Make sure ChatGPT can read accurate data about your products: on Shopify, keep Agentic Storefronts switched on; on Etsy, your listings reach ChatGPT through Etsy's own app; on other platforms, allow OpenAI's OAI-SearchBot crawler and apply to submit a product feed. Then earn mentions in reviews and comparisons. You cannot pay for placement in the product results.
OpenAI says product results are chosen independently, not sold. Selection draws on structured data such as price and description, and when several shops sell the same item, the ranking considers availability, price, quality and whether the merchant is the maker or primary seller. Below you will find each route in, what to fix, how to test it and how to track results. The caveat: this area changed twice in 2026, so check the current settings rather than relying on any guide, this one included.
What changed in 2026
- Instant Checkout came and went. Buying inside ChatGPT launched in September 2025. In March 2026 OpenAI said the first version "did not offer the level of flexibility" it wanted, let merchants use their own checkouts, and shifted its focus to product discovery.
- Shopify switched discovery on by default. Shopify's Agentic Storefronts, announced in March 2026, make products from eligible stores discoverable in ChatGPT and other AI channels without an app. Shoppers check out on your store; Shopify says there are no transaction fees beyond normal payment processing.
- Etsy launched an app inside ChatGPT in May 2026, in beta. Shoppers tag @Etsy in a prompt, browse listings and click through to Etsy to buy.
- Ads arrived, separately. OpenAI began testing ads on the Free and Go tiers in early 2026, shown below answers and labelled Sponsored. OpenAI says they do not influence answers.
The three routes your product data takes into ChatGPT
| You sell on | How ChatGPT gets your data | What you switch on | What you control |
|---|---|---|---|
| Shopify | Shopify's product catalogue, synced to AI channels | Settings, then Sales channels, then Agentic Storefronts; eligibility depends on the markets you sell to | Titles, descriptions, product category, variants, price, stock, images, shipping |
| Etsy | Etsy's app in ChatGPT | Nothing per shop | Listing titles, tags, attributes, photos, reviews |
| Your own site on another platform | OpenAI's search crawler, plus a direct product feed if accepted | Allow OAI-SearchBot in robots.txt; apply through OpenAI's merchant page for a feed | Product pages, structured data, the feed itself |
OpenAI's crawler documentation is explicit: sites that block OAI-SearchBot will not be shown in ChatGPT search answers, and robots.txt changes take about 24 hours to register. OAI-SearchBot is separate from GPTBot, the crawler used for model training, so you can block training and still allow search.
For a direct feed, OpenAI's specification requires fields including a title (up to 150 characters), a plain-text description (up to 5,000 characters), a product URL, brand, seller name, an image, availability and price, with optional review count and star rating. Even if you never submit a feed, that list is a good checklist for what your product pages should state clearly.
Step 1: find out whether ChatGPT can see you at all
Before changing anything, run five prompts a real customer might type, and log what comes back. Use a fresh chat with memory off so earlier conversations do not bias results. A filled-in test log, from an illustrative car repair garage that sells its own car-care range:
| Prompt | Our product shown? | Who was shown | What ChatGPT said about us |
|---|---|---|---|
| pH-neutral car shampoo safe on ceramic coating | No | Three national brands | Nothing |
| car wash kit gift under $40 | Yes, 4th of 6 | Two marketplaces, two brands | Correct price; said "includes wax", which it doesn't |
| interior cleaner for fabric seats and pet hair | No | Four brands | Nothing |
| [our brand name] car shampoo review | Yes | Us only | Summarised two reviews accurately |
| best car care kit for a new driver | No | Retailer bundles | Nothing |
This log tells the garage two things: ChatGPT can reach its products (they appear for a gift query and a brand query), and the data is wrong in one place ("includes wax"). Visibility is a relevance and data problem, not an access problem.
A realistic access problem looks different. Another shop found its products never appeared, even for its own brand name. Its robots.txt had been edited by a website plugin set to "block AI bots", which blocked OAI-SearchBot along with the training crawlers. Removing that one rule was the most valuable fix it made all year. Check yours by visiting yourdomain.com/robots.txt and searching the page for "OAI-SearchBot".
If you want to stay out of model training but stay visible in ChatGPT search, the relevant part of the file needs two separate groups, one per crawler:
User-agent: OAI-SearchBot
Allow: /
User-agent: GPTBot
Disallow: /What the plugin had written instead was a single catch-all group listing several AI crawlers, OAI-SearchBot among them, under one "Disallow: /". The two versions look similar at a glance, which is why the mistake survives for months. After you change the file, allow about a day for OpenAI to pick it up, then rerun the brand-name prompt from your test log. If your products appear for your own brand name, access is fixed and everything after that is about relevance.
Step 2: fix the product data ChatGPT reads
AI shopping matches products to detailed questions ("safe on ceramic coating", "for pet hair"). The facts that answer those questions have to be in the product data, stated plainly. A before and after for the garage's shampoo (illustrative):
BEFORE
Title: Premium Shampoo 500ml
Description: Our best-ever formula for a showroom shine. Try it!
