Let me describe something that is already happening, quietly, millions of times a day. A person opens ChatGPT and, instead of searching Google, simply types: "I need waterproof hiking boots under 150 dollars for wide feet." The assistant does not return ten blue links. It returns a considered shortlist of actual products, with prices and reasons, and in a growing number of cases, it offers to complete the purchase without the person ever leaving the conversation. This is what the industry has come to call agentic commerce, and for a small merchant it represents both a remarkable opportunity and, if ignored, a quiet threat.
The threat is this: if the AI cannot read your catalogue clearly, your products simply do not appear, not because they are inferior, but because the machine cannot be confident enough to recommend them. And the opportunity is the mirror image. Because most product catalogues are, to be honest, a bit of a mess: incomplete descriptions, stale stock counts, prices that do not match the live site. The merchant who puts their data in order early tends to earn the recommendation before the category becomes crowded. The truth is that in agentic commerce, clean data is not a technical nicety. It is the product itself, as far as the AI is concerned.
Here is the short version. In 2026, OpenAI launched "Buy it in ChatGPT," powered by an open standard called the Agentic Commerce Protocol, developed with Stripe. It lets people discover products inside ChatGPT and, for participating merchants, buy them via Instant Checkout without leaving the chat. Crucially, OpenAI states that product results are organic and unsponsored, ranked purely on relevance to the shopper. You cannot pay to jump the queue. To appear, a merchant needs a clean, complete, accurate product feed; genuinely descriptive product information written for a shopper rather than a search crawler; live inventory and prices that match your website; and proper structured data. This guide explains each of these in plain language, and tells you what to do first.
What Is ChatGPT Shopping, and How Is It Different From Google?
It is a way for people to find, and increasingly buy, products through a conversation rather than a search results page. When someone asks a shopping question, ChatGPT shows the most relevant products from across the web, and, according to OpenAI, these results are organic and unsponsored, ranked on relevance to the person asking rather than on who paid the most. This alone makes it profoundly different from the advertising-driven model most merchants know.
There is a distinction worth understanding here. "Conversational" commerce simply uses a chat interface to help a human shopper. "Agentic" commerce goes further: it allows an autonomous AI agent to research, compare, and even purchase on the person's behalf. The Agentic Commerce Protocol is the open standard, built by OpenAI together with Stripe, that connects merchants to this system. It is the plumbing that lets ChatGPT understand your inventory and, where you have enabled it, complete a purchase. OpenAI's Instant Checkout, launched in the United States and now available to free as well as paying ChatGPT users, is the most visible expression of it, though it must be said that OpenAI has been actively reworking its approach, so the exact mechanics continue to evolve.
For the merchant, the practical implication is refreshingly clear: you are not fighting an auction. You are being judged, more or less, on whether your product genuinely fits the request and whether the AI can read your data with confidence. That is a game a small, careful merchant can absolutely win.
How Does the AI Actually Find Your Products?
Through structured product data, chiefly a product feed (the file that lists your entire catalogue in a form machines can read) and the structured markup on your site (hidden labels in your page code that describe each product precisely), read against the natural-language request a shopper makes. When someone asks for "a gift for a ceramics lover under 60 dollars," the AI is matching the meaning of that request against the meaning it can extract from your product information. So the completeness and accuracy of that information is, quite literally, the difference between appearing and not.
The Agentic Commerce Protocol works through a structured product feed: a file, submitted in a compressed format, containing the required fields for each product: a clear title, a full description, price, availability status, images, and various eligibility flags. Alongside this sit the checkout endpoints and payment integration that make Instant Checkout possible for participating merchants. But for the vast majority of small merchants, the discovery question comes first: can the AI see your product clearly and trust what it sees?
And this is where a subtle but important point emerges. The descriptions that work for AI discovery are not the flowery marketing copy of old, nor the keyword-stuffed text written to please a search crawler. They are plain, specific, use-case-oriented descriptions: what the product is, who it is for, what it does, its real specifications. An AI recommending "waterproof boots for wide feet" needs to actually find the words "waterproof" and "wide" and the relevant measurements in your data. Write for the machine as you would explain the product honestly to a friend, and you will tend to satisfy both.
What Does "Clean Product Data" Actually Mean in Practice?
Completeness, accuracy, and freshness: the three virtues that sound dull and win quietly. Let me be concrete, because vagueness helps no one here. The signals that matter most for a merchant hoping to be discovered and recommended are these:
- Fill in every core attribute. Title, description, price, availability, images, category, key specifications: do not leave fields blank. A high fill rate across your core attributes is one of the strongest signals of a trustworthy catalogue.
