• Shopware Gold Partner & Shopify Partner

Chapter 1

How AI systems change purchase decisions

Table of contents

A customer wants a running jacket. Two years ago, he would have googled "waterproof running jacket men", got ten blue links and compared them himself. Today he types into ChatGPT: "I need a light, waterproof running jacket for winter, budget up to 150 euros." He does not get a list of links. He gets an answer. Two or three specific products, with reasons why exactly these fit.

The difference sounds small. It is not. In a Google search, the customer decides which of the ten links looks trustworthy. In an AI answer, that pre-selection has already been made. If your store does not appear among those two or three suggestions, the customer never saw it. There is no position ten to work your way up from. Either your store appears in that answer – or, for this customer at this moment, it does not exist.

What happens technically at that moment?

For specific product recommendations, ChatGPT, Claude, Perplexity and Google have long stopped relying only on the knowledge of their models. Current product data, web content and structured information play a central role – and that is exactly where it is decided whether your store is even in the running.

For ChatGPT, an analysis by Peec AI of around 43,000 product suggestions in March 2026 showed: 83 percent of the products in ChatGPT's shopping carousels were among the top 40 organic Google Shopping results. But this is exactly where you can see how fast the market moves: OpenAI is now building a direct product data infrastructure with its own product feeds and the Agentic Commerce Protocol. Merchants and platforms can use it to deliver product data straight to ChatGPT.

For store owners, the decisive point is therefore not optimizing for a single data source. What matters is keeping product data so clean and structured that it can be served through different systems.

Perplexity takes a similar approach with its merchant program: merchants can provide product catalogs and current product information directly so that products can be classified and recommended more accurately. Google, in turn, uses its Shopping Graph with more than 60 billion product listings as the central product data foundation for its shopping experiences, especially in AI Mode.

Why is this no longer a side issue?

The behavior is big enough that store owners can no longer ignore it – classic search remains relevant alongside it; one does not replace the other. For store owners, this has created a second, independent space for product search and purchase decisions, with its own rules.

What does this mean for your store?

What is new is that product data keeps gaining importance. If your product data is not under control, there is no point discussing AI readiness – it is the foundation everything else builds on, whether AI system or classic search engine.

What is also new is who is reading along. Besides Google and Bing, dedicated AI crawlers access store data. In stores we take over, we regularly see these crawlers being blocked unintentionally – for example by preconfigured rules in CDN and security services such as Cloudflare. The store works perfectly for people and still loses visibility to machines. That is exactly why technical accessibility now belongs on the basic checklist of every store. What this looks like in practice and what to check is covered in Chapter 2.

The third shift concerns the text itself: for years, many product texts were written mainly for keywords and classic search intent. AI systems add another question: does the page provide enough substance to derive a specific recommendation – suitability, use case, limitations? A page that only lists features gives an AI little to base a recommendation on.

Our assessment

Anyone who now believes a few meta tags or a new landing page will settle the matter underestimates it. And anyone who believes they have to convert their entire business model to AI visibility overestimates it.

The right reaction is fairly unspectacular: get the basics right first – product data that is correct, and accessibility for the systems that are already reading along today. The rest of this guide shows what that looks like in practice and where you should actually start.

Frequently asked questions

No, classic search remains relevant, and the two are closely connected. A second, independent space for product search has emerged – not a replacement.

What matters is not the individual interface but the data foundation. If you keep product data clean and structured, you can serve it through different systems – today to Google Merchant Center, tomorrow to a standard that does not exist yet.

The fastest way is technical accessibility: can AI crawlers reach your store at all? Chapter 2 covers how to check – or the free arboro AI Bot Checker does it in a few seconds.