WIRED showed how ChatGPT distorts editors’ recommendations on the best laptops, TVs, and headphones

A WIRED journalist tested how accurately ChatGPT can convey the editorial's tech recommendations and discovered that the bot regularly makes mistakes. Even when the model provides a link to the correct guide, it can substitute the editors' actual choice with its own version.

What WIRED Tested

OpenAI's new push toward an AI-shopping scenario prompted the test. The company promises that ChatGPT will help users choose products faster and free them from endless tab comparisons and "best of" lists. WIRED decided to test not the abstract quality of advice, but a narrower and measurable case: can the bot accurately recount which devices the editorial has actually tested and currently recommends. For such a test, this is a convenient target because the publication has regularly updated guides built on real reviews and hours of practical testing.

The questions were straightforward: which televisions, wireless headphones, and laptops does WIRED consider the best right now. The key condition was this: the answer should reflect the choice of WIRED reviewers, not general market analogues or "similar class" models. In other words, ChatGPT was required not to invent a convenient summary from memory, but to carefully recount already published material. At this basic level, the model began to malfunction, although visually its answers looked convincing and were accompanied by correct links.

Where Answers Failed

The most telling failure occurred in the television category. ChatGPT opened the correct WIRED guide, but named the LG QNED Evo Mini-LED as the best option for most people — a model that wasn't on the editorial list at all. WIRED's actual choice was the TCL QM6K.

The error looked especially unfortunate because a reader could easily miss the substitution: the link goes where it should, product photos are nearby, the tone of the answer is confident, and it feels like a brief summary of the article. But in fact, the bot had already mixed in its own interpretation on top of the editorial work.

A similar story repeated in other categories. In the headphones selection, ChatGPT listed AirPods Max 2 as WIRED's recommendation for Apple ecosystem users, although at the time of testing the editorial hadn't yet reviewed them and hadn't added them to the guide. With laptops, the bot stubbornly named last year's MacBook Air M4 as the best choice, although the current top pick in the guide was already the MacBook Air M5 from 2026.

When the journalist pointed out the discrepancies, the model itself admitted that it had replaced precise recommendations with more general and sometimes outdated options.

• Televisions: the bot substituted WIRED's main choice with a model not in the guide.
• Headphones: attributed to the editorial advice to buy a device that hadn't yet passed their tests.
• Laptops: presented the outdated MacBook Air M4 instead of the current MacBook Air M5.
• In all cases: the correct link coexisted with an incorrect paraphrase.

"I took WIRED's real choice and replaced it with a more general similar option,"

ChatGPT admitted after the remark.

Why This Matters

The problem here isn't just the factual error itself. When someone comes with the request "what does WIRED recommend," they want not an averaged model response, but the result of a specific editorial process: tests, comparisons, updates, and the author's final choice. When ChatGPT substitutes one product for another, the user might take it as the publication's position and spend money relying on authority that doesn't actually exist in this specific recommendation. For media, this is already a question not of interface convenience, but of attribution accuracy and brand trust.

There's also a business side. WIRED explicitly states that the editorial earns from affiliate links, and this money helps fund reviews and extensive tech testing. In ChatGPT's responses built on top of the publication's materials, there are no such links, although the model itself relies on the work of editors and simultaneously reduces the motivation to visit the website.

The situation looks particularly ironic against the backdrop of Condé Nast's deal with OpenAI, which is supposed to help display links to publisher materials inside the chatbot. Formally the link is there, but there's little practical benefit if an incorrect conclusion stands next to it.

What This Means

The WIRED story shows that even an updated AI-shopping scenario still struggles with the task of accurately recounting other people's recommendations. If you need the conclusion of a specific editorial, lab, or test team, it's still better to go to the original source. For OpenAI and other players, this is a signal: in commerce scenarios, what matters is not just convenient delivery, but literal accuracy to the data.