Someone asked ChatGPT yesterday "what Romanian online stores do you recommend for baby products". They got three names. Another buyer asked Perplexity "best natural cosmetics stores". They got five. A third, on Google AI Overviews, "furniture stores with fast delivery". They got four.
None of those lists were generated randomly. And none of them change fundamentally if you spend more on Google Ads.
If your store is not on those lists, you are losing sales that get decided before the buyer reaches your site, and the loss does not show up anywhere in Google Analytics.
What changed in buyer behaviour
Two years ago, the typical online purchase journey looked like this: the buyer typed a few keywords into Google, saw ten links, clicked on the first two or three, compared products, and bought from the store that convinced them.
In 2026, the journey increasingly looks like this: the buyer opens ChatGPT, asks a full sentence "what should I buy for [problem]", gets a short list of 3 to 5 recommendations, and buys directly from one of them. They do not see ten links. They do not compare ten stores. They see three or four.
Recent studies show 58% of consumers have already replaced traditional Google searches with AI assistants for at least part of their decisions. Gartner estimates traditional Google search volume will drop 25% by the end of 2026. AI assistants no longer route traffic to 10 sites, they concentrate it on 3.
For an online store, the question "are you ranking on Google" becomes secondary. The real question is: when the buyer asks the AI, are you in the answer?
How AI decides which stores to recommend
AI assistants do not pick recommendations randomly. They build a short list using four kinds of information.
1. Training data. Large models (GPT, Claude, Gemini) were trained on public text from the internet. If your brand appears on Wikipedia, in press articles, in reviews, in "top 10 stores for [category]" lists, the AI already "knows" you before the user asks.
2. Structured data on your site. AI assistants do not read your site like a buyer does. They read the technical form of the catalogue: schema markup, product feed, declared data on the page. If your products have clean schema, declared brand, GTIN, price, availability, the AI can cite you with confidence. If those are missing, it skips you.
3. Real-time retrieval. Some AI assistants (Perplexity, Google AI Overviews, ChatGPT with browsing) actively search the web at the moment of the question. That means you can enter the answer even if your brand was not in the training data, but only if your page is structured to be easy to understand and cite.
4. Brand authority signals. The AI does not recommend brands it does not "know". It needs a clear entity: consistent company name, easy-to-verify online presence, real reviews. Stores with complete Organization schema, an active Google Business Profile, mentions in credible articles, are favoured over stores that have not established that identity.
The four kinds of information work together. Missing any one of them takes you off the short list.
What this means in money, for an e-commerce store
Take a simple numerical example. A natural cosmetics store, 300 monthly transactions, average order value €50, 80% of traffic from Google Ads and organic search.
If the typical buyer spends 10 minutes on ChatGPT before purchasing and gets a list of 4 recommended stores, that means there are 4 candidates in their mind before they visit any site. If your store is on the list, you have a 25% chance of being chosen (1 in 4), without spending anything on a click.
If you are not on the list, your chance drops as close to zero as you allow yourself to imagine. The buyer does not actively search for you. They do not compare. They buy from the list they were given.
The cost of being invisible in AI answers is not a cost visible in Google Analytics, because the sale goes to a competitor without ever generating a click on your site. It is a sale that did not happen, and you will not see it being lost.
Quick test: are you in the recommendations or not?
Before any strategy, run this 10-minute test:
- Open ChatGPT (the free version works).
- Ask in English: "What online stores do you recommend for [your product category]?"
- Note the names that appear.
- Repeat on Perplexity (perplexity.ai) and on Google (with AI Overviews active).
- Repeat the question for 3 different categories in your catalogue.
If your store does not show up in any of the answers, you are already losing buyers. If it shows up only in one, you have partial visibility. You are on the AI's radar for only part of your catalogue.
This test is the starting point of the AI-Ready audit. We run it together for 5 categories of your catalogue and map exactly what is keeping you out of the recommendations: missing data, incomplete schema, unsettled brand authority, or a combination.
What you can do this week
Three steps that cost nothing and give you a clear picture of where you stand:
Step 1. Run the test above for your three most important categories. If you are absent across all three, the problem is structural, not category-specific.
Step 2. Open your site in a browser and check, on a representative product page, whether you have schema markup. The simplest check: paste the URL into the Schema.org Validator. If the tool does not detect a complete Product block, that is one of the causes of your invisibility to AI.
Step 3. Search your brand name on Google and check whether the following appear: a Wikipedia page (even a short one), an active Google Business Profile, recent press articles, real reviews on known platforms. If two of these three are missing, you have a brand authority problem that the AI sees.
These three checks do not fix anything, but they tell you where you stand. The decision about what to fix and in what order comes after you have the data.
Visibility in AI Shopping is not bought with budget. It is earned with product data written correctly and with monthly maintenance discipline.
What we do next
In the next article in this series, we go into the technical details: the 7 concrete reasons the AI does not recommend your products, how to check each one, what tools to use, what to fix in order. It is the checklist you want next to you when you open your site's schema markup.
The third article zooms out: what buyers will do through AI agents in 2026, how the economics of customer acquisition shift when the agent buys in their place, and what e-commerce stores should do in the next 12 months.
If the test above confirmed you are losing AI recommendations and you want to skip the two years of individual experiments, the AI-Ready Catalog Audit gives you a complete map of the problems and the repair order in 5 days.
Your category already has a winner in AI answers. Every month a competitor stays in the recommendations and you do not, their recommendation becomes the default. At some point, recovery is no longer fast, no matter what you spend on Google Ads.
Your competitor is already in ChatGPT answers for your category. You are not.
The AI-Ready Catalog Audit maps in 5 days exactly what is keeping you out of AI recommendations and the order of fixes. 5 categories manually tested, schema markup verified, brand authority evaluated.
Adela Mincea
Marketing Economist · Fondatoare DAFE Digital · Formator ANC
Adela is a Marketing Economist with over 10 years of paid media experience across Europe, the US and Asia. She founded DAFE Digital for one reason: serious Romanian businesses deserve the same paid media expertise companies get in any other market. That's what DAFE Digital does.
Tags
