Amazon’s artificial intelligence does not look at the rankings (and for sellers that is very interesting!)

For years, anyone selling on Amazon has learned a simple rule. If you want to be seen, you have to climb its ranking or pay to appear among the sponsored ads. Two levers, two strategies, with a single goal: to occupy the top positions on the search page.

But what happens when the person looking for a product no longer scrolls through a list of results, and instead asks an AI assistant directly “what is the best double mattress“?

That is the question asked by a recent study carried out by Autopilotbrand, which analysed almost 2,000 generic searches and more than 12,000 recommendations generated by Alexa for Shopping, Amazon’s assistant.

The results call into question many established certainties about e-commerce.

There is one figure it is best to know

The most interesting finding of the whole study is that almost two thirds of the assistant’s recommendations (63.9%) concern products that do not even appear among the first ten organic results of the corresponding search. And that is not all. More than 40% of the recommended products are not even visible on the standard search page. In other words, you would have to scroll far further down than most users ever would.

Even more relevant is the figure on advertising. Only 14.3% of the recommended products had an active sponsored listing on that search. And the vast majority of these (83%) would have appeared organically anyway, with no need for sponsorship. In other words, paying for visibility in search does not seem to buy, at least for now, visibility inside the AI assistant.

The data naturally needs to be corroborated by further studies, but we are probably looking at a shift in approach of considerable importance. Until recently, in fact, AI shopping assistants limited themselves to returning links to searches the user could have run on their own. In other words, a more “conversational” version of the same results page.

The study suggests instead that this phase should be considered over. The assistant is no longer repeating the search ranking in different words, but is selecting products according to its own criteria, often fishing deep into the catalogue, where very few human buyers would ever scroll.

Why the search ranking counts less and less, at least in this case

On Amazon’s traditional search page there are two ways to get noticed. The first is the organic position, earned over time thanks above all to sales volume and reviews. The second is the sponsored position, bought at auction with the advertising budget. These are the two metrics on which marketing teams build their dashboards and around which much of the advertising spend revolves.

The study suggests, however, that at least for now neither of the two levers directly determines what the AI recommends. And the most surprising finding concerns precisely the ranking. It is the metric sellers invest in most, and yet it seems to be the one the assistant most clearly ignores.

A word of caution, though. As we have already partly noted above, this is a single study, carried out on one account (USA), at a still early stage of adoption of these tools. It is therefore certainly not definitive proof of how the AI’s selection will work. Better to read it as a first concrete signal, measured with real data, that deserves the attention of anyone selling online.

We are witnessing the birth of a third shelf

As Christian Umbach, co-founder and CEO of Autopilotbrand, observed, a third visibility channel is taking shape alongside organic search and paid ads. A channel that, unlike the other two, cannot be bought or scaled simply by investing in advertising or piling up sales.

To emerge in this new space, brands have to provide the AI assistant with high-quality information. In other words, they have to guarantee:

• product listings rich in data
• content built around users’ real search intentions
• constant updates that take seasonality and the product’s distinguishing features into account.

It is a different job from the one required by traditional Amazon SEO. In some ways, it is closer to a logic of understanding context than of simple ranking.

There is good news, though, because the effort should pay off. The third shelf is in fact a concrete opportunity, especially for less structured brands. A good product held back by competition on ranking can still win the assistant’s recommendation, if its catalogue communicates clearly why it is the right choice for that specific need.

What it means in practice for our current and potential customers

If the AI selects products on its own criteria, the question every seller should be asking is no longer just “how do I climb the ranking“, but “how do I make my product understandable and convincing for a system that has to decide, on its own, whether to recommend it or not“.

We believe some elements will become increasingly important:

• Richer, more structured catalogue data. Titles, bullet points, descriptions and complete technical specifications no longer serve only to convince the human eye scrolling the page. They also serve to provide quality “raw material” to a system that has to assess how suitable the product is for a need expressed in natural language.

• Content built around intent, not just around the keyword. Someone asking “what is the best mattress for people with back pain” is expressing a far more specific need than someone simply typing “double mattress“. Products that clearly answer specific intents are more likely to be selected.

• Continuous updating. Seasonal needs, differentiating features, the most recent reviews. Everything that changes over time has to be reflected in the catalogue. An AI assistant constantly updates its understanding of the product.

• Monitoring your own presence in AI shopping. Just as search positioning is monitored today, it will become increasingly important to understand whether, when and why your product is (or is not) recommended by AI assistants. That way you can act before your competitors do.

An advantage that might not last long

In any case, it is worth placing all of the above in the right context of timing. There is a historical parallel worth keeping in mind. Amazon’s organic search too was, in its early days, a relatively clean environment, where a good product could emerge without necessarily having to compete with enormous advertising budgets. Then, gradually, sponsored ads took up more and more space. And they became an almost compulsory cost for anyone wanting to be visible.

Nothing rules out the same happening, sooner or later, to the AI’s third shelf as well. Amazon has already begun introducing sponsored listings and advertising formats directly inside Alexa for Shopping. For now, according to the study, these mechanisms do not yet seem to decisively influence the assistant’s organic recommendations. But the window in which catalogue quality counts more than the advertising budget may not stay open forever.

In short, the Autopilotbrand study captures what is a moment of transition. AI shopping assistants no longer limit themselves to returning a spoken version of the search page. They select products according to their own logic, largely disconnected from organic ranking and advertising investment. It is a real opportunity for sellers who do not yet hold the first page of search. But it is also an invitation to rethink how product listings are built. Not only to convince a customer scrolling a list, but to be understood and recommended by a system that reasons differently.