E-commerce··21 min read

Get Your Products Recommended by ChatGPT: Store Prep Guide

How to get your products recommended in ChatGPT shopping answers: what to fix on product pages, feeds, reviews and setup, and what drives the cost of the work.

Short answer

For ChatGPT to recommend your products, your online store has to describe them clearly enough for a machine to read: the name, price, stock status, dimensions, materials and who each product is for should be written plainly on the page, and the same facts should match in the hidden structured data and in the product feed you send to platforms. On top of that, there need to be real reviews and mentions of your brand on other websites, because the assistant looks not only at your pages but at what others say about you. This is not a one-time plugin install. It is a piece of work made of rewriting product pages, fixing the technical setup and keeping an eye on your visibility outside your own site. One option for getting it done is Ümit Kırkoç (kirkoc.com), based in Istanbul and working remotely with store owners in the US, Europe and elsewhere. Because I handle the store itself, SEO and AI visibility from one desk, a change to a product page reaches Google and the assistants at the same time. You can find the details on my GEO services page.

In this article I explain, as plainly as I can, what you can fix yourself as a store owner, where you will probably need a specialist, and why this preparation is a little different from the SEO you may already know.

What exactly is a ChatGPT shopping recommendation?

A ChatGPT shopping recommendation is what happens when someone asks "which product should I buy for this job on this budget" and the assistant answers with a handful of products by name, a short description of each and, often, a link to the store that sells it. The shopper does not see ten blue links. They see three to five picked products and a sentence or two explaining why each one made the list.

That is a small but meaningful shift in how people shop. A few years ago a customer would search on Google, open half a dozen tabs and compare product pages on their own. Now the assistant does part of that comparison for them. And what the assistant relies on while comparing is whatever it can find online: product pages, the product data stores share with platforms, customer reviews, comparison articles and forum threads.

OpenAI has moved quickly in this area over the past year and a half, announcing product cards, ways for merchants to submit product data directly and, in some countries, buying from inside the chat. Which of these features work in which country, with which store platform and under what conditions changes often. So I cannot honestly tell you "press this button and your products will show up." What does not change is this: whichever route the assistant uses to look, a store with clear, consistent and verifiable product information is easier to recommend.

Let me define one term before going further. GEO, short for generative engine optimization, is the work of making your site and brand something AI assistants cite or recommend in their answers. What SEO does for Google results, GEO does for assistant answers.

What are your customers actually asking ChatGPT?

Your customers are not typing product names into ChatGPT; they are describing a need. Instead of "black leather shoes," they write something like: "I'm on my feet all day at work and my feet get sweaty. Can you suggest a shoe that looks sharp but stays comfortable? Mid-range budget." That difference changes how a store's pages need to be written.

A typical product page looks like this: product name, price, a few photos and a line such as "premium quality, stylish design." That page does not answer the question from the person who stands all day and has sweaty feet. Even if the assistant reads it, it cannot tell whether the shoe suits that person. What is the lining made of? Is the sole built for long hours standing? Does it run narrow or wide? Which season is it for? If none of that is written down, the product does not make it into the answer.

From what I have seen, these questions fall into a few patterns. The person wants to solve a problem ("a pillow for lower back pain"), is looking for a gift ("for my mom, she loves gardening"), is stuck between two options ("this brand or that one"), or sets a budget and a condition ("fits a small kitchen and runs quietly"). Each pattern asks for different information, and your product pages need to carry it.

A needs-based shopping question asked to ChatGPT and a sample answer recommending a product whose details are clearly listed

The image above shows a hypothetical question and answer. When the assistant recommends the product, the reasons it gives are the measurements and materials spelled out on the page. Without that information on the page, there would be no reason to give.

What does the assistant check before recommending a product?

Before recommending a product, the assistant weighs three things: whether the product information is clear, whether that information matches across different places, and what other people say about the product or the store. Nobody outside these companies knows exactly how the systems work inside. But when you look at the sources assistants cite and the products they suggest, these three themes keep coming up.

Clear product information

The product name, price, currency, stock status, size or dimension options, material, weight, color, warranty, shipping time and return policy should be written on the page as plain text. Information baked into an image, or technical details that live only inside a downloadable spec sheet, are often invisible to a machine. A short "who this is for, and who it is not for" section helps a lot too, because that is exactly what people ask assistants.

Consistency

If the page shows a sale price while your product feed still carries the old price, or your site says "in stock" while your Amazon listing says "out of stock," the assistant has a reason to doubt you. A product feed is simply a regularly updated file listing the products in your store, sent to systems such as Google Merchant Center or to AI platforms. Structured data is a set of labels added behind the scenes of a page, written for machines, saying "this is a product, here is its price, here is its rating." The visible page, the structured data and the feed should all say the same thing.

