---
title: "Why ChatGPT Names One Product and Skips Another"
description: "When shoppers ask AI what to buy, clear product pages are only part of the answer. Here’s how feeds, reviews and independent mentions help a store earn a place on the shortlist."
url: "https://groas.com/post/when-shoppers-ask-chatgpt-what-to-buy-how-product-pages-get"
image: "https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/3b117b2b-aa1d-48dc-ba77-b88c783d6a7a.png"
published: "2026-10-31T06:25:46.365Z"
modified: "2026-10-31T06:25:46.411Z"
---

October 31, 2026 · 11 min read

# Why ChatGPT Names One Product and Skips Another

[Alexander PerelmanHead Of Product @ groas](https://groas.com/author/alexander-perelman)[LinkedIn](https://www.linkedin.com/in/alexander-433793253/)

![Collage of two water bottles: many cut-out hands from reviews and magazines point at one, while the other has only its own hand pointing at itself.](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/3b117b2b-aa1d-48dc-ba77-b88c783d6a7a.png)

In this article

1. [First, separate the product from the seller](#first-separate-the-product-from-the-seller)
2. [Give the assistant a sentence worth using](#give-the-assistant-a-sentence-worth-using)
3. [Earn agreement beyond your own site](#earn-agreement-beyond-your-own-site)
4. [Measure the answer, then the order](#measure-the-answer-then-the-order)
5. [Buy execution, not another visibility deck](#buy-execution-not-another-visibility-deck)
6. [Try the same reasoning on trail socks](#try-the-same-reasoning-on-trail-socks)

A shopper asks ChatGPT for an insulated bottle that will not leak in a gym bag. She gets three names and clicks one. **For that shopper, the answer is the store shelf.** Your ad might be excellent. Your page might rank first on Google. Neither guarantees a place on it.

Here is the first principle everything else follows from: **assistants tend to recommend products that the sources they read keep naming for a particular use.** If roundups, retailer listings and reviews repeatedly point to the same bottle for gym bags, that bottle has a stronger case than one making the claim only on its own page. The store’s job is to make the buying answer clear on its pages and give other people something specific worth confirming.

## First, separate the product from the seller

The bottle question contains two decisions. Which product fits the job? And where can the shopper buy it? Those are related, but they are not the same question. A shopper can see one bottle named in an answer and several sellers attached to its product card.

ChatGPT can draw on product pages, retailer listings, reviews and other material it reads. Merchants can also provide product information through an [OpenAI product feed](https://rithum.com/). The feed gives the assistant structured facts about an item and its listing; pages and outside sources give it language and evidence about the use case. **Neither a feed row nor a first-place Google ranking is a promise of a recommendation.**

That helps explain why the cheapest listing does not automatically win, either. OpenAI says its [shopping results are selected independently and are not ads](https://bigcartenergy.com/). When several stores sell the same product, seller ordering can reflect availability, price, listing quality and whether the seller makes the product or is its primary seller. First the assistant needs a product that fits the question. Then the seller has to give it a listing a shopper can actually use.

I used to tell clients that ranking first was enough. I was wrong. A ranking can put a page in reach; it cannot make other sources agree that a bottle survives a trip sideways in a gym bag. **The practical test is whether the product keeps being named for the job the shopper described.**

![Diagram showing a product recommendation drawing on product pages, reviews and roundups.](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/433c73ad-2106-4218-800c-6473a338e3a2.png)

An [analysis of shopping-agent citations in Q2 2026](https://martechcube.com/) found that ChatGPT drew 41% of the sources it cited from earned media, 37% from retailers, 19% from user discussions and 3% from brand sites. That is a breakdown of cited sources, not a formula for how every answer is chosen. Still, it makes the editorial problem plain: the page you control cannot do all the persuading. A roundup that tested six bottles, a retailer listing with detailed reviews and a discussion about leaky lids may all help establish why one bottle fits this use.

For our bottle, I would ask two separate questions. Does the product page clearly state what the bottle does? Do sources beyond that page name it when gym-bag leaks come up? The first confirms the details. The second helps make the claim credible enough to repeat. **Consensus first, confirmation second.**

## Give the assistant a sentence worth using

Start with the page you own. Most product pages describe the bottle in general terms. A useful page answers the buying question in the buyer’s terms. For the gym bottle, that means addressing who it fits, how its lid seals in a bag, how long it keeps water cold in a hot car and how it compares with the next size up. Those are the doubts inside the purchase, not just features to fit into a grid.

