Commerce GEO: How to Get Your Products Recommended by AI

Short answer: to get your products recommended by AI, make them easy for each assistant to find, read and trust. That means being in the product databases the engines draw on (Shopify Catalog, Google Merchant Center, Microsoft Merchant Center, direct feeds), having product pages whose facts a crawler can read and quote, and having enough independent reviews and mentions that an assistant believes you. A shopper who asks ChatGPT or Perplexity "what's the best non-toxic crib mattress?" gets a short list of products. If you're on it, you win a sale you didn't pay for. If you're not, you may never know the question was asked.

Commerce GEO (Generative Engine Optimization)

Optimizing a brand's product data, content and reputation so that AI assistants such as ChatGPT, Gemini and Google AI Mode, Microsoft Copilot, Perplexity and Alexa for Shopping surface, cite and recommend its products when shoppers ask buying questions.

How is Commerce GEO different from SEO?

SEO and GEO share a foundation: clean pages, structured data, real authority. The output is different. SEO earns a ranked link a person chooses to click. GEO earns a mention inside a generated answer, often with no click at all, and for products it also depends on feeds that classic SEO never touched.

GEO is additive to SEO, not a replacement: Google's AI features draw on pages that are already indexed and eligible for snippets, and the structured data that helps one helps the other. How to split budget between them is in GEO vs. SEO budget split: a decision framework.

Where does each AI engine get its product data?

Each engine documents some of where its product answers come from and keeps the ranking to itself. Below, documented means the platform says so in its own help or developer pages (cited). Observed means a pattern seen when running shopping questions through the assistant, which can change without notice and isn't a specification.

ChatGPT shopping

Documented. OpenAI's help center says product results are selected independently by ChatGPT and aren't ads. To choose products, ChatGPT considers structured metadata from first-party and third-party providers (price, product description) and other third-party content; when it lists merchants for a product, it ranks them on factors like availability, price, quality and whether they're the maker or primary seller. Shopify merchants' product data is integrated through Shopify Catalog with no additional work, and other merchants can apply to send a direct product feed (OpenAI Help Center).

Web answers depend on OpenAI's search crawler: OpenAI says sites that opt out of OAI-SearchBot won't be shown in ChatGPT search answers, while GPTBot is a separate crawler for model training. For Shopify stores, a purchase from the ChatGPT agentic storefront completes on your own checkout, opened in ChatGPT's in-app browser or a new tab. Observed: broad category questions tend to pull in editorial roundups and review content alongside product cards, so the third-party picture of your brand shapes which products make the list.

Perplexity, including its Merchant Program

Documented. Perplexity's crawler documentation says PerplexityBot surfaces and links websites in Perplexity's search results and isn't used to train foundation models; a separate agent, Perplexity-User, fetches pages when a user asks and generally ignores robots.txt. When Perplexity launched shopping in November 2024, it said its product cards aren't sponsored and are powered by platform integrations including Shopify, covering Shopify-powered businesses that sell and ship to the US. The same announcement introduced the Perplexity Merchant Program for large retailers: free to join, with richer product specs, a better chance of being a recommended product, payment integration for its Buy with Pro checkout, free API access and a dashboard. We haven't found updated program terms published since that announcement, so confirm eligibility before planning around it.

Observed: Perplexity cites its sources visibly and seems to favor pages with clear dates, plain declarative specs and ready-made comparison tables. The full set of observed patterns is in how Perplexity picks products.

Google AI Overviews and AI Mode

Documented. Google's guidance is short: there are no additional requirements to appear in AI Overviews or AI Mode, and you don't need new machine-readable files, AI text files or special schema. A page needs to be indexed and eligible to show with a snippet; structured data should match the visible text; and merchants should keep Merchant Center and Business Profile information up to date (Google Search Central).

Products reach Google's shopping surfaces through Merchant Center and the Shopping Graph, which Google said in November 2025 holds more than 50 billion product listings, about 2 billion of them refreshed every hour. Free listings are on by default for most Merchant Center accounts and can appear across Search, the Shopping tab, Images and Lens, YouTube and Gemini, though Google doesn't guarantee they'll be shown. Shopify's agentic storefront for AI Mode and Gemini runs through the Google & YouTube channel and Merchant Center. Observed: AI Overviews lean on pages that already rank well organically. More on that in Google AI Overviews for ecommerce.

