Skip to content

SEO + AIBy Mariano González Campas · 11 min read

GEO for ecommerce: get AI to recommend your store

Online store catalog with a sneaker, a backpack and a mug whose product cards —price, stock and star ratings— feed an AI assistant's answer that recommends two products and shows a cart

"What running shoes would you recommend for under $100?" That question used to end in a list of ten links. Today it often ends in an answer with three products, their prices and a button to the store. If your store isn't one of those three, it didn't exist for that shopper.

GEO —Generative Engine Optimization— is the work of getting AI engines to find, understand and recommend what you sell. For an online store there's a twist: what the AI recommends isn't an article, it's a product with a price, stock, shipping and reviews. And that data can be organized. This guide covers what each engine looks at according to its own documentation, what a store can do today, which AI checkout programs are still limited to a few countries, and what isn't worth buying yet.

Key takeaways

  • AI assistants pick products based on data: price, availability, images, reviews, shipping and returns. An incomplete product page loses to a complete one even when the product is better.
  • ChatGPT can find your store by crawling the web, without you sending anything. OpenAI's product feed gives you more control, but as of November 16, 2026 its shopping experience is live only for users in the United States.
  • Google says there are no extra requirements and no special schema for AI Overviews and AI Mode. For products it recommends structured data on the page plus a Merchant Center feed, and using both.
  • Buy buttons inside AI (Google's UCP-powered checkout, Copilot Checkout) are currently for merchants selling in the US, Canada or Australia, and Copilot's is still a pilot. Everywhere else, the work that pays off is on your own site.
  • What moves the needle most: complete, consistent product pages, Product schema with offer, shipping and returns, real reviews, and content that answers the questions people compare with.
  • Nobody can guarantee that ChatGPT will recommend your product. You can measure how often it does, and compare that against your competitors.

Why an online store needs GEO (not just SEO)

In classic Google search, a store competes to show up in a list and the shopper does the rest: clicks, compares, comes back. In an AI answer, the engine does the comparing. It reads product pages, reviews and comparisons, filters by what the user asked for —price, size, delivery to their city— and returns two or three options with a reason for each. The shopper arrives at your store with the decision nearly made, or doesn't arrive at all.

That changes what ranking well means. It's no longer enough for your "running shoes" category to rank: every product page has to hold up on its own, with the data the engine needs to choose it. And the foundation is still SEO: if Google doesn't index your product pages, or if the price only shows up after JavaScript runs, no engine can use them.

What each engine looks at to recommend a product

No engine publishes its recommendation algorithm, but the big three document which product data they accept and how they prefer to receive it. Here's what each one says, as checked on November 16, 2026:

ChatGPT

OpenAI says ChatGPT uses information from the public web, private integrations and, for retail, merchant product feeds. On its merchant page it names images, price and reviews as signals, and notes that results are tailored to each shopper's preferences. To the question "do I need a feed if ChatGPT already crawls my site?" it answers no, but that a feed gives you more control over how your products appear.

OpenAI's feed requires nine fields per product —ID, title, description, URL, brand, seller, image, availability and price— plus target countries. As of this writing, its own documentation says the standard format targets the United States and that shopping in ChatGPT is live for users there. OpenAI has also said it's moving away from Instant Checkout as a standalone experience: purchases are completed on the merchant's own website or app. So for most stores, the way into ChatGPT today is making sure its crawler can read the store well: allow OAI-SearchBot in robots.txt and keep product pages readable without JavaScript.

Google: AI Overviews, AI Mode and Gemini

Google is explicit: there are no additional requirements and no special structured data to appear in AI Overviews and AI Mode. The page has to be indexed and eligible for a snippet. For products, its documentation describes two routes —Product schema on the page and a Merchant Center feed— and says that providing both maximizes eligibility. The same AI features guide recommends keeping your Merchant Center and Business Profile information up to date.

What's new is checkout inside AI Mode and Gemini, built on the Universal Commerce Protocol (UCP): a buy button on listings that have it enabled. According to Merchant Center Help, it applies to products eligible in the United States, Canada and Australia, and the integration is rolling out gradually in the US.

