How AI shopping assistants find products, and how to get your store ready

More and more people start their shopping with a question instead of a search box. They ask an AI assistant such as ChatGPT, Gemini, Claude, Copilot or Perplexity: “Find me a handmade leather bag under €300 that ships to Lyon.” The assistant reads the web, compares a few products and answers with names, prices and links. Some assistants can now also fill a basket for the shopper.

If your store is hard for these assistants to read, you’re not in the answer. This guide explains, in plain words, what they look at. Then it gives you a checklist you can work through this week.

What an AI assistant does when someone asks what to buy

Each assistant works a little differently, but the steps are close.

1. It searches. The assistant runs web searches, often through a search engine it works with, and visits the pages that come back.
2. It reads pages as text. Many AI crawlers fetch the raw HTML and don’t run JavaScript. If your product name, price and stock only appear after a script runs, the assistant may see an empty page.
3. It looks for facts it can trust. These are price, currency, availability, delivery countries and times, returns and reviews. Structured data, a hidden machine-readable summary of the page, makes these facts unambiguous.
4. It compares and answers. It picks a few products that match the question and explains why.
5. Sometimes, it acts. New open standards, such as the Agentic Commerce Protocol (OpenAI and Stripe, September 2025) and the Universal Commerce Protocol (Google, January 2026), let an assistant ask a store directly for products and prices. It can then prepare a basket that the shopper checks and pays for.

The good news is that an assistant needs what a human shopper and Google want too: clear, fast, honest pages.

The 10-point checklist

1. Make your product pages readable without JavaScript

Open a product page and use “view source”, or turn JavaScript off in your browser. Check that the name, price, description and “in stock” are there as text. If they aren’t, ask your platform how to render pages on the server.

2. Add complete product structured data

Every product page should carry Product structured data with an Offer: the price, currency, availability and condition, plus the shipping and return details. Group products with options (sizes, colours) so the assistant understands they’re one product. You can test a page with Google’s Rich Results Test.

3. Write descriptions that answer real questions

Assistants quote what you write. Say:

• what the product is made of;
• its exact size;
• how it’s made, and who makes it;
• how to care for it;
• when it arrives.

“Hand-stitched from vegetable-tanned leather, 32 × 24 cm, ships in 3–5 days to the EU” beats “Our iconic bag, timeless elegance.”

4. Put delivery and returns on the page, in numbers

For example: “Free delivery over €150 to France, Belgium and Switzerland, 2–4 working days. Returns within 14 days.” Shoppers ask assistants about delivery all the time. If the answer isn’t on your page, the assistant will pick a store where it is.

5. Speak your customers’ languages

If you sell in several languages, give each one its own address (for example /fr or /de) and tell search engines about them with hreflang. Write each version for its readers instead of relying on a rough translation. A shopper who asks in their own language is far more likely to get a page in that language in the answer.

6. Publish a product feed

A product feed is one file that lists every product with its price and stock. Google Merchant Center reads it, and so can other tools. Keep it updated automatically, never by hand.

7. Add an llms.txt file

llms.txt is a short plain-text file at the root of your site (yourstore.com/llms.txt). It tells AI systems what your store is, what you sell, where you deliver and where to find the catalogue. It’s a young convention, not a guarantee, but it costs nothing and helps the assistants that read it understand you quickly.

8. Welcome the right crawlers in robots.txt

Check that your robots.txt doesn’t block search engines or AI search crawlers. Keep private pages, such as the basket, checkout and account pages, closed. Decide, as a business, whether you also allow crawlers that collect data to train AI models.

9. Be fast

Slow pages are crawled less and sell less. Aim for a mobile score above 90 in Google Lighthouse. To get there, compress your images (WebP), host your fonts on your own site and keep heavy scripts off product pages.

10. Keep reviews real

Assistants, shoppers and regulators all pay attention to reviews. Show only real reviews from real buyers. In the EU, fake or unverified reviews presented as genuine are illegal.

What this looks like on OumsAI

Every OumsAI store does the technical part of this checklist for you:

• pages rendered on the server, readable without JavaScript;
• product structured data with offers, options, delivery and returns, plus your shop’s details;
• a separate address for each of your store’s languages;
• a sitemap, a product feed and a public product catalogue for machines;
• an llms.txt file, and a robots.txt that welcomes search engines and AI assistants while keeping baskets and checkouts private;
• a connection for AI shopping assistants. They can search your catalogue, check prices and delivery, and prepare a basket. Your customer always reviews it and pays on your own checkout.

You still write the descriptions and set the delivery rules, and the AI store manager helps you with both, in most languages.

See how selling in AI chats works → · How AI assistants shop on OumsAI stores →

Questions

Can I pay to appear in AI answers?

Some assistants now show paid placements, but the regular answer depends on the quality of your pages. Nobody can promise you a place in an assistant’s answer, and we don’t either.

Will AI assistants take my sales away from my site?

Not if your store is built for them. A good set-up sends the shopper to your own checkout, with your prices, your brand and your customer relationship.

Do I need a developer?

For most of this list, no, although it depends on your platform. On OumsAI, the technical points are built in.

Is SEO still worth it?

Yes. The same pages that rank in Google are the ones AI assistants find and quote. Good SEO and good AI visibility are mostly the same work.