Ecomrank.ai
Start Free
Start Free
Blog

How to rank on ChatGPT: how ChatGPT Shopping picks products to show

19 min read
Illustration of a haberdashery counter seen from the front, three balls of yarn side by side on a wooden board, a hand lifting one of them towards a glowing circle, a wicker basket and a pair of scissors nearby, wide empty space at the bottom

A shopper asks ChatGPT which mug works best for filter coffee, and a list of products comes back with prices and links. Every store owner then asks the same thing: how to rank on ChatGPT, how ChatGPT Shopping picks products, and why it is never yours. The honest answer starts with a correction: there is no results page inside ChatGPT, so there is no rank to win. What exists is eligibility, and OpenAI documents a good part of it in public. This page only covers that: what is written down in its own documentation (all pages read on 27 September 2026), what it means for your catalog, and what nobody can promise you.

TL;DR

  • OpenAI describes its product results as organic. They are not ads and they cannot be bought.
  • Four factors are named for ranking merchants that sell the same item: availability, price, quality, and whether you are the maker or primary seller.
  • There are three ways product data reaches ChatGPT. Direct product feeds are limited to approved partners, so for a small store the open door is your public product page.
  • OpenAI documents three user agents. Only one of them, OAI-SearchBot, decides whether your pages can appear in ChatGPT search answers, and it is controlled from your robots.txt.
  • Shopify stores start ahead: their product data already flows through the Shopify Catalog, independently of robots.txt, with no work required from the merchant.
  • There is no results page, no position and no guarantee. What you control is whether your page is readable and complete.

Table of contents

What ChatGPT Shopping shows, and what it is not

When a question looks like buying intent, ChatGPT can show products with price, availability and a link to the store. OpenAI describes these results as organic: its help center page Shopping with ChatGPT Search says product results "are selected independently by ChatGPT and are not ads, nor influenced by any OpenAI partnerships". The same page lists what ChatGPT considers when deciding which products to surface: structured metadata from first-party and third-party providers (price and product description are its examples) and other third-party content, the model's own response before any new search results, and OpenAI's safety and product policies.

That matters if you run the store alone: there is no button to buy a place in that list. Ads do exist inside ChatGPT, but the help center states that they are separate from product results.

Second point, the one most articles miss: there is no results page, so there is no position. The help center says a product appears "when ChatGPT perceives it to be relevant to your intent", taking the query and the user's context (Memory, custom instructions) into account. Ask the same question twice, from two accounts, and you get two different answers. For a store, that changes everything: chasing a rank is a dead end, because the thing does not exist. "How to rank on ChatGPT" really means "how to be eligible every time the question comes up".

One scope note, to avoid a common mix-up: tracking whether your brand gets mentioned in answers is brand monitoring, a different job with different tooling. What follows is only about your products and your pages.

The three doors your product data can come through

Reading the help center and the developer documentation side by side, OpenAI describes three paths for product data.

1. Your public product page. This is the baseline, and the only one that needs nobody's permission. OpenAI's Overview of OpenAI crawlers documents three user agents, and names OAI-SearchBot as the one "used to surface websites in search results in ChatGPT's search features". The same page recommends allowing it in your robots.txt so that your site appears in search results. The whole mechanism is detailed below.

2. Structured metadata from third parties. OpenAI says it considers structured metadata from first-party and third-party providers, giving price and product description as examples, and that prices shown in ChatGPT are received from third-party providers. You do not steer that channel directly, but you inherit its mistakes: a stale price at an aggregator stays a stale price, and the help center itself warns that there may be some delay before a merchant's price or shipping update is reflected.

3. A direct product feed. OpenAI publishes a full product feed specification and a get started guide. It is the cleanest channel, and it comes with a line worth reading: the guide states that "onboarding product feeds in ChatGPT is currently available to approved partners", and that access is requested through a form. The help center says the same: merchants who want a direct feed can read the documentation and apply for access. In other words, a store with twelve SKUs does not sign up in one click. Saying so plainly saves you from chasing a closed door.

Practical conclusion: for the vast majority of stores, the open door is the first one. Which means the useful work happens on the product page itself, the same foundation we cover in our ecommerce SEO guide.

