OpenAI 推出一款基于 GPT-5 mini 的免费购物研究工具，旨在节日期间为用户生成个性化购物指南。与普通 ChatGPT 对话不同，该工具通过专用入口运行，会先以问答形式收集预算、颜色、尺寸等偏好，再逐步给出 10–15 个候选商品并让用户标记“更像这样”或“不感兴趣”以缩小范围。该模型特别强调引用高质量信息源，如 Reddit 等被视为较可信的用户评价，而非付费营销内容。OpenAI 表示，此功能对电子产品、美妆、家居、厨房与户外用品等信息密集型品类表现最佳，但生成完整清单可能需数分钟，简单查询仍建议使用常规 ChatGPT。
该功能在 AI 购物竞争加剧的背景下推出。OpenAI 近期陆续上线群聊、浏览器版 ChatGPT 以及面向 K-12 的免费版本，试图扩大使用场景。竞争者 Perplexity AI 等也在争抢电商导购入口，但 OpenAI 暂未将此次工具商业化，计划在 1 月前向所有免费与付费用户开放“几乎无限使用”。它与 Stripe 支持的应用内结账分离运作，不按联盟链接点击收取佣金，也不会与零售商共享聊天内容；其做法是汇总网络评价并引用商品页与评测页链接。
OpenAI 承认该模型仍有不足，尤其是在自动识别真实、未付费用户评测方面难以做到 100% 准确。团队强调模型已被训练更关注自然内容来源，但仍可能在价格、库存等细节上出错，并建议用户最终以商家页面为准。总体来看，该工具强调更精细的偏好搜集与信息过滤，以期在复杂品类中提供高质量的购物研究体验。
OpenAI has released a free GPT-5-mini–based shopping research tool designed to generate personalized buying guides during the holiday season. Unlike standard ChatGPT conversations, the tool uses a dedicated interface that begins by asking clarifying questions about budget, color, size and other preferences, then iteratively presents 10–15 candidate items for users to refine with “more like this” or “not interested.” The model prioritizes high-quality sources such as Reddit over paid marketing content. OpenAI says the tool performs especially well in detail-heavy categories such as electronics, beauty, home, kitchen and outdoor goods, though compiling a full list may take minutes; simpler queries should still use regular ChatGPT.
The feature arrives amid intensifying AI competition in shopping. OpenAI has recently rolled out group chats, a browser-integrated ChatGPT, and a free K-12 version to expand usage. Rivals like Perplexity AI are also targeting shopping workflows, but OpenAI is not monetizing the tool yet and offers “nearly unlimited usage” to all free and paid users before January. It operates separately from Stripe-powered in-chat checkout and does not rely on affiliate-link commissions or share chat data with retailers; instead, it aggregates reviews across the web and cites links to product or review pages.
OpenAI acknowledges limitations, especially the difficulty of reliably identifying authentic, unpaid user reviews. The model has been trained to favor organic sources, yet mistakes on product details such as prices or availability remain possible, and users are advised to verify information on merchant sites. Overall, the tool emphasizes granular preference collection and broader review synthesis to improve research quality for complex shopping categories.