I look for one clear sentence per doubt in the prompt. If the shopper asks for a bottle that “doesn’t leak in bag,” do not make the reader assemble an answer from a photograph, a lid name and a row of icons. State the relevant claim plainly, in words the assistant can quote. Do not claim it survives being laid sideways unless that is true. A crisp sentence is useful only if the product can back it up.

The category page has a different job: comparison. Shoppers ask for the best bottle for the gym under $40, or for steel versus plastic for daily training. A category page that groups products by use and presents weight, capacity, dishwasher safety and lid type in the same order gives each option a fair, readable comparison. It should also say who should skip each one. A bottle that is a poor fit for someone’s bag does not become a better pick because the page avoids saying so.

**Write the product page to answer the doubt; write the category page to distinguish the options.** If both merely repeat the same feature list, neither finishes the sentence the shopper started.

### Make price, stock and variants readable

Copy answers the question. Structured data helps the assistant identify which fact belongs to which product. A price embedded in an image, or stock status that does not render for a crawler, leaves too much to guess. Put [Product and Offer schema on each product page](https://mersel.ai/), with price, currency, availability, aggregate rating and review count marked as separate fields. That is how $34 stays a price rather than getting tangled with a 4.8-star rating.

![Paper-cut product page with price tag, star rating and stock label lifting off in separate layers](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/2384dd64-1c8b-4051-86df-f44cc9fc9516.png)

Apply the same discipline to the feed. The [nine fields requested for each feed row](https://alhena.ai/) are item ID, title, description, URL, brand, seller name, image, availability and price. Optional details about variants, reviews and Q\&A can give the assistant more to compare. Eligibility does not guarantee that an item appears. It does mean the listing is not missing basic facts before the comparison even starts.

For the gym bottle, the page might explain the seal while the feed identifies the bottle, its price and whether it is available. Those are different pieces of the same buying decision. **If the assistant cannot reliably read a claim or a listing fact, do not expect it to use that fact.**

### Put reviews where the claim can be checked

A store can say its bottle is leakproof. Reviews let a shopper check what happens after someone carries it to the gym. On-page reviews help, but retailer reviews and detailed Q\&A matter because buyers leave them somewhere the brand cannot simply rewrite the product description. A useful review talks about the lid, the seal or the dishwasher. “Great bottle” is pleasant. “It stayed dry in my gym bag” answers the question.

For our example, I want the same use-case claim to appear in places with different owners: the store page, a retailer listing and a discussion where people talk about carrying the bottle. Not identical marketing copy pasted three times. Independent descriptions of the same experience. **Repetition becomes useful when it is corroboration, not an echo chamber.**

![Stack of review cards and forum threads feeding into a single trusted product pick.](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/c9d160ca-04e3-4816-b84a-454654339fdf.png)

## Earn agreement beyond your own site

This is where many stores lose the answer. The citation analysis puts brand sites at 3% of the cited sources it examined; most of the cited material came from elsewhere. If you want to see [which off-site pages ChatGPT and Perplexity cite](https://groas.com/post/11-places-chatgpt-and-perplexity-find-so), look at the kinds of pages doing the work: roundups that address the purchase, retailer listings with detail and discussions that test a claim against use.

For the gym bottle, I would keep the target narrow: a roundup that names it for leakproof gym use, a retailer listing where buyers discuss that same use and a thread where gym-goers confirm how the lid behaves. Those are three kinds of evidence for one question. Thirty generic backlinks might help someone find the domain, but they do not necessarily explain why this bottle belongs in a gym bag.

That distinction also keeps outreach honest. A broad roundup about attractive water bottles may be a poor match for a prompt about leaks. The question is not merely, “Did the store get mentioned?” It is, “Did a source name this product for the reason the shopper is buying?” **Win agreement on the use case, not just attention for the category.**

### Let people answer in their own words

User discussions accounted for 19% of citations in that analysis. Forums, threads and long Q\&A answers often contain the exact doubts a product description smooths over. If gym-goers discuss whether a lid stays shut when a bag gets tossed around, they are addressing the shopper’s question more directly than a generic claim about durability.