How ChatGPT Shopping picks products between two stores selling the same item

This is the most useful question, and OpenAI answers it fairly directly. Its help center says that when a user clicks on a product, ChatGPT may show a list of merchants offering it, and that "merchants are ranked based on factors like availability, price, quality, and whether they are the maker or primary seller of that item". It adds that this is expected to evolve, including toward rankings personalized to user preferences.

Read that list as a shop owner, not as an SEO:

  • Availability. A product that has been out of stock for six weeks and left online works against you.
  • Price. Not the cheapest, but an accurate and current price, consistent with what circulates elsewhere about your product. The help center explains that the price shown first is typically the first merchant's, not necessarily the lowest, and that ChatGPT works to highlight the lowest option it knows of, sometimes with a "Best price" label.
  • Quality. The perceived quality of the offer and of the page carrying it.
  • Maker or primary seller. This is the structural advantage of the maker and the own-brand store. If you make what you sell, you are on the right side of that factor, and it is one of the rare cases where a small store starts ahead of a reseller.

There is a nuance that OpenAI's own documents place side by side, and it would be dishonest to smooth it over. The Instant Checkout announcement says merchants pay a small fee on completed purchases, that it does not affect prices, that it does not influence ChatGPT's product results, and that Instant Checkout items are not preferred in product results. The same page also says that when ranking multiple merchants that sell the same product, ChatGPT considers "availability, price, quality, whether a merchant is the primary seller, and whether Instant Checkout is enabled". Both sentences come from the same publisher: there is no preference for the item, but checkout availability counts when arbitrating between merchants. For a small store the takeaway is simple: it is not a lever you can pull today, and it does not cancel the four factors above either.

Shopify and WooCommerce do not start from the same place

This is the least known asymmetry, and it is documented on both sides. OpenAI's help center writes: "For merchants on Shopify, product data is already integrated into ChatGPT through Shopify Catalog, helping products appear more accurately and completely in relevant user conversations. No additional work is required from individual merchants." Shopify's own robots.txt.liquid page confirms it from the other end: if you sell through agentic storefronts such as ChatGPT or Microsoft Copilot, your product data is syndicated to those channels through Shopify Catalog, independently of /robots.txt.

If you sell on Shopify, your catalog already travels through a structured channel without any action from you. That moves your work: it is about the accuracy of your product data, not about getting it delivered.

If you sell on WooCommerce, that channel does not exist for you. Your catalog is visible through the first door, the public page, read by OAI-SearchBot. The gap is not dramatic, but it is real, and it justifies being stricter about how accessible and how complete your product pages are: they are your only channel.

The three OpenAI robots and your robots.txt

OpenAI's crawler page lists three user agents, and they do not do the same job. Confusing them is how a store ends up invisible while believing it is protected, or the reverse.

User agent What OpenAI says it does What it means for your store
OAI-SearchBot "Used to surface websites in search results in ChatGPT's search features." Sites that opt out "will not be shown in ChatGPT search answers, though can still appear as navigational links." The only one that decides whether your product pages can be shown. Allow it.
GPTBot Crawls content that may be used in training OpenAI's models. Disallowing it "indicates a site's content should not be used in training". Your call. Blocking it does not remove you from search: OpenAI states that each setting is independent.
ChatGPT-User Visits a page when a user asks ChatGPT a question. "Not used for crawling the web in an automatic fashion", and "not used to determine whether content may appear in Search". Because these visits are user initiated, robots.txt rules may not apply. Nothing to configure. Managing search opt-outs goes through OAI-SearchBot, not this one.

The page adds two practical details: it can take about 24 hours after a robots.txt update for OpenAI's search systems to adjust, and OpenAI publishes the IP ranges of each agent (searchbot.json, gptbot.json, chatgpt-user.json) for site owners who filter at the network level.

How to check your own file in two minutes. Open your domain followed by /robots.txt in a browser. Then look for three things, in this order:

  1. A block starting with User-agent: OAI-SearchBot followed by Disallow: /. That is a direct block: your catalog is invisible to ChatGPT search through this door.
  2. A block starting with User-agent: * followed by Disallow: /, with no separate OAI-SearchBot block that allows it. The star applies to every crawler that has no rule of its own, so this blocks ChatGPT search along with everything else.
  3. A Disallow rule on the folder that holds your products (/products/ on Shopify, the product permalink on WooCommerce) inside any of those blocks.

