ChatGPT doesn't shop on Amazon: what AI changes for brands

By the Finnex Agency team

Short answer. ChatGPT is not allowed to read your Amazon listing: since November 2025, amazon.com's robots.txt file blocks all three OpenAI bots (GPTBot, ChatGPT-User and OAI-SearchBot). And in-chat purchasing, the much-discussed Instant Checkout, was deprioritized on March 24, 2026, when OpenAI announced that merchants can use their own checkout. For a brand selling on Amazon, AI works on two separate fronts: inside Amazon, Alexa for Shopping decides (the assistant that was called Rufus until May 2026), and outside Amazon, ChatGPT forms its view from what it finds on your website, in third-party reviews and in the press. They are two different jobs, and it pays not to mix them up.

Over the past year, one idea got repeated a lot: that conversations with AI would replace the store, and that brands would sell directly inside ChatGPT. The general direction makes sense, because more and more shoppers start their search by asking an assistant. The problem is that the specific mechanism behind that story changed, and almost nobody updated it.

In this article we lay out what happened, with dates and sources, and what it means in practice for a private-label brand selling on Amazon in the United States, the United Kingdom or Spain. We look at it with a focus on supplements and wellness, the category where AI gets the most questions and where bad information costs the most.

---

1. What matters, before we start

If you're short on time, here's what you need to know. ChatGPT doesn't see your Amazon listing, so what it says about your product depends on what exists outside Amazon. Inside Amazon, Amazon's own assistant already recommends, compares and even buys on its own. And the two are worked with different tools.

Three definitions so the rest of the article reads clearly:

• Alexa for Shopping: Amazon's AI shopping assistant. It was called Rufus until May 13, 2026, when Amazon renamed it. It answers using the product catalog, customer reviews, community Q&A and information from across the web.
• Agentic Commerce Protocol (ACP): the standard OpenAI uses to receive product catalogs from merchants and show them in ChatGPT. Target, Sephora, Best Buy and Wayfair are among the retailers using it.
• GEO (Generative Engine Optimization): the work of making sure an AI assistant understands your product, can verify what you say about it and has reasons to mention it when someone asks.

None of these three things guarantees a recommendation. What you can do is raise the odds that the AI has correct and sufficient information about your brand, which is already an edge over most brands in the category, since most aren't doing it.

---

2. What the story promised and what actually happened

On September 29, 2025, OpenAI launched Instant Checkout with Stripe, Etsy and Shopify. The idea was that a US user could discover a product and pay for it without leaving the conversation. Several industry analyses framed it as the beginning of the end of the traditional store.

Six months later, the picture was different. On March 24, 2026, OpenAI wrote that the initial version of Instant Checkout didn't offer the flexibility it wanted, that merchants could move to their own checkout processes, and that the company would focus its efforts on product discovery. In most cases, the purchase goes back to being completed in the merchant's own store.

Our reading is simple: AI didn't stop mattering for shopping, but the place where the sale is decided didn't move to ChatGPT. It split. Part of the decision is made in a conversation with ChatGPT outside Amazon, and another, growing part is made in a conversation with Amazon's assistant inside Amazon.

---

3. ChatGPT doesn't see your Amazon listing

This is the point that changes strategy the most, and the one that gets explained the least. The amazon.com robots.txt file includes a Disallow: / rule for GPTBot, ChatGPT-User and OAI-SearchBot. Those are the three agents OpenAI uses for training, for fetching live information when a user asks a question, and for its search engine. We checked on October 19, 2026, and the rule is still there.

A robots.txt file is an instruction, not a technical or legal barrier, but it makes Amazon's position clear: it doesn't want OpenAI reading its catalog. In practice, when someone asks ChatGPT for the best magnesium for sleep, you shouldn't count on it reading your bullets, your images or your Amazon reviews. It may name a brand that sells on Amazon, but what it knows about that brand came from somewhere else.

That somewhere else is where your visibility in ChatGPT is decided:

• Your own website, if you have one and AI bots can read it.
• Other retailers that sell your product and share their catalog with OpenAI.
• Independent reviews and comparisons, press, blogs and forums.
• Structured data that describes the product in a way a machine can understand without ambiguity.

For a brand that only exists inside Amazon, the conclusion is uncomfortable: to ChatGPT, your product has almost no information of its own. If something has been said about it, a third party said it, and not necessarily with the right data.

---

4. Inside Amazon, the AI that decides is Alexa for Shopping

While the public debate focused on ChatGPT, Amazon was turning its own assistant into a core part of the shopping experience. According to Amazon, by November 2025 more than 250 million customers had used it that year, and shoppers who use it are more than 60% more likely to buy during that session.

What matters for a brand isn't the new name, it's where the information comes from. Amazon explains that the assistant combines the product catalog, customer reviews, community Q&A and information from across the web. In other words, it answers with what you wrote in the listing, what your customers wrote in their reviews and what is said about your product elsewhere.

