Introducing shopping research in ChatGPT
OpenAI News今天我们推出了一项名为 shopping research 的新功能,让 ChatGPT 为你做购物调研,帮你找到合适的商品。无需翻阅几十个网站,只要描述你的需求——比如“为小公寓挑一款最安静的无绳吸尘棒”、“帮我在这三款自行车里选一款”,或者“给我四岁、爱画画的外甥女买礼物”—— shopping research 会在几分钟内为你生成一份周到的购买指南。它会提出恰当的澄清问题,深度检索互联网、审阅优质来源,并基于 ChatGPT 以往与你的对话和记忆,为你量身推荐。
从今天起,已登录的 ChatGPT 用户在移动端和网页版即可逐步使用该功能,适用于 Free 、 Go 、 Plus 和 Pro 套餐。为方便节日购物,我们在节日期间向所有套餐开放几乎无限的使用次数。
找到合适的商品
数以亿计的人使用 ChatGPT 查找、理解和比较商品,他们需要帮助来厘清选项,并判断哪些更符合需求、预算和偏好。 shopping research 正是为这种需要更深层决策的场景设计:把商品发现变成一次对话——通过提问了解你的重点,从高质量来源抓取准确、最新的细节,并把选项呈现给你以便进一步筛选。它在信息密集的品类表现尤其好,比如电子产品、美妆、家居与园艺、厨房电器、体育与户外等。
对于像查价格或确认某个功能这类简单问题,普通的 ChatGPT 回答已经足够快捷;但当你需要更深入的比较、约束条件或权衡分析时, shopping research 会花几分钟时间给出更详尽、调研充分的答案。
发现新产品、寻相似款、并排比较与选礼物
无论是“帮我找一台 1000 美元以下、屏幕 15 英寸以上、适合玩游戏的高性能笔记本”,还是“我要一条像这样但到地、价低于 250 美元的连衣裙”,或是“三款儿童推车里帮我比较,要求轻便、易折叠、能在城市人行道及较差路况通行,且宝宝坐着舒适(靠背和遮阳棚好)”,以及“给喜欢钓鱼但老是空手而归的父亲买什么礼物”,再到“帮我找这些运动鞋在 Black Friday 时最划算的折扣,我是学生,告诉我是否有学生优惠码”—— shopping research 都能按你的需求检索并整理对比,指出各自的优劣和取舍。
与此同时,我们也把 shopping research 集成到 ChatGPT Pulse 中(目前对 ChatGPT Pro 用户可用)。当相关时, Pulse 会基于你过去的对话主动推荐个性化的购买指南;例如,你如果一直在聊电动自行车,未来的 Pulse 卡片可能会主动提示适合的配件。
如何使用
提出一个购物问题, ChatGPT 会自动推荐使用 shopping research。点开即可开始,或者在“+”菜单里选择“ shopping research ”。
界面会以可视化的形式展开,你可以在对话中描述需求,并就预算、使用对象、关心的功能等问题作出回答。如果开启了记忆,研究还能进一步个性化:例如当 ChatGPT 知道你喜欢玩游戏,会在找笔记本时把这点考虑进去。
在后台,系统会在互联网上检索最新信息——价格、库存、评测、规格和图片等,并将候选商品实时呈现给你。你可以通过与建议的商品互动来引导研究进程,例如把不喜欢的标为“ Not interested ”,把想要相似款的标为“ More like this ”,研究会根据你的即时反馈调整方向。
几分钟后,你会收到一份个性化的购买指南,列出最优商品、关键差异、各自的权衡点,以及来自可靠零售商的最新信息——这些结论通常能替你省去大量对比、阅读和查证的时间。若要购买,可直接跳转到零售商网站;未来在可用的情况下,你也能通过 ChatGPT 内对接的 Instant Checkout 完成购物。
技术原理
shopping research 由一个专门用于购物任务并经过强化学习训练的 GPT‑5 mini 驱动。我们训练模型去阅读可信网站、引用可靠来源,并对多处信息进行综合以产出高质量的商品调研。这个交互式体验能实时更新和精炼研究内容——加入新约束、根据用户对商品偏好的反馈调整结果,从而让回应既有扎实的调研依据又具个性化色彩。
(注: shopping research 是在 GPT‑5‑Thinking‑mini 基础上做后训练的 mini 模型。我们为此创建了一套针对多约束商品发现问题的评估体系,用以衡量响应中商品满足用户要求(如价格、颜色、材质、规格等)的准确率。)
透明与信任
shopping research 的设计以透明和有用为目标。你的对话不会与零售商共享。结果来源于公开可得的零售网站——直接阅读商品页面、引用来源,并尽量回避质量低下或垃圾站点。希望确保能出现在搜索结果中的商家,可按官方流程申请白名单。
局限性
尽管该模型在内部评测中比我们的其他模型更擅长准确引用网页上的商品细节,但仍非完美。 shopping research 可能在价格或库存等细节上出现错误,我们建议你在下单前到商家官网核实最准确的信息。
这只是个开始
我们的目标是让选购合适商品变得更简单。未来 ChatGPT 会在理解你的偏好、覆盖更多品类以及提供更直观的比较与发现方式上持续进步。
Today, we’re introducing shopping research, a new experience in ChatGPT that does the research for you to help you find the right products. Instead of sifting through dozens of sites, you can just describe what you’re looking for—“Find the quietest cordless stick vacuum for a small apartment,” “Help me choose between these three bikes,” or “I need a gift for my four year old niece who loves art”—and shopping research builds a thoughtful guide to help you decide. It asks smart clarifying questions, researches deeply across the internet, reviews quality sources, and builds on ChatGPT’s understanding of you from past conversations and your ChatGPT memory to deliver a personalized buyer’s guide in minutes.
