How to appear in ChatGPT and other AI assistants when they recommend products

Customers no longer always type "women's trail running shoes" into a search engine. More and more often they ask an assistant: "I need running shoes for mountain trails, I overpronate, my budget is around a hundred euros, which would you recommend and where can I buy them?". The assistant answers with a conversation, compares options and sometimes links to stores.

This is different from the AI summaries that appear within Google's results page. Here there is no list of ten links: there are a few recommendations and, with luck, the sources they come from. Let's look at what can be done to make your store one of those sources, without promises nobody can keep.

Where they get their information

Conversational assistants combine two things:

1. What they learned during training: general knowledge, with a cut-off date, that knows nothing about today's price or your stock.
2. Real-time searches: when the question calls for it, they query the web on the spot, with their own crawlers or relying on search engine indexes, and read the pages they find.

For shopping questions, the second one is what counts. And there the rules are very similar to traditional SEO: the assistant needs to be able to reach your page, read it and understand it. In addition, some assistants already accept product feeds from merchants, just like Google Shopping, which makes a well-structured catalogue even more important.

Step one: let them in

Many stores block AI crawlers without knowing it, with robots.txt rules copied from somewhere or with a firewall that rejects anything that doesn't look like a browser. Each provider has several crawlers with different purposes. In OpenAI's case, for example:

• GPTBot collects content to train models.
• OAI-SearchBot crawls for ChatGPT search.
• ChatGPT-User visits a page when a user asks for it in the conversation.

Other providers follow a similar pattern. You can block the training crawler if you don't want your content used for that and allow the search crawlers, which are the ones that can bring you visits.

Step two: make sure they can read it

• Content in the HTML. Many AI crawlers don't run JavaScript. If your price, description or features are loaded later with JavaScript, for them they don't exist.
• Speed. A crawler waiting for a response has limited patience. A product page that takes several seconds to respond may go unread.
• No walls: popups that cover all the content until you accept something, or product pages only visible after logging in.

This is where the usual thread comes in: a store on PrestaShop 1.6 or 1.7 with old PHP, an unmaintained theme and ten modules loading scripts is usually slow and outputs incomplete structured data. Keeping your store up to date isn't just about security: it is about the machines that read your website being able to understand it.

Step three: make sure they understand it

An assistant that has to recommend products looks for concrete data it can compare and cite. The product page that wins is the one that provides it unambiguously:

• Full name with brand, model and variant.
• Features as data: dimensions, weight, materials, compatibility, power, size. Better as product features than buried in a paragraph.
• Who it is for and who it isn't for. "Suitable for neutral runners; if you overpronate, look at this other model" is exactly the kind of sentence an assistant reuses.
• Real FAQs about the product.
• Price, availability and delivery time visible and consistent.
• Your own texts, not the manufacturer's description that a hundred other stores also have.

Structured data: necessary, not magic

Mark up your products with Product and Offer (price, currency, availability, condition), identifiers such as the GTIN when you have them, the brand and, if they are genuine, the ratings. Add breadcrumbs and your organisation's details. None of this guarantees a citation, but it reduces ambiguity: an assistant that reads "in stock, €89.90, shipped within 24 hours" in a standard format has fewer reasons to get it wrong.

What is said about you outside your website

Assistants don't recommend based solely on your product page. They also read reviews, forums, comparisons and the press. A store that nobody mentions anywhere is less likely to appear, just as in traditional SEO. Verified reviews, presence in comparisons in your sector and useful content that others link to still count.

How to measure it

• In your analytics, look for visits originating from the assistants' domains. Some add a source parameter to their links, such as utm_source=chatgpt.com.
• In the server logs, see which AI crawlers visit your store and which pages.
• Ask the questions your customers would ask yourself and note whether you appear, with which product and with what data. Repeat every month.

What doesn't work

• Hidden text "for the AI". It is the same as hidden text for Google: it deceives and it gets detected.
• Magic files. There are proposals such as llms.txt, but there is no evidence that the big assistants use them to recommend products. They are no substitute for well-made product pages.
• Anyone promising you rankings in an assistant's answers. Nobody controls that.

Checklist

• Check robots.txt and the WAF: allow the search crawlers you are interested in.
• Make sure price, description and features are in the HTML without JavaScript.
• Improve the response time of product pages.
• Fill in features, uses, compatibility and FAQs.
• Correct Product and Offer structured data, with no duplicates.
• Measure visits from assistants and run your own queries every month.

How to do it in PrestaShop

With Rekire's AI-powered technical SEO service, if we are already your technical team we apply the improvements ourselves in your PrestaShop's code, database and configuration: metadata at scale with artificial intelligence, on-page SEO and performance, instead of a report that never gets acted on. If you'd rather do it yourself, Easy AI Magic Catalog Suite works on the catalogue in bulk from the back office: it completes product pages with data by EAN, extracts features from descriptions, writes key benefits and FAQs, translates and writes alt texts, all as drafts that you review before applying and with a history for undoing, and it audits each category with data from Google Search Console.