similar cart through a logged-in browser workflow, so the article is a comparison rather than evidence that only one product can shop. The reported result is a prepared basket, not a completed purchase.

The useful pattern is to ask an agent to gather the requested items while keeping the final review and payment with the person ordering. Grocery catalogs can contain several close matches, package sizes and substitutions; Schwartz’s example is a reminder to inspect the exact products and quantities before paying. The source does not report a delivery, price saving or independent accuracy score, and the available integrations may change over time.

Prompt
Suggested prompt (written by Agents In Use, not the author's)

Source: TechRadar · Eric Hal Schwartz

Frequently asked questions

What items were in the test list?

Schwartz used infant formula, vanilla ice cream and fruit in the grocery request.

Did ChatGPT complete the purchase?

The article describes a cart that the shopper could review before checkout. It does not report that ChatGPT placed or paid for the order.

Was Claude tested too?

Yes. Schwartz gave Claude the same task and reports that it also produced a similar cart through a browser session.

More details

About this account

ChatGPT searched products and assembled a cart. The source leaves review and checkout with the shopper.

The article describes the grocery list and test but does not provide a verbatim prompt. The prompt shown is editorially suggested.

Credit: Eric Hal Schwartz — TechRadar

Source published

Added to Agents In Use

Source checked

Tools and topics

ChatGPT agent · Instacart · Claude

grocery shopping · Instacart · online cart

Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.