- Keep inventory synchronised in real time. Nothing erodes trust, for a shopper or for the AI, like a recommended product that turns out to be out of stock. Your availability status must reflect reality, continuously.
- Match prices exactly to your live site. A price in your feed that differs from the price on your product page is a red flag that can quietly disqualify you. Accuracy here is non-negotiable.
- Write descriptions for use, not for keywords. Explain what the product is and who it suits, in plain specific language. Marketing fluff and keyword stuffing both hurt you now.
- Use high-quality images and complete details. The AI, and the shopper reading its answer, both rely on clear imagery and full specifications to build confidence.
- Apply correct structured data on your site. Proper product schema markup (price, availability, reviews, brand) helps machines read your pages without ambiguity.
None of this is glamorous. But that is precisely why it works. In agentic commerce, as the practitioners like to say, data quality is not merely a ranking problem; it is a revenue problem.
Where Do Product Reviews and Structured Data Fit In?
They fit in as trust signals, and they matter more than they first appear. An AI deciding among several merchants selling a similar product weighs factors such as availability, price, and whether you are the primary seller, and the broader reputation of your store, expressed through genuine reviews and consistent information, feeds the model's confidence in recommending you at all. This is where classic SEO hygiene and modern AI-readiness converge beautifully.
Structured data, specifically schema markup, is the quiet hero of this whole story. When your product pages carry proper markup declaring the price, the availability, the brand, and the aggregate review rating, you remove ambiguity for every machine that reads your site. We wrote a plain-language walkthrough in our 15-minute schema markup guide for small businesses, and the principles there apply directly to product pages. Get this foundation right and you serve not only ChatGPT Shopping but Google's own shopping and AI systems at the same time: one effort, many doors.
What Should a Small Store Do First?
Audit your product data before anything else; this is not the moment for clever content strategy. The correct sequence, borrowed from the merchants doing this well, is to fix the infrastructure first and worry about optimisation second. Start by reviewing your product feed for completeness and accuracy. Clean up your structured markup. Ensure your Google Merchant Center feed is compliant and current, since much of this ecosystem draws on the same underlying data. Then automate your feed updates so that inventory and prices stay fresh without daily manual effort.
The reason to move now rather than later is straightforward. The window to become the default recommendation in your category is not permanently open; early, well-structured merchants tend to earn a durable position while their competitors are still leaving fields blank. The parallel to the broader shift in search is exact, and if you want the fuller picture of getting a business recommended by AI, our guide on how to rank on ChatGPT in 2026 sets out the wider strategy.
At Licheo, this is the kind of unglamorous, high-return work we do for clients as a done-for-you service: auditing feeds, cleaning schema, and building the machine-readable structure that lets AI shopping agents recommend you with confidence. If you would like to see how your products and store appear to AI systems today, our SEO Standings assessment shows you plainly, and you can read about the full approach on our done-for-you SEO page.
Frequently Asked Questions
Can I pay to rank higher in ChatGPT Shopping?
According to OpenAI, no. Product results are organic and unsponsored, ranked purely on relevance to the shopper. This is a genuinely important difference from advertising-driven marketplaces. Your path to visibility runs through the quality, completeness, and accuracy of your product data rather than through a bid. For a small merchant, this is very good news, because careful data work is something you can control entirely.
Do I need Instant Checkout enabled to appear in ChatGPT?
Not to be discovered. Product discovery and recommendation depend chiefly on clean, structured product data that the AI can read and trust. Instant Checkout is a separate capability that lets participating merchants complete a sale inside the chat, and OpenAI has noted that enabling it does not, by itself, make your products preferred in the results. Focus first on being discoverable; enabling checkout is a further step you can weigh separately.
What is the Agentic Commerce Protocol in plain terms?
It is an open standard, developed by OpenAI with Stripe, that acts as the connective layer between merchants and ChatGPT. It defines how a merchant supplies a structured product feed, how checkout works through a set of endpoints, and how payment is handled. In everyday language: it is the agreed set of rules that lets ChatGPT understand your catalogue and, where you allow it, help a customer buy from you directly.
I run a tiny store. Is this even worth my time?
Very much so, and arguably more so than for the giants. Because most catalogues are incomplete and out of date, a small merchant who puts their product data genuinely in order can earn AI recommendations that punch far above their size. You are not fighting an ad budget; you are competing on data quality and product fit, which is a contest a careful small business can win. Start with your feed and your structured data, and you have taken the step that matters most.