Amazon’s AI Recommender Ignores Your Organic Rank

A single analyst’s read of Alexa for Shopping suggests a third discovery shelf is forming on Amazon, one that does not consult the rankings sellers have spent years earning.

For a decade, Amazon selling has run on one working assumption: rank high in organic search, and you get found. A new analysis from Marketplace Pulse, an independent analyst outlet, argues that a growing slice of Amazon discovery no longer plays by that rule. Marketplace Pulse’s own study of Alexa for Shopping, the AI-driven recommendation layer Amazon has been building into its search and voice experiences, found that most of what it surfaces has nothing to do with where a product sits in the organic results. That is a single-source finding, not a fact Amazon has confirmed, and it should be read for what it is: one analyst’s reading of one study. But the shape of the finding is specific enough, and the sample large enough, that European and global sellers who treat SEO as the whole game should at least know it exists.

What the study measured

Marketplace Pulse built its analysis around 1,963 non-branded search queries and captured 12,810 Alexa for Shopping recommendations against them, gathered in May and June. Non-branded queries matter here: these are searches like “wireless earbuds” or “kids raincoat,” not “Sony earbuds,” so the results reflect open competition rather than a shopper hunting for a specific brand they already trust.

Against that dataset, the outlet reports that 63.9% of Alexa for Shopping’s picks fell outside the organic top 10 for the matching search term. More strikingly, 40.9% never appeared on the visible results page at all, meaning the product Alexa recommended was not one a shopper would have seen by scrolling the normal search results, no matter how far they scrolled. Put plainly: four in ten of the AI’s picks come from products the shopper’s own search would not have surfaced.

If a company selling in the UK, Germany or France has built its Amazon strategy entirely around climbing the organic top 10, this analysis suggests that ladder no longer reaches every shelf shoppers browse.

The choice to study non-branded queries specifically is worth sitting with. A branded search, someone typing the exact product or company name, is a low-risk moment for any recommendation engine: the shopper has already decided, and surfacing an alternative risks looking wrong. A non-branded search is the opposite: the shopper is still comparing, still open to being shown something new, and that is precisely the moment a recommendation layer has the most room to substitute its own judgment for the ranking algorithm’s. Marketplace Pulse chose the exact scenario where an AI recommender has the freedom to diverge from organic order, and found that it does, in nearly two out of every three cases studied.

Watch: the breakdown on The E-Commerce Weekly, Ep. 3

We walk through the key takeaways in this week’s episode. Full written analysis continues below.

Why this is not paid placement wearing a disguise

The obvious suspicion is that “AI recommendation” is just a rebrand for sponsored inventory, a way to sell placement under a friendlier name. Marketplace Pulse’s own numbers argue against that reading. Only 14.3% of the recommendations it captured were products running a sponsored listing on that same search page. And of that already-small sponsored slice, 83% had already earned an organic ranking anyway, meaning the sponsorship was topping up a product that was already competitive, not buying its way past one that was not.

So the pattern is not “pay Amazon and skip the queue.” It looks more like a separate recommendation logic sitting alongside organic and paid results: a third shelf, in Marketplace Pulse’s framing, that a shopper encounters without asking for it and that answers to signals sellers cannot see or bid on directly.

What it would mean for a European seller, if it holds up

We say “if it holds up” deliberately, because this is one analyst’s study of one Amazon surface, not a disclosure from Amazon itself and not yet corroborated by a second source. Read cautiously, though, the implication for operators is not small. A seller in Warsaw or Lyon who has spent two years optimizing listing copy, backend keywords and PPC bids to hold a top-10 organic spot has been optimizing for a version of Amazon discovery that, per this analysis, covers well under half of what one AI layer actually shows shoppers.

That does not make organic rank worthless. It remains the one lever sellers can measure, test and directly influence. But it reframes what “winning search” means on Amazon. If a growing share of product discovery runs through a recommendation layer that, according to this study, ignores the visible results page four times in ten, then a listing’s fate increasingly depends on signals outside a seller’s console: purchase history patterns, return rates, catalogue completeness, review substance, delivery reliability, the kind of behavioural data Amazon holds and a seller does not.

For sellers running across multiple EU marketplaces, that is a familiar discomfort in a new location. The same asymmetry that shows up in Amazon’s Buy Box logic or its A9/A10 organic algorithm now appears to extend into a layer that sits above both. The practical takeaway from Marketplace Pulse’s numbers is not a new tactic to chase. It is a reason to stop assuming that organic-rank optimization is a complete Amazon strategy, and to watch, rather than act on, what a single analyst has flagged as an emerging pattern.

It also reframes how a team should read its own dashboards. A catalogue manager watching organic rank climb steadily for a hero SKU has, by every existing measure, done the job correctly. This analysis suggests that climb may say less than it used to about whether shoppers actually encounter the product, because a growing discovery surface sits outside the metric being watched. That is not a reason to abandon rank tracking. It is a reason to hold it more loosely, and to keep asking what else, beyond the search page, might now be doing the introducing.

A caveat worth repeating

Everything above rests on a single source: Marketplace Pulse’s own analysis, published without a second outlet corroborating the methodology or the numbers independently. Marketplace Pulse is a respected independent analyst voice on Amazon’s ecosystem, and its dataset (nearly 13,000 recommendations across close to 2,000 queries) is not small. But it remains one organization’s read of how Alexa for Shopping behaves, not something Amazon has confirmed, disclosed or explained. We are presenting it here as exactly that: an analyst’s reading of a study, not a confirmed description of how Amazon’s systems work. Sellers should treat the specific percentages as a signal worth tracking, not as an audited fact to build a strategy around this week.

What is harder to dismiss is the direction of the signal. Amazon has spent two years pushing generative and AI-assisted shopping tools, from Rufus to voice search to now Alexa for Shopping, and every one of them sits on top of, not inside, the organic ranking system sellers already know. Whether or not this particular study’s numbers prove exactly right, the underlying question it raises, whether a hard-won organic rank still governs how often a product gets found, is one every Amazon seller in Europe and beyond will need an answer to well before this AI layer matures.