If none of the three is there, you are not blocking it. If you want the two settings OpenAI describes as independent, allow search and refuse training, the file looks like this:

User-agent: OAI-SearchBot
Allow: /

User-agent: GPTBot
Disallow: /

On Shopify. Shopify's help center says all stores have a default robots.txt that starts with Allow: / for public content (products, collections, pages, blog, policies) and only disallows admin, cart, checkout, account and filtered or sorted collection addresses. So a Shopify store that never touched the file is not blocking OAI-SearchBot. Customizing it requires adding a robots.txt.liquid template to the theme, which Shopify labels an unsupported customization; the page notes that changes are instant but that crawlers do not always react immediately. Remember also that blocking AI crawlers at that level only affects open-web discoverability: your Shopify Catalog data still reaches the agentic storefronts you have activated.

On WooCommerce. The file is served by WordPress, and an SEO or security plugin, or a firewall in front of the site, can add rules you never wrote. Do the same three-line check on the live file rather than trusting the plugin's settings screen, and check it again after installing anything that promises to "protect against AI".

The nine fields you can use as a checklist

OpenAI's product feed spec is not open to every store, but it is public, and that is what makes it useful: it states what OpenAI treats as the minimum description of a sellable product. Read on 27 September 2026, the feed reference asks for one row per purchasable item or variant with nine required fields: item_id (a stable identifier), title, description (described as a "factual product description"), url (the product page, kept stable), brand, seller_name, image_url, availability and price. The spec asks for a real brand and seller name rather than placeholders, and recommends keeping titles to at most 150 characters and descriptions, in plain text, to at most 5,000. Its own minimal example is a blue ceramic mug of 350 mL, described as "dishwasher-safe glazed ceramic mug with a handle".

Use it as a checklist on your own pages, even if you never submit a feed. Open three of your product pages and confirm those nine pieces of information are there, explicitly, as text, and not only inside an image or inside the owner's head. On maker stores the same ones are always missing: brand, seller name, and a factual description instead of mood copy.

That overlaps with what we have been writing about product pages for a long time, which is reassuring: the work assistants ask for is not separate work. What reviews change on a page is covered in customer reviews on product pages.

A quotable page states facts, not atmosphere

A machine writing an answer needs sentences it can repeat without getting them wrong. Compare two ways of describing the same candle. "Lovingly handmade in small batches with respect for tradition" gives nothing to copy. "Soy wax candle, 8 oz, cedar and vetiver, 45 hour burn time, ships in 2 business days, returns accepted for 30 days" gives six usable facts. The second one serves the hesitating buyer first, and it is the only one an assistant can repeat without inventing something. How to write that copy for the buyer before the machine is covered in how to write product descriptions.

Shopify now says the same thing in its own words. Its page Optimizing your store for AI (read on 27 September 2026) recommends, for AI systems that crawl product pages, detailed specifications and technical details, comparison information with similar products, comprehensive descriptions with relevant keywords, structured data and product attributes, and sizing guides, material information and care instructions where applicable. It adds a useful placement rule: content written for a machine but not for your customers can go lower on the product page.

Two rules follow from it. Keep the facts in the text of the page: a price inside a banner or a size chart saved as a photo is invisible to a machine. And make the invisible layer say the same thing as the visible one. Your Shopify or WooCommerce theme already publishes product structured data for you; Google's own guidance on AI features in Search lists making sure structured data matches the visible text on the page, and keeping Merchant Center and Business Profile information up to date.

The same goes beyond your site. Your marketplace pages, your business listing, your social profiles and the old site you never took down should show the same name, the same prices and the same stock. Remember the second door: a stale price elsewhere stays a stale price. Getting those sources to agree is a short, occasional job, and one of the few in this field with a visible payoff.

Collection pages answer the choice questions

Go back to the mug for filter coffee. Questions put to an assistant are rarely a product name: they are questions of choice, which teapot for two people, which bedding for summer, a gift for someone taking up pottery. A collection page that is only a grid of thumbnails answers none of them. The same page with three short paragraphs at the top, explaining how to choose between the models and who each one suits, becomes an answer, and it keeps the choice inside your store. The method is in optimize collection pages.

The three brand pages that settle the doubt

Sending a buyer to a store they have never heard of is a bet, and three ordinary pages, usually written in a hurry, exist to settle it. They also answer the questions buyers ask before checkout, which makes them factual material any reader, human or machine, can repeat. Shopify's AI page points the same way with its free Knowledge Base app, which lets a merchant see and edit the FAQs that AI shopping agents use to answer questions about the store's policies.

