WIRED在2026年4月1日的测试文中以Reece Rogers为作者，对ChatGPT在三个类别中的推荐回应进行对比，结果显示其对WIRED审稿人实际推荐的还原率接近零。WIRED的Gear Reviews主张其产品评论依赖大量实机测试与持续更新，属于可追溯的来源。虽然OpenAI宣称已优化ChatGPT的产品发现流程，但在「TV、无线耳机、笔电」三组验证中，回应都偏离了官方推荐，形成每3个类别中全错的趋势。
在电视测试中，ChatGPT虽连结到正确的 WIRED「最佳大萤幕」购物导览，却把LG QNED Evo Mini-LED列为总体第一，但该机型在导览中根本没有登场；实际对应应为TCL QM6K。对于无线耳机，模型将AirPods Max 2描述为深度生态使用者的首选，实际上WIRED尚未完成上手测试前不能列入推荐，且还将产品公告误当成测评内容。这两类都出现「看似合理却脱离来源」的幻觉式插播。
在笔电测试中，WIRED最新榜首是Apple MacBook Air (M5, 2026)，但ChatGPT反复输出旧模型MacBook Air (M4, 2025)，并承认先前是用过时框架推导并臆测了Windows排名。ChatGPT常在连结到正确页面后仍产生过时或凭空补全，显示抽取与核对流程缺口。更严重的是，推荐清单本身不含附属连结时，AI中转会抽走导流流量，压缩出版商可透过导购分润维持深度测评的能力。最终结论是：若要准确知道WIRED或任何测评媒体到底推荐什么，仍以直接读原站最可靠。
In Reece Rogers’ WIRED article dated April 1, 2026, a controlled test across three categories found that ChatGPT did not reliably reproduce WIRED reviewers’ actual recommendations. WIRED says its Gear Reviews are based on hands-on testing and frequent updates, while OpenAI says ChatGPT has improved product discovery. Yet across TV, wireless headphones, and laptops, ChatGPT’s recommendation mapping was wrong in all cases, making the success rate effectively 0/3 for top-pick accuracy.
For TVs, ChatGPT linked to WIRED’s “best large TVs” guide but introduced LG QNED Evo Mini-LED as the overall best pick, even though it is not in the guide; the WIRED top pick was TCL QM6K. For wireless headphones, it framed Apple AirPods Max 2 as WIRED’s recommendation for users in the Apple ecosystem, but WIRED had not yet reviewed it at the time. The model also treated a launch-news story as if it were a hands-on judgement. Both cases show confident additions that were not grounded in WIRED’s published evaluations.
For laptops, the current WIRED top pick is Apple MacBook Air (M5, 2026), while ChatGPT repeatedly claimed MacBook Air (M4, 2025) instead. It later implied it had relied on outdated framing and guessed Windows ordering without strict verification. The pattern is systemic: it can link to the right page yet output outdated or fabricated ranking details. Because AI referrals can siphon traffic away from publishers, and because affiliate-supported listings are often central to funding deep testing, the article argues the safest way to trust recommendations is to consult the source directly rather than intermediated AI summaries.