Microsoft Copilot

Copilot Checkout lets shoppers buy inside Copilot, with the merchant as the seller of record. It requires a Microsoft Merchant Center account, a UCP-compatible feed, a return policy and customer support set up, and an approved payment provider. It's currently a pilot, and only English-language merchants selling to US buyers in US dollars are eligible. Copilot and Bing's AI summaries, on the other hand, cite web pages from anywhere: that's where your store's content competes, and you can measure it with the AI Performance report in Bing Webmaster Tools.

Perplexity

Perplexity had a merchant program and "Buy with Pro" shopping, but as of November 16, 2026 we couldn't find current requirements, countries or a feed on its official pages. We don't treat it as a channel you can work with your own data. Its answers do cite web pages, so everything that applies elsewhere applies here too.

What a store can do today, as of November 16, 2026
EngineHow it finds productsMerchant program or checkoutOpen to any store?
ChatGPTCrawling the public web and merchant feedsOpenAI product feed; shopping live for US usersCrawling, yes; feed and shopping, US first
Google AI Overviews and AI ModeIndexed pages, Product schema and Merchant CenterUCP checkout in AI Mode and Gemini (US, Canada, Australia)Schema and Merchant Center, yes; checkout, limited
Microsoft Copilot and BingPages indexed by Bing and Microsoft Merchant CenterCopilot Checkout (pilot, US merchants in USD)Web citations, yes; checkout, limited
PerplexityCited web pagesNo current program confirmed in an official sourceWeb citations, yes

What to do in your store, in order

The list below is ordered by what pays off most for the least effort. The first two steps help with Google and with every AI engine, so none of the work is wasted.

1. Make sure engines can read your product pages

  • Check your robots.txt: if it blocks OAI-SearchBot, Claude-SearchBot or PerplexityBot, those engines can't cite your store. Blocking training bots (GPTBot, ClaudeBot, Google-Extended) is a separate decision and doesn't remove you from search.
  • Confirm that the name, price and stock are in the HTML the server returns, not only after JavaScript loads. Many custom-built stores, or stores with heavy apps on top, fail here without knowing it.
  • Look at Search Console to see how many product pages are indexed. If a large part of the catalog shows up as "Crawled – currently not indexed," that's the first problem to fix.

2. Complete product structured data

Product schema with its offer (price, currency and availability) is the minimum. What's usually missing, and what Google documents for merchant listings, is shipping —cost, destination and delivery time— and the return policy. Those are exactly the data points an assistant uses to answer "will it arrive before Friday?" or "can I return it?" Add the brand, the GTIN if you have one, and the aggregate rating when the reviews come from real buyers.

If your store runs on Shopify, much of this comes built in or with an app, but it's worth checking: custom themes and review apps often leave the schema incomplete or duplicated. Google's Rich Results Test shows you in a minute what it's reading.

3. A Merchant Center feed

Google recommends combining schema with a Merchant Center feed and free listings. Check in Merchant Center Help that free listings are available in the countries you sell to. The feed adds another source of product data for Google, and it's the same file you'll need when AI shopping programs open up in your market. What matters is that the feed and the page say the same thing: a different price in each place is the fastest way to lose trust, with Google and with shoppers.

4. Product pages that answer what people ask

An AI recommends a product when it can justify it: "it's waterproof," "it works for wide feet," "it fits in an overhead bin." If that information isn't written on your product page, the engine will look for it in another store or a third-party review. For each important product, write the answers to the three or four questions customers ask you most often: dimensions, compatibility, who it's for, how it differs from the cheaper version. You don't need long copy; you need the facts to be there.

5. Your own buying guides and comparisons

Shopping questions rarely name a product. They sound like "which backpack works as a carry-on" or "moka pot or drip coffee maker." Pages that answer those with judgment —a table, honest pros and cons, who each option suits— are the ones engines cite, and they link to your product pages from there. One solid guide per category is worth more than twenty generic blog posts.