Microsoft Copilot

Documented. Microsoft announced Copilot Checkout on January 8, 2026, rolling out in the US on Copilot.com with PayPal, Shopify and Stripe, and said product feeds from Microsoft Merchant Center help inform organic Copilot results. Shopify merchants were enrolled automatically, with controls in the Shopify admin; you remain the merchant of record (Microsoft Advertising). Microsoft's agentic commerce page adds that UCP-ready feeds are available in Microsoft Merchant Center for US businesses, and that Copilot Checkout is currently limited to English-language merchants selling to US buyers in USD, with a Merchant Center account required to onboard.

Brand Agents, announced the same day, are a different surface: Microsoft's AI shopping assistant on your own site, set up through Microsoft Clarity rather than Copilot. They're covered in Brand Agents explained. Microsoft doesn't publish how Copilot ranks products, so treat it like the others: get the feed in and keep your pages readable.

Alexa for Shopping (formerly Rufus)

Documented. Amazon renamed Rufus to Alexa for Shopping on May 13, 2026. Amazon says Rufus, now Alexa for Shopping, was trained on its product catalog, customer reviews, community Q&As and information from across the web (About Amazon). It answers inside Amazon, so your levers are the listing itself: complete product-type attributes, bullets and descriptions that answer real questions, images, reviews and Q&A. The Amazon side has its own playbook: Amazon listing optimization for Alexa for Shopping. Across all five engines, the common thread is the same: each one reads the clearest, most consistent facts it can find about a product, from a feed or from your pages.

What is the feed path, and which feeds do I need?

The feed path is how your products get into the databases the engines query, as opposed to the web pages they crawl. For most merchants it's fewer feeds than it looks, because Shopify Catalog and Google Merchant Center cover most of the surface.

Agentic storefront

Shopify's name for a store's presence inside an AI channel such as ChatGPT, Microsoft Copilot, Google AI Mode and Gemini, or Meta. Products reach those channels through Shopify Catalog, and the merchant controls per-channel settings such as direct checkout from the Shopify admin.

Shopify's agentic storefronts are active by default for eligible stores, and for ChatGPT the purchase completes on your own online store checkout; for AI Mode and Gemini, Copilot and Meta, customers can buy in a Shopify-powered checkout inside the channel if you turn direct checkout on (Shopify Help Center). Being in a feed only makes you eligible. Whether you're recommended depends on what the feed and your pages say, which is why the feed path starts with catalog enrichment.

One rule ties the two paths together: your feed, your page markup and your visible page must agree on price, availability and identifiers. When they don't, an engine either distrusts the listing or repeats the wrong number. Which schema.org types to use, and how to keep markup in step with the page, is in structured data for AI shopping agents.

How do AI assistants decide which products to recommend?

No assistant publishes its ranking for product answers. What's documented is thin but consistent: OpenAI names availability, price, quality and being the maker or primary seller as merchant ranking factors, and Google says it relies on your data to match products with what people search for. Beyond that, we work from a three-question model. It's our model, not a platform's specification, but every engine above behaves consistently with it:

1. 01

   Can it read you?

   Machine-readable product facts: a feed in the right database, schema.org markup in server-rendered HTML, specs stated as fields rather than buried in prose or a JavaScript tab. If an assistant can't work out what the product is and who it's for, it can't recommend it.

2. 02

   Can it trust you?

   Agreement across independent sources: reviews, editorial mentions, marketplace listings, forum discussion. A claim only your own site makes carries little weight. A claim that five independent sources repeat starts to read as fact.

3. 03

   Can it quote you?

   Short, factual, self-contained statements that answer the buying question directly: "the strap is 22mm wide", "safe for kids over 3". Assistants lift sentences like these and skip adjectives.

The trust question is usually the slowest to move, because you can't publish your way to it; how independent sources add up is in reviews, Reddit and the corroboration graph. The quote question is a writing discipline: see the quotable content style guide.

Do I need special markup, llms.txt or AI files to be recommended?

No special markup. Google says it outright for AI Overviews and AI Mode, and none of the other platforms' shopping documentation cited here asks for an AI-specific format. What you do need is the ordinary markup done well: Product and Offer structured data that matches the visible page, with price, currency, availability and identifiers that agree with your feed. Most failures are mundane: an Offer with no numeric price, availability as free text, or JSON-LD that only appears after JavaScript runs. Those patterns are in what a schema audit actually finds.

