I Analyzed 8,000 AI Product Recommendations. The Results Were Not What I Expected.

Over the last few months, I've been researching a question that I couldn't stop thinking about.

Everyone is talking about AI visibility.

Can AI find your business? Can AI mention your brand? Can AI cite your website?

But I started wondering if we were focusing on the wrong question.

If AI is recommending products to consumers, does it actually recommend the businesses with the best e-commerce stores?

I decided to test it.

The experiment: I analyzed two e-commerce categories.

Beauty and Supplements.

For each category, I generated 4,000 AI recommendations using high-intent shopping prompts.

Questions like: Best vitamin C serum, Best moisturizer for oily skin, Best protein powder and Best magnesium supplement

Each prompt was run repeatedly to capture recommendation patterns rather than single responses.

In total, the dataset included:

8,000 AI recommendations
609 brands
40 shopping prompts
Hundreds of real e-commerce stores

For every brand, I measured how often it was recommended and compared that against the quality of the store behind it.

My assumption was simple.

If a store is easier for AI systems to understand, it should be recommended more often.

The data told a different story.

What I expected:
I expected strong stores to win. Stores with clear structure. Strong product information. Good machine readability. Better signals for AI systems.
The logic seemed straightforward. Better stores should earn more recommendations.

What actually happened...

In Beauty, the relationship between recommendation frequency and store quality was almost nonexistent.

Correlation: r = 0.17**

In Supplements, it disappeared entirely.

Correlation: r = -0.015**

Two independent categories.

The same outcome.

Brands that were recommended most often were not necessarily the brands with the strongest stores.

The brands that kept appearing

Certain names showed up repeatedly: Clinique. SkinCeuticals. Kiehl's. NOW Foods. Nature's Way.

Many of these brands appeared frequently in recommendations despite having relatively weak AI Commerce Scores™.

At the same time, some of the strongest stores in the dataset appeared only rarely.

That forced me to reconsider the original assumption.

A different explanation:

The more I looked at the data, the more one pattern stood out.

AI systems appear to recommend many brands because they already know them.

Not because they have the best stores.
Not because they are the easiest to understand.

Simply because they are familiar.

The brands have been mentioned across the internet for years.
They have strong awareness.
They have been seen repeatedly during model training.

My working hypothesis is that much of today's AI shopping behavior is driven by what I call Recommendation by Memory™.

The model reaches for brands it already knows.

Why this matters:

Most of the conversation today focuses on visibility.
Visibility is important.
But visibility and recommendation are not the same thing.

A brand can be visible without being recommended.
A brand can be recommended without having a strong store.

Understanding that difference may become increasingly important as AI becomes a larger part of e-commerce discovery.

The question I am becoming most interested in is no longer can AI see my business?

The question is why does AI choose my business instead of someone else's?

What happens next:

I'm currently analyzing three additional categories:

Coffee
Pets
Home & Living

If the same pattern appears again, it may suggest that recommendation behavior is influenced far more by brand familiarity than most people currently assume.

Either way, the data is already challenging some of my own assumptions.

And those are usually the most interesting discoveries.

Have you seen similar patterns while working with AI search, e-commerce, or recommendation systems?