6. Reviews and presence beyond your site

Engines don't rely only on what a store says about itself. Buyer reviews on your site and your Business Profile, press mentions, third-party comparisons and your listings on marketplaces like Amazon or Mercado Libre are all sources an AI cross-checks before recommending you. Ask for a review after every delivery and reply to them: there's no shortcut, and buying reviews is the surest way to get penalized.

How to tell if it's working

GEO measurement for a store comes from three sources. The first two are platform data; the third is your own sample, and it's worth keeping them apart.

  • Search Console: the generative AI reports show your impressions in AI Overviews and AI Mode by page. Check which product pages and guides appear.
  • Bing Webmaster Tools: the AI Performance report counts how often Copilot and Bing's summaries cite your pages, and for which queries. We explain it in our Bing AI Performance guide.
  • Your analytics: visits from chatgpt.com, perplexity.ai, gemini.google.com or copilot.microsoft.com, and how much they buy. They tend to be few and high-intent: compare their conversion rate with everyone else's.
  • Your own sample: ten or twenty real shopping questions from your category, each phrased a few different ways, repeated in ChatGPT, Gemini, Perplexity and Copilot. Note whether you're named, which product, and who gets recommended when it isn't you. Repeat it every month with the same questions.

A screenshot of ChatGPT recommending your product proves nothing: it's one run among many possible ones. What counts is the rate —how many out of ten answers include you— and how it moves month to month.

Common (and expensive) mistakes

  • Paying for a guaranteed "ranking in ChatGPT." There's no stable ranking and no way to guarantee it.
  • Buying AI feed tools for a program that isn't available in your market yet. Product pages and schema first; the feed when there's somewhere to use it.
  • Filling product pages with AI-generated text that rewrites the manufacturer's description. Engines already have that description: what they're missing is your information, the kind that isn't anywhere else.
  • Having different prices or stock across the product page, the schema, the feed and your marketplace listings. Engines compare, and inconsistency counts against you.
  • Forgetting about categories. Many shopping questions are answered by a category page or a guide, not a single product page.

Frequently asked questions

What is GEO for ecommerce?

It's the work of getting AI engines —ChatGPT, Gemini, AI Overviews, Copilot, Perplexity— to find, understand and recommend an online store's products. It combines technical SEO, product structured data, product pages and guides that answer shopping questions, reviews, and repeated measurement of whether the engines recommend you.

How do I get ChatGPT to recommend my products?

Nobody can guarantee it, but you can raise the odds: allow OAI-SearchBot in robots.txt, keep price, stock, shipping and reviews readable without JavaScript, add complete structured data, and publish content that answers the questions people compare with. OpenAI's product feed gives you more control, but as of November 16, 2026 its shopping experience is live only for US users.

Can I sell directly inside ChatGPT or Google AI Mode?

Only in a few markets so far. Google's UCP checkout applies to products eligible in the US, Canada and Australia; Copilot Checkout is a pilot for merchants selling to US buyers in US dollars; and OpenAI moved away from standalone Instant Checkout, so purchases finish on the merchant's site. Anywhere else, what you can do is get those engines to recommend and link to your store.

Do I need special schema to appear in AI Overviews?

No. Google says there are no additional requirements or special structured data for AI Overviews and AI Mode. For a store, what helps is the usual Product schema done completely —offer, shipping, returns, reviews— plus a Merchant Center feed with the same data.

Doesn't Shopify already handle GEO?

It covers part of it: it generates basic structured data and crawlable product pages. It doesn't write the information that sets your product apart, build buying guides, or measure whether AI recommends you. And custom themes or apps can break the schema, so it's worth checking.

How long does it take to see results?

Technical fixes —robots.txt, schema, readable product pages— can show up within weeks. Getting engines to recommend you consistently usually takes months and depends on how much outside presence your brand has. Measure in 90-day cycles.

At loco22 we do GEO for online stores as part of our SEO + GEO service: we review what the engines see in your catalog, fix product pages and structured data, and measure every month which answers include you and which include your competitors. If you'd rather start with the technical side, our GEO Radar is free.

ShareXLinkedInWhatsApp

Found this useful? Add us as a preferred source in Google.

(Keep reading)