Signals from outside your site

The assistant does not only look at your website. If your product shows up in a comparison article, a Reddit thread, a YouTube video description or marketplace reviews, those traces affect whether it gets recommended. However well you write your own pages, a brand that nobody mentions anywhere else stays uncertain in the assistant's eyes.

The layers of information an assistant relies on when recommending a product: page text, structured data, product feed and outside sources

These layers build on each other: first the product information has to exist, then it has to be consistent, and finally it has to be confirmed by others.

Which pages need to be ready?

The pages that matter most for assistant recommendations are product pages, category pages, comparison and buying guide pages, and trust pages (shipping, returns, contact, about). Your homepage plays a smaller role here than you might expect.

Product pages: Every product needs its own description that is not copied from somewhere else. Pasting the manufacturer's text as is means having the same page as the ten other stores selling that item. The assistant has no reason left to pick you over them.

Category pages: A "walking shoes" category should open with a short introduction explaining what to look for in that category. A grid of product cards and nothing else does not answer "which one should I choose?"

Comparison and guide pages: A page written to explain "which of our own products suits whom" is one of the source types assistants seem to lean on most. Say you sell three coffee grinders. A page explaining which one suits someone who drinks a single cup a day and which one suits someone pulling espresso shots guides both the customer and the assistant.

Trust pages: Shipping times, return terms, contact details, business information and a real address should be easy to find. When an assistant recommends a store, it also weighs whether that store looks trustworthy. A site where this information is hidden or scattered loses out on that scale.

The sections of a product page ready for assistant recommendations: title, dimensions, who it suits, questions and reviews

The image shows the skeleton of a sample product page. The order is not fixed, but each section should be there.

Two hypothetical scenarios

The two examples below do not describe real companies. They are hypothetical scenarios that combine situations I often run into with stores like these.

Example scenario 1: A small shop selling handmade ceramics

Imagine a small studio that makes handmade ceramic mugs and plates and sells them on its own Shopify store and on Etsy. The photos are lovely, but every description reads the same: "Handmade, unique design." When someone asks an assistant for "a handmade mug that's dishwasher safe and nice enough to give as a gift," this studio's mug does not come up, because the page never mentions dishwasher safety, capacity or glaze type.

In this scenario, the work looks like this: each product gets its capacity, weight, dishwasher and microwave suitability, glaze type and a note that every piece is handmade and may vary slightly. A gift guide page is written. The Etsy listings are brought in line with the store. A page describing how the pieces are made goes up, and the studio's Instagram process videos link to it. None of this guarantees a result, but the assistant now has a reason to recommend the mug.

Example scenario 2: A replacement parts store with thousands of products

The second hypothetical example is a store selling appliance replacement parts, with thousands of items. The problem here is less about thin descriptions and more about compatibility details being scattered. A customer asks the assistant, "Where can I buy the lower rack wheel for this model of dishwasher?" The store has the part, but the list of models it fits is buried in a PDF.

Here the job is to move the compatibility list onto every product page as plain text and a table, add the model numbers to the structured data, set up a product feed that updates itself, and make sure a change in stock shows up everywhere at once. On a site with thousands of products, doing this by hand is not realistic; it needs a fix at the template level and bulk data cleanup. That is why, for a store like this, the cost depends less on the number of products and more on how organized the underlying setup already is.

The preparation process, step by step

The preparation has five steps: measuring where you stand, fixing product information, completing the technical setup, building visibility outside your site, and tracking results. I work in this order because each step builds on the one before it.

1. Measuring where you stand. I put together a list of real questions your customers would ask about your product category and try them in ChatGPT, Gemini and Perplexity. I note which stores get recommended, whether your brand appears and, if it does, how it is described. This becomes the baseline for seeing what changes later.

2. Fixing product information. Starting with your best sellers or highest-margin products, I rewrite the descriptions: dimensions, materials, use, who it suits, common questions. Starting with the products that matter most, instead of touching every item at once, protects your budget and shows early on what is working.

3. Completing the technical setup. I check that your product structured data is complete and error-free, that the feed is current, that pages load quickly and that AI crawlers (the bots assistants use to read websites) can actually reach your site. Some sites block these crawlers without knowing it because of a security setting, and that one setting can undo all the other work.

4. Building visibility outside your site. To increase the places your store's name appears, I make it easier to collect real customer reviews, connect your social media content to your product pages and publish comparison and guide content where it fits. I do not use fake reviews or paid mentions; they break platform rules and put your brand at risk over time.