Category: (blank)
AFTER
Title: pH-Neutral Car Shampoo 500 ml, Safe on Wax and Ceramic Coatings
Description: A pH-neutral car shampoo for hand washing. Safe on waxed,
sealed and ceramic-coated paint; not suitable for matte wraps.
One 500 ml bottle makes about 20 buckets at 25 ml per bucket.
Made by us in our workshop. Rinse off in the shade.
Category: Car wash solutions (from the platform's category list)
Price: $14 Availability: in stock Brand: [garage brand]The "after" version answers the question from the test log that the garage lost. It also says what the product is not for, which helps an assistant avoid recommending it to the wrong person. (Always pick the category from your platform's own list rather than typing your own.)
Work through your range with this checklist:
- Title: product type first, then the attribute people search on. No slogans.
- Description: plain facts, compatibility, what it is not suitable for, size and quantity.
- Category and variants set properly, so a 250 ml and a 1 litre bottle are not one vague product. The mistake shows up in answers: the garage's interior cleaner was listed as one product "from $8", and a test prompt came back with "[garage brand] Interior Cleaner, 1 litre, $8". The 1 litre is $19. Set each size as a variant with its own price, stock and option value ("250 ml", "1 litre"), and the assistant has no reason to mix them up.
- Price and stock accurate. An assistant recommending an out-of-stock item is the fastest way to lose that shopper's trust.
- Barcode (GTIN) and brand where you have them, so your listing is matched to the right product.
- Shipping and returns stated on the product page, not only in a policy page.
Sometimes the wrong fact doesn't come from your own data at all. An illustrative answer from a later test, for "interior cleaner for fabric seats and pet hair":
"[garage brand] Interior Cleaner ($12, 250 ml) is a good budget option. It lifts pet hair from fabric and is also safe on leather, according to customer reviews."
The product label says not for leather. The claim came from one review ("used it on my leather seats, came up fine"), which the assistant treated as a product fact. You can't edit a customer's review, and shouldn't try. What you can do is make your own data unambiguous: add "Not for leather or suede; use our leather balm instead" to the description and to the "not suitable for" line, and reply publicly to that review with the same advice. Your own product page should always be the clearest statement of what the product is and isn't for.
Etsy sellers work the same list through Etsy's own fields rather than a feed. A small candle maker, for instance, might find its listing had a poetic title ("Autumn Evenings") and empty attributes. The rewrite puts the product type and scent first ("Fig and cedar soy candle, 200 g, 40-hour burn"), fills in the attributes Etsy offers for the category (scent, material, and so on), and uses the tags for the phrases shoppers actually type ("fig candle", "woody candle gift"). Since Etsy's app in ChatGPT draws on Etsy's listing data, those fields are the only lever an Etsy seller has. Check each rewritten listing in Etsy itself afterwards, since a title that reads well in a spreadsheet can be cut short in search results.
AI assistants can draft these rewrites from your existing data; the method in writing Shopify product descriptions with AI works for any platform.
Step 3: be the maker or the primary seller where you can
OpenAI names "whether the merchant is the maker or primary seller" among the factors when several shops sell the same product. For small businesses this cuts both ways.
If you make your own products, as the garage does with its car-care range, say so in the product data ("made by us in our workshop"), use your own brand consistently, and make your own store the most complete source of information about the product. Resellers of your products should not have better descriptions than you do.
An independent bookshop that sells other people's books
A bookshop selling new books is rarely the primary seller of any title; publishers and large retailers are. Competing on "novel about grief" queries is unlikely to work. What a bookshop can own are products nobody else has: signed copies, its own curated bundles ("five books for a first-time gardener"), bookplates and its own gift boxes. Those get full product data, a clear description of what makes them unique, and reviews. The shop's regular stock still benefits from accurate data, but its visibility strategy rests on the things only it sells.
Step 4: earn the outside signals
ChatGPT's answers draw on what the wider web says, not only your own pages. The signals a small shop can realistically build:
- Reviews on your own site, collected after every order. The feed specification has fields for review count and star rating, which shows they matter.
- Mentions in genuine comparisons: send products to relevant independent reviewers, enter trade awards, answer questions in the forums where your customers ask for recommendations. Always disclose that you are the seller.
- Consistent facts everywhere: the same product name, size and price on your site, marketplaces and social profiles.
The reviewer approach is where small shops are most tempted to cut corners, so here is honest wording the garage might use: "We're a car repair garage that makes its own car-care range. Would you like to try our pH-neutral shampoo? We'll send a 500 ml bottle; there's no need to review it, and if you do, we'd rather you said what you really think, including anything you don't like. We're the maker, so please say so if you mention us." It asks for nothing in return, makes the relationship clear, and produces the kind of independent mention an assistant can reasonably trust.
Never buy or fake reviews. Apart from the legal risk, AI summaries surface patterns, and a burst of identical five-star reviews reads as exactly what it is. The broader approach is in getting your business recommended by ChatGPT and AI search.