You cannot buy a convincing conversation, and you should not try to plant one. The less glamorous approach is to answer real questions publicly and specifically. Explain the seal design, the dishwasher test or the warranty replacement rate where those facts are available. Put a detailed comparison on your own blog that someone can examine and link to. Then let owners describe their experience in their own words. A comparison can give a discussion something concrete to inspect; it cannot dictate what that discussion concludes.

This is the point where the analogy of a “store shelf” stops being exact. A physical shopkeeper decides what to stock. An assistant assembles an answer from information it can read, and that answer may change. **Give independent sources a precise claim to examine, not a slogan to repeat.**

## Measure the answer, then the order

Rankings alone will not tell you whether the bottle makes the shortlist. Track a fixed set of buying prompts. I would start with 30 to 60 exact questions buyers ask, from “best insulated bottle for gym under $40” to “steel bottle that does not sweat in car cup holder.” Run the same list monthly across ChatGPT, Perplexity, Gemini and AI Overviews. Record whether the product is mentioned and whether a source is cited.

Keep that prompt measure separate from the commercial one. [Tracking prompts alongside GA4 segments for AI referrers](https://geolikeapro.com/), including chat.openai.com, perplexity.ai and gemini.google.com, lets you compare visibility with sessions, conversion rate and revenue. A name in an answer is progress. An order tells you whether that progress is worth protecting.

[Adobe found that AI-referred retail traffic during the 2025 holidays converted 31% higher, stayed 45% longer and viewed 13% more pages](https://neuraladx.com/) than other sources. That result is a reason to check your own numbers, not to assume your store will match it. If the gym bottle appears in 12 of 50 tracked prompts, look at what happens after shoppers arrive. If those referrals close at twice the site average, protect the work that keeps that bottle visible. If they do not, a mention count alone is not the win.

**Measure prompts for presence and referrers for money.** Do it by product, so a strong result for one bottle does not conceal a weak result for the rest of the catalog.

## Buy execution, not another visibility deck

Most agencies will happily sell a monthly report of mentions. That is the wrong job if the product pages still fail extraction, the feed omits variants and no independent source names the bottle for gym use. A store needs the work connected in order:

1. Rewrite the priority product and category pages so they answer buying prompts clearly.
2. Fix Product and Offer schema and the feed so price, availability and variants parse cleanly.
3. Work toward relevant outside mentions for each hero product.
4. Report prompt coverage beside revenue from AI referrers.

If an agency cannot show how those pieces connect from page change to prompt appearance to order, you are paying for charts. I have written more about [how brands get recommended by AI chatbots](https://groas.com/post/how-to-get-your-brand-recommended-by-ai-chatbots); use that chain to test any pitch, including ours.

That execution gap is why I point stores at groas. Its [SEO and AI Search service](https://groas.com/pricing) handles new pages, technical fixes and authority-building on the sites engines read, with a named strategist responsible for direction. It is offered at a flat $1,499 a month rather than a cut of ad spend. A percent-of-spend fee rises when spend rises; it does not, by itself, get a bottle named in an answer. **The useful question is whether the work improves visibility that turns into orders.** Treat that as an ongoing operation, not a one-time audit.

## Try the same reasoning on trail socks

Now change the shelf. Say you sell merino trail socks. A shopper asks ChatGPT for socks that prevent blisters on 10-mile runs and survive frequent washing. Start where we started with the bottle: identify the job, then ask what the sources would need to agree on. For this example, the claims are a no-blister heel, true sizing after five washes and a price under $25. Do not write any of those on a product page unless the sock can support them.

The page should state the relevant details in plain sentences. The feed should make the size variants, color availability and review count available for comparison. Outside the store, look for a running-site roundup that names the sock for long runs, a retailer listing with reviews that discuss washing and a thread where runners talk about fit. None of those sources is a substitute for the others: the page provides the product facts, while outside accounts test whether those facts matter in use.

I would work through one hero product in this order:

1. Choose one use and 10 to 15 exact buying questions about it.
2. Rewrite its product page and category comparison to answer the doubts those questions contain.
3. Check that schema and the feed present price, stock, variants and ratings without guesswork.
4. Pursue relevant outside mentions, then compare prompt appearances with GA4 revenue from AI referrers.
5. Keep the named product available, price it clearly and make it easy for buyers to leave useful reviews where shoppers already read them.

If the outside sources name three different socks for that long-run question, there is no clear agreement for the assistant to carry into its answer. If they independently keep naming yours for the same reason, the case gets stronger. That is the model working on a new shelf.