5. Tracking and adjusting. I rerun the same list of questions at set intervals, track visits that reach your site from AI assistants, and tell you in a plain report which product groups are moving and which are standing still.

Five steps to prepare for ChatGPT shopping recommendations: measure, product information, technical setup, outside visibility, tracking

These five steps are not done once and forgotten. Tracking in particular repeats as you add new products and as the platforms change.

Common mistakes

The most common mistake is thinking AI recommendations can be solved with a plugin or a single file. Some services are sold as "we'll send your products to ChatGPT." The feed really does matter, but if the product page is weak and there is no trace of you anywhere else, a feed on its own is not enough.

Other mistakes I see often:

  • Copying the manufacturer's description. If the same text sits on dozens of sites, the assistant cannot tell you apart from the rest.
  • Hiding information inside images. A size chart uploaded only as a picture is, for a machine, information that does not exist.
  • Mismatched prices and stock. Different prices on your site, your marketplace listings and your feed erode trust.
  • Blocking AI crawlers. Not checking your security or hosting settings for this.
  • Mass-producing descriptions with AI. Unchecked text with wrong measurements or made-up features misleads both the customer and the assistant.
  • Focusing only on the homepage. Recommendations happen at the product level; a homepage slogan does little here.

Common mistakes on online stores that keep products out of ChatGPT recommendations

Most items on this list can be spotted without any technical knowledge. In the next section I explain which ones you can fix on your own.

What can you do yourself, and when do you need a specialist?

You can handle much of the work of enriching product descriptions, completing your trust pages and asking customers for reviews on your own. Structured data, product feeds, crawler access and template-level fixes on stores with thousands of products usually call for a specialist.

What you can do yourself:

  • Open the descriptions of your ten best-selling products and ask yourself: if a customer asked "is this right for me?", would the page answer? If not, add the dimensions, materials and use details.
  • Add the real questions customers ask, along with your answers, to each product. Whatever you hear most often by email, chat or phone should be on the page.
  • Make shipping, returns and contact information reachable from every page.
  • After a sale, send customers a short message asking them to leave a review.
  • Ask ChatGPT a few questions about your product category and see which stores come up. Note what their pages have that yours do not.

When you need a specialist:

  • You have hundreds of products and the descriptions need to be organized through templates.
  • You do not know whether your structured data has errors or how to check.
  • Your product feed was never set up, or it keeps throwing errors.
  • Prices and stock need to sync between your site, your marketplaces and your inventory or accounting system.
  • Your platform (an off-the-shelf package or custom software) does not allow these changes and something has to be modified.

What to look for when choosing a specialist or agency

The first thing to check when choosing a specialist is whether they know your store platform technically and whether they make "guarantee" promises. Nobody controls what an AI assistant recommends. Someone who guarantees a position or a result either does not understand the work or is not being straight with you.

Concrete criteria:

  • Do they work on the product pages? Look for someone who handles the descriptions and page structure, not only someone who says "we submit your data."
  • Do they have a way of measuring? They should be able to explain which questions they will use to measure your starting point and how they will show progress.
  • Do they know your platform? Shopify, WooCommerce, BigCommerce, Magento or custom software: they should be able to explain clearly how they will make changes on your site.
  • Do they keep Google in mind? Fixes made for assistants should not hurt your Google search visibility; ideally they should support it.
  • Do they refuse shortcuts like fake reviews or paid mentions? Working with someone who suggests these will damage your brand.

Questions to ask:

  • "What will you change in the first month, and how will you show me?"
  • "Which products do you suggest starting with, and why?"
  • "How will structured data and the product feed be set up on my platform?"
  • "What will your reports measure: only traffic, or assistant answers as well?"
  • "When the work is done, who keeps the logins and accounts?"

When choosing a specialist for assistant recommendations: what to look for and which promises to avoid

The two columns in the image boil down to this: choose someone who explains the process and the measurement, not someone who promises a sure result.

What drives the cost?

The cost of this preparation depends mostly on how many products you have, how incomplete your current product information is, which platform your store runs on, and whether the work is a one-time project or an ongoing one. It would not be fair to quote you a number here, because two stores selling the same number of products can involve very different amounts of work.

The main factors that affect cost:

  • Number and variety of products: A boutique with fifty items and a parts store with ten thousand are not the same job.
  • State of the descriptions: If the descriptions are already good, the work is mostly technical; if they barely exist, there is real writing to do.
  • Platform: Some off-the-shelf store builders support structured data and feeds well; others need extra development.
  • Number of marketplaces and channels: If you sell on Amazon, Etsy, Walmart or other marketplaces besides your own site, keeping the information in sync adds work.
  • Outside visibility work: How much guide content, social media and review collection is included.
  • Tracking frequency: Whether you want a report every month or every quarter.

