Your webshop’s next customer may no longer be searching on Google
One of the biggest changes in e-commerce is not that yet another marketing channel has appeared. It is that part of the purchase decision is slowly moving out of webshops and traditional search engines.
More and more people no longer open ten different Google results when they look for a product. Instead, they tell an AI assistant what they need:
“Which robot vacuum should I choose under HUF 150,000 if I have a dog and lots of carpets?”
“Which running shoe is good for 5-10 kilometres a day if soft cushioning matters?”
“Find me garden furniture that can stay outside all year round and seats six people.”
At first glance the difference seems minor. In reality, it creates a completely different buying process.
A search engine gives you results.
AI, on the other hand, interprets, narrows down, compares and recommends.
And by the time the user reaches an online store, they may already have made a large part of their decision.
The purchase no longer necessarily starts on the website
For a long time the classic e-commerce model was fairly simple. Someone searched for something on Google, saw an organic result or an ad, clicked on it, browsed the webshop, compared products, then bought or left.
SEO, Google Ads, analytics and, in fact, the whole conversion funnel were built on this.
AI assistants are starting to take this process apart. Getting to know products, finding your way between categories, comparing alternatives and, in many cases, even weighing pros and cons can all happen before the shopper ever visits the retailer’s site.
And this has an interesting consequence. A webshop may get fewer visitors while the visitors who do arrive become more valuable.
Little AI traffic, surprisingly strong purchase intent
For many online stores, direct AI traffic is still a tiny slice of total traffic today. That, however, can easily be misleading.
In Growww Digital’s analysis of data from more than 900 Hungarian online retailers, AI traffic still accounts for a relatively small share on average, but a very different picture emerges at the best-performing webshops.
At the ten webshops with the largest AI share, very little revenue was linked to AI traffic in April 2025. The AI yield index they used was just 0.18 at the time. By May 2026, the same index had risen to 1.54.
In plain terms: at these online stores, traffic coming from AI already accounted for a larger share of revenue than of visits.
This is not just a Hungarian phenomenon. According to Adobe Digital Insights’ US retail data, in March 2026 visitors arriving from AI sources had a 42 percent higher conversion rate than visitors from other sources. A year earlier the situation was exactly the opposite: AI traffic converted 38 percent worse.
| Metric | Before | Now | Source |
|---|---|---|---|
| AI yield index at the ten Hungarian webshops with the largest AI share | 0.18April 2025 | 1.54May 2026 | Growww Digital |
| Conversion rate of visitors arriving from AI compared with other sources (US retail) | 38% lowerMarch 2025 | 42% higherMarch 2026 | Adobe Digital Insights |
That is a turnaround in a very short time. And there is a logical explanation for it.
AI is becoming a kind of pre-filter
If someone arrives at a webshop after a traditional search, they may only now be starting to understand what kind of product they need. An AI assistant, however, may already have asked them:
- What will you use it for?
- What is your budget?
- What matters more, price or quality?
- What problem do you want to solve?
- Is there a feature you definitely want?
That is why a user sent on by AI is, in many cases, not a simple visitor. They are already a prepared buyer.
In the same research, Adobe measured that users arriving from AI spend more time on websites and view more pages than those coming from non-AI sources.
This is no longer just a traffic question. This is purchase intent.
SEO is not dying. The meaning of a “result” is changing
Over the past twenty years, SEO has been declared dead roughly as many times as someone on the internet discovered a new acronym. This is not about that either.
Traditional search engines still deliver huge traffic, and organic search presence still matters. But another competition has appeared next to it.
You no longer have to fight only for this:
“My site should be among the top results.”
But also for this:
“AI should treat my brand, my product or my content as a relevant source when it answers.”
This is where terms like AEO, GEO or AI Search Optimization come from. The names change, but the business problem behind them is the same:
does an artificial intelligence understand what your company offers, to whom, at what price, with what features, and why it may be relevant to a given buyer?
If not, your webshop can be as beautiful as you like. To the machine, it will simply be a less useful source of information.
The next big KPI may not be traffic at all
This is perhaps where analytics will get the biggest headache. Traditional digital marketing likes counting clicks.
- Impression.
- Click.
- Session.
- Add to cart.
- Purchase.
A neat little funnel. Everything has its own box, and for a few years humanity believed it had managed to turn shopping into mathematics.
AI tangles this up too. Suppose someone asks an AI:
“Which three brands should I look at if I’m after a premium coffee machine?”
The AI lists three brands. The user doesn’t click any of the links. The next day they search for one of the brand names on Google, then buy.
Classic analytics can easily see this as a Google organic or brand search conversion. Yet the moment of choosing the brand actually happened earlier, in the AI conversation.
That is why so-called AI visibility is becoming more and more important: examining in which topics, questions and buying situations a brand appears in the answers of generative systems. INTREN, for example, already treats AI appearances, AI referral traffic and the resulting conversions separately.
The measurement chain of the future may therefore look more like this:
- AI appearance
- brand mention
- click or later search
- webshop
- conversion
And even that may not be enough for long.
What happens when the shopper no longer even has to enter the webshop?
The next level of AI commerce is so-called agentic commerce. Here AI doesn’t just help choose the product. It takes part in the transaction itself.
The user can state what they are looking for and under what conditions they want to buy it, and the AI may be able to search for products, compare offers and handle an ever larger part of the buying process.
In this world the webshop may remain a key back-end system, while the shopper sees less and less of it. The retailer’s database, stock, product feed, prices, shipping information and structured product data will communicate with other systems.
And all the shopper may notice of this is:
“I recommend this. It can be delivered tomorrow. This is what it costs.”
This fundamentally changes what a “good webshop” will mean.
A beautiful product page is no longer enough
In the coming years, an online store has to be understandable to two audiences at once. One is the human. The other is the machine.