It also no longer just answers questions. According to the same Amazon announcement, it can track a product's price, alert the customer when it hits a target price and, for Prime members, buy it automatically when it does. It can also add previously purchased products back to the cart. For replenishment categories like supplements, this changes how a customer is lost or retained.

---

5. Two fronts, two different jobs

Mixing up these two fronts is the most common mistake we see. Brands try to "show up in ChatGPT" by rewriting their Amazon listing, which ChatGPT doesn't read, or try to improve with Amazon's assistant by working only on their website, which carries much less weight there.

The two columns meet at one point: consistency. If your website says 400 mg per capsule and the listing says 500 mg, neither assistant will trust the figure. And in supplements, a number that doesn't add up is a reason for the AI to pick another brand.

---

6. What changes for a supplement brand

Supplements are one of the categories where shoppers ask an AI the most questions before buying. The questions are specific: what dose, who it's for, what not to combine it with, whether it has certifications, how long it takes to notice an effect. An assistant that can't find those answers in your information will look for them in a competitor's.

In practice, we see four concrete changes:

Questions replace keywords. The shopper no longer types "magnesium glycinate 400 mg". They type "which magnesium should I take if I have trouble sleeping and I drink coffee". The listing has to answer that question explicitly, without relying on the assistant to infer it.

Reviews become content. Amazon's assistant summarizes what your customers say. If a pattern of complaints repeats, for example about capsule size or taste, it will show up in the answer even if your listing never mentions it.

Claims have to be defensible. An AI repeats what it reads. If a claim on your website or listing can't be backed up, the risk is no longer just that Amazon flags it: it can be reproduced in answers you don't control. This ties directly into compliance in the category.

Replenishment becomes automated. If the assistant buys automatically when the price reaches a target, a price increase or a stockout can send the customer to another brand without anyone making a conscious decision. Price and inventory stability become a retention lever.

---

7. What to review this month

You don't need to redo everything. This is the list we use for a first review:

1. Your website and AI bots. Check your own site's robots.txt. If it blocks GPTBot, OAI-SearchBot or ChatGPT-User, ChatGPT can't read you there either, which leaves it with no first-party source about your brand.
2. A product page on your website for each main product, with the same data as the listing: dose, ingredients, format, certifications and who it's for.
3. Data consistency across sources. Compare your website, listing and other retailers field by field. Every mismatch is a reason for the AI to distrust you.
4. The listing as an answer. Take the ten questions customers ask most, in reviews and in the Q&A section, and check that the listing answers them explicitly.
5. Review patterns. Group negative reviews by theme. Whatever repeats is what Amazon's assistant will summarize.
6. Price and stock on replenishment products. Identify the ASINs where a stockout would cost you repeat customers.
7. A simple test. Ask ChatGPT the questions your buyer would ask, without naming your brand, and note which brands show up and with what data. Repeat it every month to track the trend.

---

8. Frequently asked questions

Can you sell directly inside ChatGPT?

In some cases it's still possible, but it's no longer the main path. Since March 24, 2026, OpenAI has prioritized product discovery and lets merchants close the sale in their own store. For Amazon sellers, there is no integration that lets you sell an Amazon listing inside ChatGPT.

Can ChatGPT recommend a product that's only sold on Amazon?

It can mention it, but it shouldn't be relying on the listing's information, because Amazon blocks its bots. If it mentions the product, it's because it found information in other sources. The better and more consistent that external information is, the better you'll be represented.

What happened to Rufus?

On May 13, 2026, Amazon renamed it Alexa for Shopping. It's still Amazon's shopping assistant, and it still uses the catalog, reviews, community Q&A and web information to answer.

Does optimizing for AI replace Amazon SEO?

No. Ranking in Amazon search still depends on relevance and sales. What changes is that the listing, besides ranking, has to be able to answer questions in natural language, because that's what the assistant reads and summarizes.

Is there any way to guarantee that AI will recommend my product?

No. No agency or tool can guarantee it. What you can do is make sure the AI has correct, verifiable and consistent information about your brand, which is the precondition for it to consider you at all.

---

To wrap up

AI is changing how people shop, but not in the way that was announced in 2025. The sale didn't move to ChatGPT. It split between two assistants that read different sources: ChatGPT, which doesn't see your Amazon presence, and Alexa for Shopping, which lives inside Amazon and already buys on its own.

For a supplement brand, that translates into two concrete tasks. Outside Amazon, build a first-party source about your product that is readable and consistent. Inside Amazon, turn your listing and reviews into clear answers to your buyer's real questions. Both share one goal: when an AI talks about your product, it should say the right thing.

This article reflects publicly available information as of October 19, 2026. AI assistant features change frequently, so we will keep it updated.

---

Keep reading

• 7 Amazon Listing Mistakes Killing Your Conversion Rate
• How to Stop a Bad Review from Killing Your Amazon Conversion
• Supplements on Amazon US: compliance without losing sales

---

If you sell on Amazon and don't know what AI is reading about your brand, inside and outside Amazon, you can request your free audit. We review your listing the way Amazon's assistant reads it, data consistency between your website and Amazon, and whether your site is readable by AI bots.