The bottle shopper who started this piece does not need to see an ad or scroll ten results. She can pick from three names. Stores that understand [why the old agency economics no longer work](https://groas.com/post/ai-agency-revenue-model-2026-broken-economics-what-businesses-do-instead) stop buying reports about that shelf and start doing the work that earns a place on it: clear pages, clean data, repeated proof. Do it for one product, measure the orders, then repeat. **Consensus first, confirmation second, revenue to prove it.**

## Frequently Asked Questions

### How do AI assistants decide which products to recommend?

AI assistants tend to recommend products that the sources they read repeatedly name for a specific use case. A product must keep being mentioned for the job a shopper described, not just on its own page.

### Is a high ranking on a search engine enough to get a recommendation from ChatGPT?

No, a first-place Google ranking is not a guarantee of a recommendation. A ranking can put a page in reach, but it cannot make other sources agree that a product meets the specific requirements of the use case.

### What types of sources does an AI assistant use when generating a shopping answer?

AI can draw on product pages, retailer listings, reviews, and structured information provided by product feeds. These various sources give the assistant both structured facts and the necessary language and evidence about the use case.

### What is the difference between the goals of a product page and a category page?

Product pages should be written to answer specific buying doubts in the buyer’s terms using clear sentences. Category pages have the job of comparison, grouping products by use and presenting features for a readable distinction.

### What technical elements should be used to help AI read listing information?

Merchants should put Product and Offer schema on each page, marking price, currency, and availability as separate fields. Structured data fields in the feed ensure the assistant can reliably read listing facts before making comparisons.

### Why are independent reviews and user discussions important for product credibility?

Independent sources like retailer listings and user discussions provide corroboration of a product's claims. A claim about a bottle is only credible if sources beyond the brand itself can confirm its performance in use.

### How should businesses track the success of their product visibility in AI?

One should track a fixed set of specific buying prompts across multiple AI tools and record whether the product is mentioned and cited. This allows a business to measure both visibility and the subsequent revenue generated from AI referrers.

## Paid Ads & SEO, Run For You

Businesses pay one flat monthly fee per service, agencies resell it under their own brand, and groas does the work either way.

[See Pricing](https://groas.com/pricing)

