![\"ChatGPT](\"https:\/\/teknomers.com\/en\/wp-content\/uploads\/2025\/10\/The-ChatGPT-Atlas-agent-made-my-purchase-at-Mercadona-and.jpeg\")
\n <\/div>\n<\/div>\nPrior discussions highlighted that AI-based browsers could be vulnerable to \u00a0prompt injection attacks\u00a0, and OpenAI is aware of this risk. Before beginning, an advisory note stated that \u00a0using agent mode carries risks\u00a0, giving me the option to use it either logged in or logged out. I chose to log in to simplify the process, but I first deleted my payment details on the Mercadona website for safety.<\/p>\n
<\/p>\n
Making the Purchase<\/h2>\n
Once I accepted the risks, agent mode kicked in, and I observed as the \u00a0mouse navigated\u00a0 the Mercadona website. The sidebar displayed the model’s thought process while selecting ingredients for the \u00a0chickpea curry\u00a0. The video documented the entire purchase sequence, showcasing how the agent was making decisions about various items.<\/p>\n
<\/p>\n
For instance, the recipe called for one onion, but it opted for a \u00a01kg package\u00a0 instead. When selecting spinach, it favored baby spinach over the cheaper larger package. After choosing all ingredients, the agent asked me for confirmation, and when I requested a replacement for the spinach, it complied promptly.<\/p>\n
<\/p>\n
The process came to a halt due to a significant hurdle: my total was only \u00a010.28 euros\u00a0, while Mercadona’s minimum order was \u00a050 euros\u00a0. I instructed the agent to include ingredients from an additional recipe for \u00a0baked salmon\u00a0. However, that recipe also fell short of the minimum, prompting me to request that it prepare the meal for \u00a0four people\u00a0, specifying that I wanted \u00a0fresh salmon\u00a0 instead of frozen.<\/p>\n
<\/p>\n
The agent adjusted quantities and swapped the salmon for fresh, yet the total still didn’t meet the minimum order. Thus, I suggested it seek out \u00a0recently viral products\u00a0 at Mercadona to fill the gap.<\/p>\n
<\/p>\n
The Purchase is Made for You, but There\u2019s a Problem<\/h2>\n
Upon completing the selections, it was time to review the basket. Surprisingly, the agent had added both regular and \u00a0purple garlic\u00a0. While the normal garlic was acceptable, the purple option created confusion. It mistakenly searched for purple onion, interpreting Mercadona’s term “red onion” as an indicator for purple garlic, demonstrating a lack of precision in ingredient selection.<\/p>\n
Regarding the trending products, my chosen list included a \u00a0makeup advent calendar\u00a0, \u00a0smoked raclette cheese\u00a0, \u00a0cookie nougat\u00a0, and \u00a0pistachio cake\u00a0. The \u00a0total came to 66 euros\u00a0, which exceeded my original budget. While I had not explicitly requested a strict adherence to 50 euros, the final sum felt excessive.<\/p>\n
<\/p>\n
![\"Mercadona's](\"https:\/\/teknomers.com\/en\/wp-content\/uploads\/2025\/10\/1761746684_349_The-ChatGPT-Atlas-agent-made-my-purchase-at-Mercadona-and.jpeg\"\/)
\n <\/div>\n<\/p><\/div>\n<\/p><\/div>\n<\/div>\nWhile the agent had taken charge and completed the shopping for me, it revealed a critical flaw: \u00a0speed\u00a0. I hadn’t anticipated the need for adjustments, like handling minimum orders or modifying quantities, which resulted in a drawn-out process. The total time spent was almost \u00a015 minutes\u00a0, with over \u00a02:14 minutes\u00a0 solely to add \u00a0eight items\u00a0 to the cart. I couldn\u2019t help but feel that I could have completed the order far more efficiently myself.<\/p>\n
<\/p>\n
As for accuracy, I found that the agent made fewer mistakes than I had feared. However, it remains essential to verify the contents of the basket, as it could easily misplace items like garlic instead of onions; and I definitely have enough garlic stashed away.<\/p>\n
<\/p>\n
Much More Practical in Other Scenarios<\/h2>\n
OpenAI had promoted agent mode as useful for various applications, including making purchases. After this experience, it’s apparent that the \u00a0ChatGPT Atlas agent mode\u00a0 has immense potential, but not particularly for shopping. I decided to test its usefulness in another scenario: \u00a0organizing a trip\u00a0.<\/p>\n
I requested the agent to identify \u00a0getaway locations\u00a0 within a two-hour drive from Valencia over the December long weekend, complete with a budget and accommodation options from \u00a0Booking\u00a0 and \u00a0Airbnb\u00a0. To my satisfaction, within six minutes, it provided options for two distinct destinations, neatly arranged in a table that included the price per night and highlights. After selecting a destination, I simply had to provide my personal information to finalize the reservation.<\/p>\n
<\/p>\n
![\"ChatGPT](\"https:\/\/teknomers.com\/en\/wp-content\/uploads\/2025\/10\/1761746684_257_The-ChatGPT-Atlas-agent-made-my-purchase-at-Mercadona-and.jpeg\")
\n Organizing a trip is practical.<\/span>\n <\/div>\n<\/p><\/div>\n
The act of making a purchase is a mechanically simple process that can easily be executed manually within a few minutes. However, when faced with barriers like a minimum order or unclear preferences, the method may consume more time than it saves. On the other hand, the agent excels at more complex tasks requiring multiple steps and information filtering. For instance, organizing travel involves numerous elements\u2014picking a destination, planning routes, finding accommodation, and staying within budget. In such instances, we can allow agent mode to think while we attend to other matters, effectively saving us considerable time.<\/p>\n
<\/p>\n
Images | Amparo Babiloni, Xataka<\/p>\n
In Xataka | Some researchers created a company where all employees were AI agents. They didn’t even complete a quarter of the work.<\/p>\n