AI search engines and assistants have become a new entry point to a purchase. ChatGPT had roughly 800 million weekly users in autumn 2025 and some of those people no longer buy through classic search, they ask an AI for a recommendation. Traffic from AI is still small in volume, but US data puts its conversion rate about three times higher than classic search. At the same time the share of searches that end without a click keeps growing, so value is shifting from being first to being cited in the answer. For a store that means writing content an AI can read and recommend.
Comparison table
The facts in one place first. Figures as of July 2026.
| Metric | Classic search | AI search |
|---|---|---|
| Visit volume | high, established | small so far, growing fast |
| Visit conversion | around 3 % | around 3x higher per US data |
| Visit value | reference point | roughly 4.4x higher (US data) |
| The goal | be first in the results | be cited in the answer |
| What helps | SEO, links, speed | schema.org, llms.txt, reviews, Q&A content |
Figures on conversion and the value of AI traffic come predominantly from the US market (Adobe, Envive) as of 2025. The Czech market lags behind but moves in the same direction. Treat them as orientation, not a guarantee.
How many people already buy through AI
Try putting a question you used to type into Google into ChatGPT instead. Something like which coffee machine suits an office of ten people for under eight thousand. You get an answer with specific models, not ten blue links. Millions of people do exactly this every day.
The figures are mostly American so far, but they are unambiguous. ChatGPT had around 800 million weekly active users in October 2025, roughly double February of the same year. Perplexity handles around 780 million queries a month. Among AI platforms, ChatGPT holds a share of around 60 percent.
It is seeping into shopping quickly. Adobe reports that 38 percent of consumers have already used generative AI for an online purchase and 52 percent plan to this year. Newer surveys go higher still: 61 percent of people say they have used a tool like ChatGPT to buy something and 64 percent expect to in 2026. Most often for product research, recommendations and hunting for deals, which is precisely the stage where the decision about who gets the sale is made.
AI traffic is small so far, and expensive to ignore
Look only at visit volume and you will wave AI away as a couple of percent. That is a mistake, because the quality of the visit sits somewhere else entirely. Adobe reports AI traffic to retail sites growing 393 percent in a single quarter, and year on year figures for some sites run into thousands of percent, because they start from zero.
More important than the pace is the value. By the same data a visit from AI converts roughly three times better and is around 4.4 times more valuable than one from ordinary search. On visits where an AI assistant advises the shopper, conversion of up to 12.3 percent is reported against 3.1 percent for everyone else. It makes sense: somebody arriving on an AI recommendation arrives with the decision already made, not to start comparing.
I want to be clear that these figures come predominantly from the United States and the Czech market is behind. The direction is the same though, and in the Czech Republic the first wave of AI shopping is expected during 2026. Wait for hard Czech data and you give up the head start.
The zero-click era: being first is no longer enough
Here is the uncomfortable part. Roughly 60 percent of searches in 2025 ended without a single click, because the person got the answer on the page and went nowhere. Add a Google AI overview above the results and the click-through rate of the first position drops to something like 2.6 percent, with organic click-through falling by tens of percent overall.
Does that make the old race for the top spot pointless? Not entirely, but the rules have changed. The value moves from being first to being cited. A brand the AI mentions in its answer picks up around 38 percent more clicks by measurement, along with better performance on paid advertising. So the game is not beating a competitor in a ranking. It is getting into the sentence the AI writes for your customer.
How to get into AI answers
The good news is that the levers are not magic. AI pulls its answers out of content it understands, which makes a few concrete things true. Structured formatting, meaning headings, bullets, numbered lists and tables, raises the probability of a citation by 28 to 40 percent. Content written as a clear question with a direct answer gets cited more readily than long padding.
Technically, schema.org structured data helps (Product, Offer, Review, AggregateRating) so a machine can tell what is a product, a price and a rating, along with an llms.txt file that tells AI models what matters on the site. Reviews and domain authority carry a lot of weight too. Citations are concentrated: the first five domains collect roughly 38 percent of all citations and the first ten take over half. Credible content and ratings get you into answers repeatedly.
This approach is called GEO, generative engine optimisation, and we take it apart in a separate article on GEO and AEO for online stores. The market for GEO services is growing at a furious pace, from around one billion dollars in 2025 towards an estimated seventeen billion by 2034. Most large companies have started. Most smaller ones have not, and that is exactly where a small store's window sits.
What a small store should actually do about it
You do not need to rebuild the company. It is enough to stop writing product pages purely for Google and start writing them so a machine can quote from them too. In practice: a clear description on every product, specifications, real reviews and answers to common questions right there on the page. Plus structured data and a tidy technical foundation.
The second thing is measurement. If you do not know how many visits and orders reach you from AI sources today, you cannot tell whether the effort is paying. This is where most stores are lost, because classic analytics tends to bury AI traffic among other referring sites.
How Behio handles it
The Behio storefront template is built SEO-first and with AI search in mind: it generates structured data, sends ecommerce events to GA4 out of the box and ships its own llms.txt so models know what matters on the site. You are not assembling that from plugins, it is part of the store.
The second piece is analytics of our own. Behio Analytics shows revenue and orders by traffic source right in the admin, so AI search engines appear as a channel in their own right and you can see whether they genuinely bring orders rather than clicks. And because it measures every visitor rather than only the ones who clicked accept on a cookie banner, these are exact figures rather than an estimate from a sample. Tom, the AI partner, will also help you fill in product descriptions in a question and answer style. To my mind that pairing, content for AI plus measurement of what it produces, matters more than chasing every new acronym in marketing.
The verdict
AI search engines are not a bogeyman, they are a new entry point to a purchase, and in 2026 they arrive in force in the Czech market too. The traffic is still small in volume but according to US data it is markedly better and converts better. Practically, that means you stop writing product pages purely for Google and start writing them so an AI can quote from them: clear answers, structured data, reviews and a tidy foundation. And above all measure how much genuinely arrives from AI. Start now and you catch a window that closes within a year.
Common questions
Adobe reports that 38 percent of consumers have already used generative AI for an online purchase, and over 60 percent say they have tried a tool like ChatGPT for buying. Most are still at the research and recommendation stage rather than final payment. The figures are mainly American, with the Czech wave expected during 2026.
According to US data, yes, roughly three times better, and it is about 4.4 times more valuable. The reason is logical: somebody arriving on an AI recommendation arrives with the decision made. Volume is still small, though, so treat it as a high quality supplement rather than a main channel.
Some organic traffic is falling, because around 60 percent of searches end without a click and an AI overview cuts the click-through rate of the first position. The defence is getting into citations inside AI answers, which brings clicks back. The game moves from being first to being cited.
It is a simple file that tells AI models what matters on your site and where to take information from. It is not mandatory, but it belongs to the basic GEO kit alongside schema.org structured data. In Behio the storefront template ships one.
You need analytics that breaks revenue down by source and does not hide AI sites among other referrals. Behio Analytics shows orders and revenue by source right in the admin and measures every visitor, so these are exact figures rather than an estimate from whoever accepted the cookie banner.
Build a store that never surprises you on the invoice.
Stock, invoicing, B2B, product feeds and an AI teammate, all native and all in one subscription. Try Behio free for 14 days, no card needed.