The human still needs good design, clear product descriptions, trust, good photos, a fast site and a simple checkout.
The machine, however, needs other information as well:
- A clear product structure.
- Accurate attributes.
- Consistent naming.
- Structured data.
- Reliable price and stock information.
- A well-built product feed.
- Content that makes it clear what problem the product solves.
- And an external digital presence that makes it possible to verify that the brand really exists, is relevant and is trustworthy.
This technical background is becoming especially important because the same data structure can be used in several places: in classic SEO, in shopping systems, in AI-based search and later in automated commerce systems too.
Meanwhile, a new advertising surface has appeared
AI search and AI advertising are two separate fields. A brand that features well in an AI answer has not paid to get in. And by advertising, a brand does not buy the generated answer itself. That is an important difference.
In August 2026, OpenAI expanded ChatGPT Ads to 31 European markets, and from August 31 the self-serve Ads Manager also became available in these markets.
The model is fundamentally different from classic keyword-based search advertising. The user doesn’t necessarily type in two or three keywords. They can describe their whole problem, goal, budget and preferences. This gives the ad system a completely different context.
At the same time, ChatGPT ads are kept separate from the model’s answer: according to OpenAI, advertisers cannot modify or rank ChatGPT’s answers.
This is why two parallel competitions are starting to take shape.
Organic
does the brand get into the sources and recommendations the AI considers relevant?
Paid
can it appear at the right moment next to a conversation that shows commercial intent?
Just as today we handle SEO and Google Ads separately. Only now their AI-era counterparts are being born.
What does all this mean for a webshop in 2026?
Not that you have to switch off your Google Ads campaigns tomorrow. Not that SEO has become unnecessary. And not that the whole marketing budget has to be pushed into AI.
For most retailers, the direct share of AI traffic is still much smaller than that of the big, established channels. According to Growww Digital’s data on more than 900 Hungarian webshops, average direct AI traffic remains low, while some segments and players already stand out significantly.
So the current share of traffic is not what really shows the significance of the change. The direction is.
- Search is slowly turning into conversation.
- The results list into recommendation.
- The keyword into intent.
- And the click into just one possible stop.
That is why five things are already worth putting in order:
- Measure AI traffic and conversions separately.It easily gets lost in total traffic, while commercially it can be more valuable than average.
- Check what AI systems say about your brand.It is not enough to measure whether you get clicks. You also need to know whether you make it into the answers at all.
- Make your product data easy for machines to understand as well.Product feeds, structured data, attributes and accurate stock information are becoming strategic infrastructure.
- Don’t pit SEO against AI optimization.Well-structured, credible and informative content is valuable in both systems.
- Handle AI visibility and AI advertising separately.In one, you are trying to be used as a relevant source. In the other, you are buying media placement.
The question is not whether Google will disappear
We are probably asking the wrong question when we argue about whether AI will “replace” search engines. Shoppers are fairly indifferent to what the marketing industry calls the channel they happen to be using. They simply choose the tool that gets them to a good decision fastest.
And if an artificial intelligence can narrow the choice down to three relevant products in thirty seconds instead of five minutes of searching, people will use it.
For retailers, the next competition is therefore no longer only about who ranks higher in the results list. It is also about who is recognized, understood and recommended by a system that plays an ever bigger role right before the purchase decision.
And that is a much bigger change than the arrival of yet another marketing channel.
Frequently asked questions
Why is a visitor arriving from AI more valuable?
Because the AI assistant may already have asked them before the purchase, for example what they will use the product for, what their budget is, and what matters more, price or quality. According to Adobe Digital Insights’ US retail data, in March 2026 visitors arriving from AI sources had a 42 percent higher conversion rate than visitors from other sources.
How big is AI traffic at Hungarian webshops today?
Still small on average. According to Growww Digital’s analysis of data from more than 900 Hungarian online retailers, however, at the ten webshops with the largest AI share the AI yield index rose from 0.18 to 1.54 between April 2025 and May 2026, which means AI traffic there already accounted for a larger share of revenue than of visits.
What are AEO and GEO?
Names for optimizing for AI search: Answer Engine Optimization, Generative Engine Optimization, also called AI Search Optimization. Behind them is the same business question: does an artificial intelligence understand what your company offers, to whom, at what price, with what features, and why it may be relevant to a given buyer?
Can I pay for ChatGPT to recommend my brand?
No. According to OpenAI, ChatGPT ads are kept separate from the model’s answer, and advertisers cannot modify or rank the answers. AI visibility and AI advertising are therefore two separate competitions: in one you have to get into the answers as a relevant source, in the other you buy media placement.
What should a webshop put in order now?
Five things: measure AI traffic and conversions separately, check what AI systems say about your brand, make your product data easy for machines to understand, don’t pit SEO against AI optimization, and handle AI visibility and AI advertising separately.
Sources
- Growww Digital: Many AI visitors, little money: where is AI revenue actually generated? (in Hungarian), August 13, 2026
- Growww Digital: AI is turning e-commerce upside down (in Hungarian), May 13, 2026
- Adobe Digital Insights: Adobe report: U.S. retailers see surge in AI traffic, but many websites are not entirely readable by machines, April 16, 2026
- INTREN: The impact of Google AI Overviews and AI Mode on SEO (in Hungarian), September 21, 2026
- OpenAI: ChatGPT Ads expands across Europe, August 18, 2026, updated August 31
- OpenAI Help Center: Ads in ChatGPT, September 2026
A note on the numbers: the AI yield index is AI’s share of revenue divided by AI’s share of visits. Growww Digital itself points out that part of the gap between the best webshops and the rest may be a measurement difference: at other webshops, traffic from AI may show up in GA4 as direct or unassigned traffic.