Crawlers are where merchants hurt themselves. Search crawlers and training crawlers are separate: OAI-SearchBot surfaces sites in ChatGPT search while GPTBot collects training data, PerplexityBot powers Perplexity's results and isn't used for training, and Google says its Google-Extended token doesn't affect inclusion or ranking in Search (Google Search Central). Blocking a search crawler removes you from that engine's answers; blocking a training crawler is a separate choice. Check your CDN's bot settings as well as robots.txt, because security defaults block crawlers nobody chose to block. The full decision is in should you block AI crawlers?

llms.txt is a proposal, first published in September 2024, for a markdown file that gives AI agents a curated map of a site. It's cheap to publish and does no harm, but none of the shopping documentation above mentions it, and Google says you don't need AI text files for its AI features. Treat it as optional housekeeping, not a growth lever: llms.txt for ecommerce.

What does a Commerce GEO plan look like, step by step?

In order of leverage, for a store starting from scratch:

1. 01

   Fix the product facts

   Hero SKUs first: complete attributes, valid GTINs or brand plus MPN, variant-level price and availability. Every later step repeats whatever the catalog says.

2. 02

   Open the feed path

   Confirm your Shopify agentic storefronts are on and your products are eligible, connect Google Merchant Center, and add Microsoft Merchant Center if Copilot matters to you. Clear feed errors before anything else.

3. 03

   Let the search crawlers in

   Allow OAI-SearchBot, PerplexityBot, Googlebot and Bingbot, check your CDN's bot rules, and keep prices and specs in server-rendered HTML.

4. 04

   Make product pages quotable

   Open each PDP with the facts that decide the purchase, stated plainly, then add a comparison table and FAQ where shoppers compare.

5. 05

   Earn corroboration

   Reviews on your site and marketplaces, editorial mentions, and honest participation where your category is discussed. Never fake it; astroturfing is detectable and burns the channel.

6. 06

   Cover constraint questions

   Budget, recipient, occasion, compatibility and size: the filters shoppers put in their prompts. If the answer isn't a field or a clear sentence on the page, you're filtered out before the shortlist.

7. 07

   Measure monthly

   Run the same panel of buying questions every month and fix the biggest gap each time.

Constraint questions are where mid-market brands lose most often, especially at peak. The worked example is why AI skips mid-market brands for gift queries.

Why does AI recommend my competitor instead of me?

When an assistant names a competitor instead of you, the reason is rarely that their product is better. It's usually that their data is easier to use. The usual causes, roughly in the order they're cheapest to fix:

• Their products are in a feed the engine reads (Shopify Catalog, Merchant Center); yours are missing, disapproved or incomplete there.
• Their PDPs carry full Product and Offer markup in the HTML; yours render it client-side, or not at all.
• Their key facts are stated as fields and plain sentences; yours sit in images, tabs or marketing copy.
• They have consistent reviews and editorial mentions; your third-party footprint is thin or contradicts your PDP.
• They publish a clear comparison an assistant can quote; you publish copy it can't.
• Your robots.txt or CDN blocks a search crawler, so the engine never sees your pages at all.

Finding which cause applies to you means testing the answers directly, question by question, and recording who gets named and which sources the assistant leans on. That's the method in how to run an AI citation audit.

How do I measure whether AI recommends my products?

Use a fixed panel of real buying questions, run across the assistants on a schedule with identical wording, and scored the same way each time. The audit guide linked above owns that method, including how to build the panel and score answers, so it isn't repeated here. Around the panel, use the data the platforms do give you:

• Google Search Console. Google counts traffic from AI features in the Performance report under the Web search type. It isn't broken out separately, so watch the pages and queries you're optimizing.
• Merchant Center AI performance insights. A report on how your products perform for shopping queries in AI Mode and AI Overviews, compared with competitors. It covers English-language queries for accounts in the US, Canada, Australia, New Zealand and India, and organic traffic only.
• Referral traffic. Visits referred by assistant domains in your analytics. Observed: these undercount, because many answers send no click.
• Brand search. A rise in branded searches after you start appearing in answers is one of the few footprints of a recommendation without a click.

What a "good" result looks like depends on your category and brand age; the bands are in the Commerce GEO benchmark. If you'd rather start with a baseline than build a panel, the free AI Citation Check report shows whether ChatGPT, Perplexity and Google AI Overviews name you, cite you or leave you out.