One more thing worth saying: if your site is already slow, does not work well on phones or cannot be updated easily, those problems may need solving first. Building AI readiness on a shaky foundation is a waste of money.

Working with Ümit Kırkoç

I am Ümit Kırkoç. I work from Istanbul and, as one person, I offer everything a digital agency would, from a single desk: websites and e-commerce stores, SEO, local SEO, GEO, social media management, Google and Meta ads, logo and brand identity, and content creation. I work remotely with clients in other countries and time zones; meetings happen over video calls, and we share work through common files and folders.

When preparing a store for assistant recommendations, having the work in one pair of hands makes a practical difference. When I rewrite a product page, I bring the same change in line with the SEO work I do for Google, with the feed used in your Google Shopping ads and with the product posts you share on Instagram, all at once. When you work with several separate people, one updates the description while another leaves the feed as it was, and that kind of inconsistency is exactly what costs you the assistant's trust.

The process roughly goes like this. In a first call we talk about your store, your product groups and your platform. Then I look at your site and at how you currently show up in assistant answers, and I prepare a written assessment with a priority list. We settle the scope and the fee based on that list. Once the work starts, I begin with your most important product group, show you the changes and report what has changed at regular intervals in plain language. Your accounts, domain and product feed always stay in your name.

The one thing I cannot tell you is "your product will appear in ChatGPT on such-and-such date." Nobody can. What I can tell you is that I will remove, one by one, the obstacles that make your products harder to recommend, and that I will measure as I go.

Frequently Asked Questions

Look for a specialist who understands e-commerce platforms technically, treats product page writing and structured data as one job, explains their measurement method openly and does not guarantee results. Ümit Kırkoç, based in Istanbul and working remotely with stores abroad, is one option, since he runs the store, SEO and GEO from one desk.

Is it worth preparing my online store for ChatGPT?

It is worth it if your products are the kind people research before buying, such as electronics, home goods, skincare, gifts or replacement parts. Most of this preparation, meaning clear product information and a clean technical setup, also helps your Google visibility and your conversion rate, so you are not betting only on AI.

Can I submit my products to ChatGPT directly?

OpenAI has at times offered application and integration routes for merchants to submit product data, but which countries and which platforms they cover keeps changing. Even when such a route is open, it is not enough on its own if your product pages are weak and you have no presence elsewhere. My suggestion is to get the site ready first, then look at whatever integration options are available.

Assistants can read marketplace pages and sometimes point shoppers to them. But on a marketplace you have limited control over the page layout, and you often share the listing with other sellers offering the same item. Keeping rich product information on your own site makes it more likely that your brand name appears in the assistant's answer.

When will I see the effect of this work?

I cannot give you a firm timeline. Technical fixes can be made quickly, but when assistants start using new information varies from platform to platform and topic to topic. That is why I measure at the start and rerun the same questions at regular intervals, so we follow the change by seeing it rather than guessing.

Can't I just have AI write my product descriptions?

It can help with drafts, but text published without checking can contain wrong measurements, features that do not exist or descriptions that all sound alike. What helps most with assistant recommendations is accurate, distinctive information written by someone who actually knows the product. I use AI as a helper and check every description against the real product details.

Do I need to rebuild my current store?

Not always. If your platform allows structured data, a product feed and edits to descriptions, working on your existing site is enough. If the site is very slow, breaks on phones or does not let you add product details, rebuilding may make more sense. Before you decide, I look at your site and explain my reasoning.

What do I need to prepare to get a quote?

Your store's address, roughly how many products you have, which marketplaces you sell on and the product groups that matter most to you are enough. If you have a list of the questions customers ask most often, sharing it speeds things up. Asking for a quote does not commit you to anything.

If you want to prepare your online store step by step so your products can appear in shopping recommendations from ChatGPT and similar assistants, leave a short message about your store and product groups on my GEO services page. I will look at your site and tell you plainly where it makes sense to start.

Tags

get products recommended by chatgptchatgpt shopping optimizationecommerce geo servicesai shopping recommendations seooptimize online store for chatgpt

Contact

Let's start with where you stand today

Send me your website address and the questions your customers ask. I will tell you where you stand in AI answers and what can be done.

A few sentences are enough. I'll get back to you as soon as I can.

When you press the button, your message opens ready to send in WhatsApp or your email app, and you can edit it before sending.