How to get your products recommended by ChatGPT

ChatGPT now answers shopping questions with named products, prices and links to the stores that sell them. Ask it for “a good burr grinder under $100” and it may show a carousel of grinders and, when you open one, the stores that carry it. Your store is either in that answer or it isn’t, and there is no second page to scroll to.

There is no trick that puts you there. But ChatGPT can only recommend what it can read, trust and match to the question, and each of those is something a store can work on. This guide covers what OpenAI has said about how it works, and seven things to do about it.

How ChatGPT decides which products and stores to show

ChatGPT answers from two places. One is what the model learned in training, which is fixed until the next model comes out. The other is search: when a question needs current information, like prices or what’s in stock, ChatGPT searches the web and reads pages as it answers.

For shopping, search is what matters. OpenAI runs a separate crawler for it, OAI-SearchBot, and its documentation is blunt about what blocking it does: sites that opt out “will not be shown in ChatGPT search answers.”

When ChatGPT shows a product, it can list the stores that sell it. OpenAI’s shopping help page says that list comes from product and merchant data, from third-party providers or from the merchants themselves, and that stores are ranked on things like availability, price, quality and whether the store is the maker or main seller of the item. It also says product results are chosen by ChatGPT and aren’t ads.

So there are three levers: can ChatGPT read you, does it have accurate data about your products, and do other sources give it reason to trust you.

1. Let OpenAI’s crawlers in

Open yourstore.com/robots.txt and look for these names. OpenAI documents three that matter here:

A store that wants to be recommended should allow OAI-SearchBot at the very least. Then check your host or CDN. Some offer a one-click setting that blocks AI crawlers, and it can block search crawlers along with training ones. If your robots.txt looks fine but you never appear, that setting is the first place to look.

A last technical check: open a product page with JavaScript turned off. If the price and the description vanish, crawlers that don’t run JavaScript see an empty page too.

2. Put each product’s facts on the page

AI answers lean on facts they can quote: price, size, material, fit, weight, compatibility, what’s in the box, who it’s for. Marketing copy is hard to quote. “Crafted for the modern explorer” tells a model nothing. “Waterproof to 10,000 mm, 1.2 lb, packs into its own pocket” answers three questions.

Two places to put those facts:

• On the page, in words. A short spec table and one or two plain sentences near the top. Write the sentence you’d want quoted.
• In structured data. Product markup with name, brand, price, currency, availability, GTIN or SKU, and reviews if you have them. Most store platforms add the basics, so check yours is complete and matches the page.

Then compare yourself with the stores ChatGPT does name. Open their product pages for the same item. What do they state that you don’t? That gap is usually the fastest fix.

3. Get your catalog to ChatGPT directly

You don’t have to wait to be crawled.

• Shopify stores. Since March 24, 2026, products from eligible Shopify stores can be found and bought through ChatGPT via Shopify’s catalog, with checkout on the store’s own site (Digital Commerce 360, March 25, 2026). Check your Shopify admin for whether your store is included and how your products appear.
• Everyone else. OpenAI publishes a product feed specification for sending a catalog to ChatGPT: titles, descriptions, images, prices and availability, refreshed at least daily. Onboarding is open to approved partners, and you apply at chatgpt.com/merchants.

Either way, the data has to be right. A feed that says in stock when the page says sold out is the kind of mismatch that costs trust.

4. Be talked about where ChatGPT reads

When ChatGPT searches, it doesn’t only read stores. It reads reviews, forums, magazine roundups and comparison articles, and it tends to recommend what several independent sources agree on.

This is the slow part, and it can’t be faked:

• Reviews. On your own pages and on the review sites people in your category use.
• Buying guides. Get your products in front of the people who write “best X for Y” roundups in your niche. Send samples, answer questions, correct mistakes.
• Forums and communities. Be useful where your buyers ask questions. Don’t post ads.
• Find the sources that cite your competitors. The quickest list of sites to work on is the sites AI answers already cite for your searches, for a rival and not for you.

5. Answer the questions shoppers ask

Shoppers ask ChatGPT questions they’d never type into a search box: “I have wide feet and run on trails, which shoe?” The stores that get named often have a page that answers a version of that question.

Write those pages. A sizing guide that says who each size suits. A comparison of your own products. A “which one should I buy” page for your category. An FAQ with real answers. Each should start with the answer in one or two sentences, then explain.

6. Keep prices and stock accurate everywhere

ChatGPT ranks stores partly on price and availability. If your page, your structured data, your Google Merchant Center feed and your ChatGPT feed disagree, at least one of them is wrong, and an answer may quote the wrong one. Make one system the source of truth and keep the rest in step with it.

7. Check what ChatGPT says about you and your competitors

You can’t improve what you don’t look at. Ask ChatGPT the questions your buyers ask, in the country you sell in, and note which stores it names and which sources it cites. Do it again next week. The stores it names instead of you, and the pages it quotes from them, are your to-do list.

Doing that by hand for dozens of questions gets old fast. Google’s AI Overviews answer the same kind of question at the top of Google, and Over the Fence checks those every week: how often they cite your store against each competitor, with the questions and the sources behind each count. We don’t track ChatGPT itself yet. For that, ask it yourself or use a prompt tracker such as OtterlyAI or Peec AI, both in our comparison of monitoring tools.

What doesn’t work

• Writing for the model instead of the shopper. Stuffing pages with “best” and question phrases reads badly to people and gives a model no new facts.
• Fake reviews and planted forum posts. They break platform rules, and they get spotted.
• Hidden text and cloaking. Showing crawlers one page and people another is the oldest trick in search, and search engines treat it as spam.
• Treating llms.txt as a ranking switch. There’s no public evidence OpenAI uses it to choose recommendations. Publish one because it’s cheap and helps some agents, then spend the time on your product pages.

A checklist for this week