I Inquired with ChatGPT About Recommendations from WIRED’s Reviewers, but Its Responses Were Incorrect

WIRED’s Gear Reviews team ranks among the finest, evaluating a wide range of products to assist you in finding the best options. These buying guides and reviews require extensive hands-on testing and continuous updates to guarantee that readers, like you, searching for headphones or running shoes, have the most current information while shopping. (WIRED may also receive affiliate commissions when readers click on certain retailer links to purchase recommended products.)

In previous tests, AI tools, including ChatGPT, have typically underperformed in product recommendations. However, OpenAI has recently enhanced its product recommendation capabilities within ChatGPT to deliver a more comprehensive user experience. This allows you to spend more time interacting with the chatbot and less time sifting through websites and conducting your own research. An increasing number of people are integrating AI into their online shopping experiences, prompting me to evaluate ChatGPT’s current performance.

OpenAI asserts that improvements are being made to its product discovery tools. Nevertheless, based on my tests, the most reliable method to find out what WIRED reviews actually say about a product remains visiting the website. ChatGPT often generated mistakes or provided unrelated product options when asked about WIRED’s recommendations across various categories.

When I sought comment, an OpenAI representative directed me to a recent blog discussing the new AI shopping assistant feature in ChatGPT. “Shopping online is straightforward if you already know what you want,” the announcement states. “However, when making decisions, it often involves toggling between tabs, reading repetitive ‘best of’ lists, and assembling the right information. ChatGPT addresses that: determining what to purchase.”

Condé Nast, the parent company of WIRED, has a business arrangement with OpenAI allowing links to its website to appear in the chatbot. Despite this collaboration, OpenAI seems to undervalue the work of human reviewers, diminishing the importance of these “best” lists as an inconvenience that readers need not consult directly. If you disregard these lists, you might mistakenly believe a product was recommended by WIRED reviewers, when in fact ChatGPT included its own selection.

The Best TVs

One thing about generative AI that hasn’t changed in recent years is how confidently incorrect a chatbot can be in its responses. When I inquired about the best TVs to purchase according to WIRED reviewers, ChatGPT did link to the appropriate buying guide. However, the very first TV it suggested as the top overall choice was the LG QNED Evo Mini‑LED, which isn’t featured in WIRED’s guide at all.

If you were quickly scrolling through ChatGPT’s responses and glancing at the images, you could easily miss this error. When I pointed out the mistake, ChatGPT acknowledged its fault directly: “I took WIRED’s actual top pick (the TCL QM6K) and substituted it with a more generic ‘similar category’ Mini-LED option. That’s not faithful to your request for what WIRED reviewers recommend.”

As more individuals experiment with generative AI as a search tool, errors like these could undermine reader trust when they assume they are purchasing a publisher’s top pick—whether from WIRED, Consumer Reports, or Wirecutter—and end up with a TV that’s not even part of their recommendations.

What About Headphones?

A similar unexpected selection emerged when I asked for the best wireless headphones to buy now, according to WIRED’s reviewers.

ChatGPT suggested that Apple’s AirPods Max 2 are WIRED’s top choice for those deeply embedded in the Apple ecosystem. This might be accurate in a few weeks—after we’ve tested the headphones—but our reviewers haven’t yet included them in the guide; ChatGPT acted prematurely. Only products that our reviewers can physically test and use can be added as recommendations.