## Structured data

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://graph.groas.com/entity/groas","name":"groas","url":"https://groas.com/","description":"groas is a done-for-you marketing service, not software: elite people working with proprietary technology systems run search ads, Meta ads, SEO and AI search for businesses, and white label for agencies, with a dedicated senior strategist owning every account.","logo":{"@type":"ImageObject","url":"https://groas.com/icon-512.png","width":512,"height":512},"sameAs":["https://www.linkedin.com/company/groas/"]},{"@type":"WebSite","@id":"https://groas.com/#website","name":"groas","url":"https://groas.com/","inLanguage":"en-US","publisher":{"@id":"https://graph.groas.com/entity/groas"}},{"@type":"WebPage","@id":"https://groas.com/post/when-shoppers-ask-chatgpt-what-to-buy-how-product-pages-get#webpage","url":"https://groas.com/post/when-shoppers-ask-chatgpt-what-to-buy-how-product-pages-get","name":"Why ChatGPT Names One Product and Skips Another","description":"When shoppers ask AI what to buy, clear product pages are only part of the answer. Here’s how feeds, reviews and independent mentions help a store earn a place on the shortlist.","inLanguage":"en-US","isPartOf":{"@id":"https://groas.com/#website"},"about":{"@id":"https://graph.groas.com/entity/groas"},"primaryImageOfPage":{"@type":"ImageObject","url":"https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/3b117b2b-aa1d-48dc-ba77-b88c783d6a7a.png"},"breadcrumb":{"@id":"https://groas.com/post/when-shoppers-ask-chatgpt-what-to-buy-how-product-pages-get#breadcrumb"},"dateModified":"2026-10-31T06:25:46.411Z","mainEntity":{"@id":"https://groas.com/post/when-shoppers-ask-chatgpt-what-to-buy-how-product-pages-get#article"}},{"@type":"BreadcrumbList","@id":"https://groas.com/post/when-shoppers-ask-chatgpt-what-to-buy-how-product-pages-get#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://groas.com/"},{"@type":"ListItem","position":2,"name":"Blog","item":"https://groas.com/blog"},{"@type":"ListItem","position":3,"name":"Why ChatGPT Names One Product and Skips Another","item":"https://groas.com/post/when-shoppers-ask-chatgpt-what-to-buy-how-product-pages-get"}]},{"@type":"BlogPosting","@id":"https://groas.com/post/when-shoppers-ask-chatgpt-what-to-buy-how-product-pages-get#article","url":"https://groas.com/post/when-shoppers-ask-chatgpt-what-to-buy-how-product-pages-get","mainEntityOfPage":{"@id":"https://groas.com/post/when-shoppers-ask-chatgpt-what-to-buy-how-product-pages-get#webpage"},"isPartOf":{"@id":"https://groas.com/blog#blog"},"headline":"Why ChatGPT Names One Product and Skips Another","description":"When shoppers ask AI what to buy, clear product pages are only part of the answer. Here’s how feeds, reviews and independent mentions help a store earn a place on the shortlist.","datePublished":"2026-10-31T06:25:46.365Z","dateModified":"2026-10-31T06:25:46.411Z","image":{"@type":"ImageObject","url":"https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/3b117b2b-aa1d-48dc-ba77-b88c783d6a7a.png"},"author":{"@id":"https://groas.com/author/alexander-perelman#person"},"publisher":{"@id":"https://graph.groas.com/entity/groas"},"keywords":"product, bottle, answer, page, gym, chatgpt, pages, reviews","wordCount":2394,"inLanguage":"en-US"},{"@type":"Person","@id":"https://groas.com/author/alexander-perelman#person","name":"Alexander Perelman","jobTitle":"Head Of Product @ groas","description":"My background includes Goldman Sachs and Stanford Computer Science.","image":"https://groas.com/media/blog/4f17cac3a81acc224cc1b2cfab8f94bbc295a3089d13e8421125771ef8d4064f.jpg","worksFor":{"@id":"https://graph.groas.com/entity/groas"},"sameAs":["https://www.linkedin.com/in/alexander-433793253/"]},{"@type":"FAQPage","@id":"https://groas.com/post/when-shoppers-ask-chatgpt-what-to-buy-how-product-pages-get#faq","mainEntity":[{"@type":"Question","name":"How do AI assistants decide which products to recommend?","acceptedAnswer":{"@type":"Answer","text":"AI assistants tend to recommend products that the sources they read repeatedly name for a specific use case. A product must keep being mentioned for the job a shopper described, not just on its own page."}},{"@type":"Question","name":"Is a high ranking on a search engine enough to get a recommendation from ChatGPT?","acceptedAnswer":{"@type":"Answer","text":"No, a first-place Google ranking is not a guarantee of a recommendation. A ranking can put a page in reach, but it cannot make other sources agree that a product meets the specific requirements of the use case."}},{"@type":"Question","name":"What types of sources does an AI assistant use when generating a shopping answer?","acceptedAnswer":{"@type":"Answer","text":"AI can draw on product pages, retailer listings, reviews, and structured information provided by product feeds. These various sources give the assistant both structured facts and the necessary language and evidence about the use case."}},{"@type":"Question","name":"What is the difference between the goals of a product page and a category page?","acceptedAnswer":{"@type":"Answer","text":"Product pages should be written to answer specific buying doubts in the buyer’s terms using clear sentences. Category pages have the job of comparison, grouping products by use and presenting features for a readable distinction."}},{"@type":"Question","name":"What technical elements should be used to help AI read listing information?","acceptedAnswer":{"@type":"Answer","text":"Merchants should put Product and Offer schema on each page, marking price, currency, and availability as separate fields. Structured data fields in the feed ensure the assistant can reliably read listing facts before making comparisons."}},{"@type":"Question","name":"Why are independent reviews and user discussions important for product credibility?","acceptedAnswer":{"@type":"Answer","text":"Independent sources like retailer listings and user discussions provide corroboration of a product's claims. A claim about a bottle is only credible if sources beyond the brand itself can confirm its performance in use."}},{"@type":"Question","name":"How should businesses track the success of their product visibility in AI?","acceptedAnswer":{"@type":"Answer","text":"One should track a fixed set of specific buying prompts across multiple AI tools and record whether the product is mentioned and cited. This allows a business to measure both visibility and the subsequent revenue generated from AI referrers."}}]}]}
```