  • Who you are. Facts, not intentions: what you sell, since when, where you are based, who makes it, how to reach you. "Ceramics studio in Portland since 2019, thrown on site, two people" does more than three paragraphs about passion.
  • How you ship. Zones, lead times, costs, carrier, and what happens when a parcel arrives damaged.
  • What you take back. Return window, expected condition, who pays for the return, refund delay. The rules depend on your country and your status, so point to your official source rather than copying a window you read somewhere.

What you can do this week

Four actions, in this order, none of which needs a developer.

  1. Check that you are not blocking OAI-SearchBot. The three-line check is above. If a rule disallows that crawler, your catalog is invisible through the only door open to you, and OpenAI notes that it can take about 24 hours after a robots.txt change for its search systems to adjust.
  2. Fix your availability. Go through the products that have been out of stock for a long time and decide: restock, unpublish, or state the lead time clearly. Availability is the first factor OpenAI names.
  3. Complete the nine fields on your best sellers. Start with five products, not the whole catalog. Brand, seller, factual description, price, availability, image, and a URL that does not change.
  4. Review your three brand pages. Who you are, how you ship, what you take back: fix whatever is vague or out of date.

Mistakes that cost you

  • Paying for a slot. Nobody sells a position in organic product results. An offer that promises that placement is selling something that does not exist.
  • Blocking every AI crawler on principle. That is a defensible call for a content site. For a store, it means walking out of an aisle where buyers are asking what to buy. If you want to refuse training, OpenAI documents a way to do it without leaving search: disallow GPTBot, allow OAI-SearchBot.
  • Polishing the homepage and leaving product pages bare. Product pages are what gets read and cited, not your about page.
  • Trusting a number with no source. This field is saturated with assistant market share statistics nobody can trace. If a number is not attached to a document you can open, do not build on it.
  • Waiting for a product feed. While access stays limited to approved partners, waiting means doing nothing for months.

How to rank on ChatGPT: what is in your hands, and what is not

Part of the outcome is not yours, and it is better to know it than to ignore it.

OpenAI says a product shows up when ChatGPT sees it as relevant to the intent, taking the query and user context into account, for example Memory or custom instructions. Its own example: if the buyer states a thirty dollar budget, ChatGPT will focus more on price; if price is not mentioned, it may focus on other aspects instead. It also says that not all available products will necessarily be shown. Two different buyers asking the same question will not see the same list.

So nobody can guarantee you an appearance, and no serious provider will. What you can get is eligibility: a page that is open to OAI-SearchBot, accurate and complete, with a fair price and real stock. The rest belongs to the model and to the buyer.

FAQ

Can you pay to appear in ChatGPT product results? No. OpenAI presents those results as organic, not ads, and not influenced by partnerships. Advertising in ChatGPT exists separately from product results.

Do you need a product feed to exist in ChatGPT Shopping? No. The direct feed is an extra channel, currently limited to approved partners who apply for access. Your public product pages remain the main channel for an independent store, and Shopify stores also travel through the Shopify Catalog.

Does blocking AI crawlers protect my store? Blocking every one of them also removes your products from answers. OpenAI documents its three user agents separately and states that each setting is independent: you can disallow GPTBot, used for training, and still allow OAI-SearchBot, used for search.

Do I need a special file or markup for AI? No. Google writes that you do not need to create new machine readable files, AI text files or markup to appear in its AI features, and that no special schema.org structured data is required. The product structured data your theme already publishes is enough, as long as it matches the page.

How long before my products appear? No delay is documented, and any promised timeline is made up. The only documented figure is on the other side: about 24 hours for a robots.txt change to be taken into account by OpenAI's search systems.

To wrap up

Ranking on ChatGPT does not ask for a new trade. It asks for accurate product pages, open to OAI-SearchBot, with a fair price, real stock and a factual description. That is exactly the work that already sells products in a classic search engine, which is good news when you run a store with one or two people.

Ecomrank connects your Shopify or WooCommerce store, imports your catalog, writes SEO articles around it, publishes them on your store blog with links to your product pages, and then measures honestly what they produced at seven, fourteen and twenty-eight days. The trial is free for 7 days and you can cancel anytime.

Start your free trial