Daniel Pokorný

Posted on Jun 17 • Edited on Jun 19 • Originally published at atomfoundry.dev

The State of AI Recommendations in Pets

#ai #ecommerce #machinelearning #marketing

Recommendation Intelligence Research™ · Atom Foundry · June 2026

We asked one AI model 20 high intent pet shopping questions, 20 times each.

Then we checked the same thing we checked in beauty, supplements, and coffee.

Does being recommended by AI have anything to do with how AI ready a store actually is?

Pets produced the strongest answer yet.

Across another 4,000 recommendations, recommendation frequency was not positively related to store readiness.

In fact, it moved in the opposite direction.

Methodology

Same method as the first three reports, so all four categories are directly comparable.

Everything below is computed from real captured responses.

Nothing is estimated or projected.

Category: Pets

Model: one model (gpt-4o-mini)

Intents: 20 high intent shopping prompts

Runs per intent: 20

Prompt runs total: 400

Recommendations captured: 4,000

Distinct brands: 405

Three metrics carry the analysis.

Recommendation Share™ is a brand's share of all recommendations captured. The field sums to 100 percent.

Recommendation Frequency™ is the percent of the 400 prompt runs in which a brand appeared at least once.

Recommendation Position™ is the average rank in the answer when the brand appeared. Lower is better.

Marketplaces are not brands.

Retailers and marketplaces such as Chewy, Petco, Amazon, PetSmart, and Frisco were separated out and excluded from the brand level analysis.

The contest measured here is between single brands and their own stores.

A note on pets.

Several of the largest pet food brands including Purina Pro Plan, Royal Canin, Hill's Science Diet, and Taste of the Wild are not in our store index, so they appear as off index and are excluded from the correlation analysis.

One brand, Wellness, was removed from the on index set after we found its name had auto matched to an unrelated domain. Rather than attach an incorrect store, we classified it as off index.

Other limitations remain the same.

One model.

One category.

One point in time.

Brands map only to stores we actually measure.

Anything else stays off index.

The Pet Recommendation Leaderboard

Top brands by share of voice across all 4,000 recommendations, with retailers excluded.

The AI Commerce Score™ column reflects each brand's actual score from our index where a measured store exists.

Pet category results showing Recommendation Share™, Recommendation Frequency™, Recommendation Position™, and AI Commerce Score™ for leading AI recommended brands.

The Central Finding

If AI recommended the stores that are easiest for AI systems to read, recommendation frequency and AI Commerce Score™ would move together.

They do not.

In pets, they move slightly in the opposite direction.

Across the on index single brand stores, the correlation between Recommendation Frequency™ and AI Commerce Score™ is r = -0.366 (n = 39).

That is the first category where the relationship is meaningfully negative.

The brands that appear most often in AI recommendations tend to have weaker stores.

Blue Buffalo appears in 27.8 percent of prompts while scoring only 30.

Merrick appears in 24.8 percent of prompts while scoring 42.

Purina Pro Plan appears in 26.8 percent of prompts despite not even being represented in our store index.

Meanwhile some of the strongest stores in the category barely appear at all.

**Benebone scores 79.

Casper scores 73.

Pawstruck scores 72.**

Yet none of them approach the recommendation frequency of the legacy pet food brands.

This is the clearest inversion we have measured so far.

The brands AI remembers are not the brands with the most AI ready stores.

The average AI Commerce Score™ across all on index recommended brands is just 52.7.

We have now measured four categories.

Beauty: r = 0.17
Supplements: r = -0.015
Coffee: r = 0.019
Pets: r = -0.366

Across more than 16,000 recommendations, recommendation frequency is never positively associated with store readiness.

In three categories the relationship is effectively zero.

In pets it becomes negative.

Brands, Not Retailers

Pets is also the most marketplace driven category we have measured.

Retailers and marketplaces such as Chewy, Petco, and Amazon accounted for 7.1 percent of all recommendations.

That is significantly higher than beauty, supplements, or coffee.

Only 19.8 percent of recommendations map to a single brand store that we actually measure.

That is the lowest share we have seen.

Two factors drive that result.

First, pet commerce is heavily concentrated around large retailers and marketplaces.

Second, some of the most recommended brands in the category are not represented in our store universe.

Purina
Royal Canin
Hill's Science Diet
Taste of the Wild

These brands are recommended because of brand familiarity, not because an AI system evaluated their stores.

That is itself an important finding.

What It Means

Visibility is not recommendation.

Recommendation is not readiness.

Today AI still recommends largely from memory.

It reaches for names that appeared most often in training data and public discussion.

That is why the dominant pet food brands continue to win recommendations even when their stores are weak, absent, or difficult for AI systems to evaluate.

But that advantage is temporary.

As AI shopping evolves toward retrieval, browsing agents, comparison engines, and autonomous purchasing, recommendation decisions will increasingly depend on what agents can actually read, understand, verify, and trust.

The brands relying on historical recognition have the most to lose.

The brands building readable, trustworthy, machine legible stores today are the ones most likely to keep the recommendation when memory is no longer enough.

That is exactly the gap the Recommendation Intelligence Framework™ was built to measure.