Step 5: track whether it is sending you shoppers
- Shopify shows orders from AI channels with their referral source in the admin.
- Web analytics: visits from links clicked in ChatGPT commonly arrive tagged
utm_source=chatgpt.com, so filter your analytics by that source. - Your test log: rerun the same five prompts monthly and note changes.
- Ask customers at checkout or in a post-purchase question how they found you. "ChatGPT told me" answers are appearing on small stores.
Read the sources together, because each one undercounts. An illustrative month for the garage: analytics shows 41 sessions tagged chatgpt.com, and 2 orders attributed to them. The post-purchase question ("How did you hear about us?") got 64 answers, 4 of which said ChatGPT or "an AI assistant". The gap is normal: some people read a recommendation, then search for the shop by name or type the address, and arrive untagged. So the realistic range for that month is 2 to 4 orders from ChatGPT, and the trend across three or four months tells you more than any single figure.
The tutorial on tracking whether AI search is sending you customers sets this up in more detail.
The garage's range over three months
Following the same illustrative garage through a quarter: it sells 18 own-brand car-care products on a Shopify store, and Agentic Storefronts is on by default. Its first test log (above) showed it appearing for one of five non-brand prompts.
- Month 1, about six hours: rewrote all 18 titles and descriptions from a fact sheet per product (compatibility, what it is not for, quantities), set categories and variants properly, and fixed the "includes wax" error at its source, an old bundle description.
- Month 2, about two hours: added a review request to every order confirmation and put shipping and returns on each product page. Sent the shampoo and interior cleaner to two independent car-care reviewers, with no conditions attached.
- Month 3: rerun of the test log showed the shop for two of the five non-brand prompts, including the ceramic-coating one. Analytics showed a handful of sessions a week tagged from ChatGPT, and three orders the owner could trace to it.
Small numbers, and no guarantee they would repeat, but the work also improved the store for every other visitor, which is the safest reason to do it.
What does not work
- Stuffing titles with keywords. Assistants read for meaning; a 140-character title of search terms reads as spam to people and machines.
- Anyone promising guaranteed ChatGPT placement. OpenAI says the product results are not for sale.
- Blocking OpenAI's crawlers wholesale and expecting to appear.
- Claiming to be the maker when you are not. It is misleading, and it is easily contradicted by the real maker's data.
ChatGPT Shopping: what merchants ask
Can I pay to have my products recommended in ChatGPT?
Not in the product results. OpenAI says shopping results are chosen independently and are not ads. Separately, OpenAI has been testing ads in ChatGPT on its Free and Go tiers since early 2026, shown below answers and labelled Sponsored. Anyone selling guaranteed placement in the organic product recommendations is selling something OpenAI says does not exist.
Do I still need a product feed if I'm on Shopify?
Usually not. Shopify's Agentic Storefronts pass your catalogue to ChatGPT and other AI channels by default for eligible stores, with no separate app. What you control is the quality of the product data Shopify sends, which comes from your titles, descriptions, categories, prices, stock and images. Check the Agentic Storefronts page in your settings to confirm your store is included.
How long does it take to show up after I fix things?
It varies and nobody can promise a timescale. OpenAI says robots.txt changes can take about 24 hours to be picked up. Product data changes on Shopify sync in near real time to the AI channels, but whether you then appear for a given question depends on relevance and competition. Test the same prompts monthly rather than daily.
Further reads
- What to Do When ChatGPT Gets Facts About Your Business Wrong — What to do when ChatGPT describes your products incorrectly.
- Can a Small Online Shop Use AI to Compete With Bigger Brands? — Where small shops can realistically beat big retailers.
- How to Turn Customer Reviews Into Marketing Copy With AI — Collect and use reviews, one of the signals AI shopping reads.
- How to Use AI for Etsy Titles, Tags and Descriptions — Etsy listing data that also feeds Etsy's app in ChatGPT.
- How to Choose an SEO Agency That Understands AI Search — Questions to ask before paying anyone for AI visibility.
- Taking a Local Shop Online With AI: Product Listings in a Weekend — Getting a range online with clean product data from day one.
- How Small Amazon Sellers Use AI for Listings and Reviews — How a small Amazon seller can use AI for listings under the 2026 title rules, analyse reviews for fixes, and grow reviews without breaking policy.
- Is AI Frame Advice and Virtual Try-On Worth It for Opticians? — What virtual try-on and AI frame advice cost an optician, the frame-digitising cost most people miss, three practices' verdicts and a 30-day test.
- What Is Generative Engine Optimisation and Does Your Business Need It? — Where the GEO idea came from, how AI answer engines choose sources, a decision table by business type, and five checks that cover most of it.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: OpenAI developer documentation on crawlers (OAI-SearchBot) and the Agentic Commerce product feed specification; OpenAI statements on product discovery (March 2026) as reported by Digital Commerce 360 and CNBC; Shopify newsroom and Shopify explainer on Agentic Storefronts (March and June 2026); TechCrunch on Etsy's ChatGPT app (May 2026).