Shopping research is starting to roll out today on mobile and web for logged-in ChatGPT users on Free, Go, Plus, and Pro plans. To help with holiday shopping, we’re making nearly unlimited usage available to all plans through the holidays.
Finding the right products
Hundreds of millions of people use ChatGPT to find, understand, and compare products. They want help making sense of the options and deciding what fits their needs, budget, and preferences.
Shopping research is built for that deeper kind of decision-making. It turns product discovery into a conversation: asking smart questions to understand what you care about, pulling accurate, up-to-date details from high-quality sources, and bringing options back to you to refine the results. It performs especially well in detail-heavy categories like electronics, beauty, home and garden, kitchen and appliances, and sports and outdoor.
For simple shopping questions like checking a price or confirming a feature, a regular ChatGPT response is quick and all you need. But when you want depth—comparisons, constraints, tradeoffs—shopping research takes a few minutes to give you a more detailed, well-researched answer.
Discovering new products
Help me find a powerful new laptop suitable for gaming under $1000 with a screen that’s over 15 inches
Finding lookalikes
I’m looking for a dress that looks like this, but is full length and less than $250
Comparing items side by side
I need a stroller. Something lightweight, compact, and easy to fold. Comfy for my toddler (good recline and canopy), can handle city sidewalks and some rougher terrain, and fit a mid to slightly-premium budget.
Please compare these three models and share the best fit and trade-offs.
Choosing gifts
I need a gift for my dad who is a fishing fanatic, but never seems to catch anything.
Finding deals
Help me find the best black friday deal for these sneakers. I'm a student, tell me if I qualify for any discount codes
We’re also making shopping research available in ChatGPT Pulse, which is currently available to ChatGPT Pro users. When relevant, Pulse can suggest personalized buyer’s guides proactively based on your past conversations. For example, if you’ve been discussing e-bikes, a future Pulse card could suggest useful accessories for you.
How to use it
Ask a shopping question and ChatGPT will suggest shopping research automatically. Tap it to begin, or you can also select ‘shopping research’ from the (+) menu.

Describe what you need
When you start, ChatGPT opens a visual interface where you can chat and share feedback on product options to guide the research. You’ll explain what you’re looking for, and answer clarifying questions about things like your budget, who the item is for, or which features you care about. If you have memory turned on, the research can be tailored even further. For example, if ChatGPT knows that you’re into gaming, it can factor that in when helping you find a new laptop.
Guide the research
In the background it looks across the internet for up-to-date information like price, availability, reviews, specs, and images, and brings options back to you as it goes. You can guide the research along the way by interacting with suggested products and marking items as “Not interested” or “More like this.” This allows the research to adapt based on your real-time feedback.
Receive a personalized buyer’s guide
After a few minutes, you’ll receive a personalized buyer’s guide with the top products, key differences, tradeoffs, and up-to-date information from reliable retailers. It’s a clear summary that normally would take a lot of comparing, reading, and checking on your own.
If you’d like to purchase an item, you can click through to the retailers site to do so. In the future, you’ll be able to purchase directly through ChatGPT for merchants who are part of Instant Checkout where available.
How it works
Shopping research is powered by a version of GPT‑5 mini trained with reinforcement learning specifically for shopping tasks. We trained it to read trusted sites, cite reliable sources, and synthesize information across many sources to produce high-quality product research. We also designed it to be an interactive experience that could update and refine its research in real time—incorporating new constraints and adjusting to feedback on user product preferences—resulting in a response that feels both well-researched and personalized.
*Shopping research is a mini model post-trained on GPT‑5-Thinking-mini. We created a new evaluation composed of difficult product discovery queries with many constraints. We measure product accuracy as the percentage of products in each response that meet the user requirements (e.g. price, color, material, specs, etc).
Transparency and trust
Shopping research is designed to be transparent and helpful. Your chats are never shared with retailers. Results are organic and based on publicly available retail sites—reading product pages directly, citing sources, and avoiding low-quality or spammy sites. Merchants who want to ensure they are available to appear in shopping research results can follow the allowlisting process here.
Limitations
While the model powering shopping research performs better relative to our other models on internal evaluations that measure its ability to accurately cite product details from the web, it is still not perfect. Shopping research might make mistakes about product details like price and availability, and we encourage you to visit the merchant site for the most accurate details.
Just the start
Our goal is to help simplify finding the right products. ChatGPT will continue to get better at understanding your preferences, covering more categories and offering more intuitive ways to compare and discover products.
Generated by RSStT. The copyright belongs to the original author.