ChatGPT Instead of Google? How Your Shop Gets Recommended in AI Answers

Hardly any topic currently generates as much excitement – and as much half-knowledge – as AI search in e-commerce. Some proclaim the end of Google, others wave it off. Both are wrong. The truth lies in the data, and it paints a nuanced picture: AI is noticeably changing the way people research products – but it hasn’t yet overturned the act of buying itself. Whoever understands this can position themselves early and with a sense of proportion. This guide shows what is changing, what the honest data looks like, how AI systems select shops and what you can concretely do.

What is changing in search behaviour

Classic search works as a list: you enter a term and get ten blue links from which you choose yourself. AI search works as an answer: you ask a question – often a complex one, with context and requirements – and get a curated recommendation. Instead of “best running shoes”, someone types “which running shoes for beginners with wide feet under 120 euros”.

For e-commerce this shifts above all the research phase. People use AI to narrow down options, compare products and clarify decision questions before they even land on a shop page. The consequence: whether your product appears in this upstream recommendation increasingly decides whether it makes the shortlist.

For you as a shop operator, the central question thereby shifts. It is no longer just: do I rank for the right search terms? But: am I selected by an AI system as the fitting answer to a prospective buyer’s question? That is a different competition – one in which the page with the most backlinks doesn’t necessarily win, but the source that most clearly and completely answers the decision question.

The honest data

Here a sober look pays off, because the numbers tell two stories at once.

On the one hand a clear growth trend: according to Adobe data, AI referral traffic to retail sites grew by 393% year on year. And these visitors are valuable – they convert around 42% better than non-AI traffic and spend 48% more time on the site. Whoever comes via an AI recommendation is thus pre-qualified: the AI has already made the preselection, the user arrives with clear intent.

On the other hand the context that likes to get lost in the hype: according to a study of 973 shops, organic LLM traffic so far accounts for under 0.2% of all visits. In absolute terms the channel is therefore still small. AI is changing primarily the research phase, not the checkout.

What matters is reading both numbers together and not selectively choosing the one that suits your own agenda. Whoever only cites the 393% growth creates hype; whoever only names the 0.2% sleeps through a development. Only together do they yield a usable picture – a small but fast-growing and high-quality channel that you should take seriously but not overrate.

The right conclusion lies between the extremes: AI search is no reason for panic and no replacement for your existing channels – but an early-growing, high-quality channel in which a good position can be built with manageable effort, before everyone does it. It is an opportunity with a good risk-reward ratio, not tomorrow morning’s revolution.

How AI systems select shops

To get recommended, a basic understanding of where AI systems draw their information from helps. One important building block: the Bing index feeds ChatGPT search. Your visibility in classic search indexes thus remains relevant – it is one of the sources from which AI answers draw.

Beyond that, AI systems prefer content they can easily understand and reliably classify. Machine-readable structure, complete product information and content that answers real questions have a clear advantage over unstructured pages. That is the logic behind Generative Engine Optimization (GEO): you don’t optimise for a ranking position but to be included in an answer as a trustworthy, well-structured source.

From this follows a principle that distinguishes GEO from classic SEO: it’s less about keywords and rankings than about clarity and trust. An AI system must be able to grasp at a glance what you offer, who it’s meant for and why it’s credible. Contradictory details, missing prices or unclear product descriptions are a small annoyance for a human – for an AI system a reason to rather not recommend you, because it can’t classify the information reliably.

GEO checklist for shop operators

Concretely, the starting position can be improved with a few levers that reinforce each other:

• Mark up structured data. Product, Offer and Review schema make your product pages machine-readable. That way an AI system reliably understands what you offer, what it costs and how it’s rated.
• Keep product feeds complete. Complete, clean product feeds are the basis for your products being captured and compared correctly. This data maintenance simultaneously pays into your paid channels – how this works in detail we show in the article on our Feed Hacking.
• Allow AI crawlers. If you want to appear in AI answers, the relevant crawlers must be allowed to read your pages. In practice this means allowing relevant AI crawlers such as the OAI-SearchBot in your robots.txt. Whoever locks them out excludes themselves.
• Create buyer-guide content. Content that answers real decision questions – comparisons, guides, purchase criteria – is exactly the material from which AI systems build recommendations. A good guide that honestly answers your audience’s questions is perhaps the most effective GEO measure of all.

None of these points is exotic. They are solid foundations of clean e-commerce work – which now gain a second benefit.

How to make AI traffic visible in analytics

A practical problem at the end: you can’t steer a channel you can’t see. AI referral traffic is often misattributed in standard analytics and wrongly ends up under direct or organic traffic. Whoever doesn’t look specifically underestimates the contribution of AI search systematically – and overlooks a channel that converts above average.

Just as important is the right expectation. Because the channel is still small in absolute terms, you shouldn’t budget GEO as a replacement for existing channels but as cheap groundwork that pays into many goals on the side: better structured data helps your classic SEO, cleaner feeds your paid channels, clearer buyer guides your conversion. GEO is therefore rarely wasted effort – even if AI traffic stayed small for the foreseeable future, you would have made your shop fundamentally more understandable and trustworthy. That is exactly why getting started with a sense of proportion pays off, rather than waiting for the perfect moment.

The first step is therefore to make AI referrals visible as their own source in your analytics setup, instead of letting them disappear into the “direct” catch-all. This attribution question is closely related to the measurement problems that automated paid campaigns bring – why you shouldn’t rely on a single platform figure, you can read in the article MER instead of ROAS.

The long-term building of visibility and trust – whether in AI answers or in social feeds – is moreover closely connected to organic community building, as we support it in our Organic Social CRM work.

Conclusion

AI search does not replace Google in e-commerce – at least not today and not at the checkout. But it changes the research phase, brings above-average valuable visitors and grows fast. The smart approach is neither panic nor ignorance, but early, solid preparation: structured data, complete feeds, accessible pages for AI crawlers and honest buyer-guide content. Whoever lays these foundations now positions their shop for a channel that starts small but delivers good quality – and that will very probably become more important.