The Oatly Problem

Ask an AI what the best oat milk is.

Go ahead. Open ChatGPT, Gemini, Perplexity, whatever you use. Type it in. "What is the best oat milk?"

Oatly will not be first. In most responses, it will not be second. Depending on the engine, it might land third, fourth, fifth. Behind Califia Farms. Behind Chobani. Behind store brands that did not exist when Oatly was teaching an entire category how to matter.

I spent two weeks tracing how AI recommendation engines construct their answers to a simple purchase query about oat milk. The methodology is not complicated: these systems ingest editorial taste tests, aggregate review sources, and Reddit discussions. They weigh independent corroboration. They surface consensus. And then they produce one to three brand recommendations per response.

There is no page two. You are named or you are invisible.

Here is what I found.

Serious Eats tested 14 brands. Oatly was in the field. It did not win. Food and Wine did not include Oatly in its top picks. Simple Green Smoothies gave it to Califia Farms. Sporked mentioned Oatly positively but did not rank it first.

Reddit -- the conversational corpus that AI models lean on heavily for purchase-decision queries -- mentions Oatly constantly, but almost always alongside cheaper alternatives. "Oatly is great but have you tried..."

In every major independent editorial source that AI engines use to construct recommendations, Oatly appears. It is in the corpus. It is never on top.

This matters because of how AI recommendation works at a structural level.

These systems operate on a corroboration threshold. When two or three independent sources agree that a product is the best in a category, that product appears in 55 to 97 percent of AI responses. Below that threshold, appearances become intermittent. You show up sometimes. You disappear sometimes. You have no control over which version of reality the user encounters.

For a system that produces one to three recommendations per query, the difference between "frequently mentioned" and "frequently recommended" is the difference between existing and not existing.

Here is the part that should make every brand leader reading this uncomfortable.

The packaging that reads like a conversation with someone who does not take themselves seriously. The "Wow No Cow" slogan that sounds like it was written at 2am by someone who genuinely finds dairy absurd. The decision to publish a 172-page lawsuit when the Swedish Dairy Association came after them -- and turn it into a brand moment. The coffee-shop-first strategy that earned credibility through baristas before it ever chased grocery shelf space. The entire Post Milk Generation movement.

None of this transfers.

AI cannot read tone. It cannot parse rebellion. It does not know that Oatly's packaging is funny, or that the brand's position on dairy is ideological, or that the act of choosing Oatly used to feel like joining something. AI processes claims, rankings, and editorial consensus.

Oatly's brand archetypes -- Jester, Outlaw, Sage -- are the exact archetypes that generate cultural resonance and zero algorithmic signal. The weird, the defiant, and the knowing. Three things a recommendation engine cannot measure, cannot weight, and will never surface.

There is one exception. And the exception proves the rule.

When the query is specifically about coffee -- best oat milk for lattes, best oat milk for espresso, best barista oat milk -- Oatly's Barista Edition still dominates. It appears first or second in nearly every AI response to a coffee-specific query.

Why? Because in the coffee-specific corpus, the editorial consensus is clear and corroborated. Multiple independent sources agree: Oatly Barista Edition is the best oat milk for coffee. The threshold is met. The recommendation follows.

This is not brand equity at work. This is editorial consensus at work. Oatly Barista Edition wins in AI because it wins in taste tests -- in the specific, narrow context of coffee preparation. The brand's cultural identity has nothing to do with it. The product's functional performance in a specific use case has everything to do with it.

The business numbers confirm what the AI layer reveals.

Oatly's 2025 revenue was $862 million. It was their first profitable year. That sounds like a success story until you look at where the growth is and where it is not.

In the market where Oatly built its cultural mythology -- the Brooklyn coffee shops, the influencer adoption, the Super Bowl ad -- the brand is in retreat. Meanwhile, it is growing in Europe and China. Markets where the brand competes on distribution, product quality, and price -- not on cultural identity.

I keep coming back to what this means beyond Oatly.

Because Oatly is not the only brand that built its edge in culture. It is not the only brand whose competitive advantage lives in voice, in attitude, in the feeling of choosing it. It might be the most visible example of a brand whose cultural equity has near-zero transfer value to AI-mediated discovery -- but it is not the last.

Every brand whose differentiation lives in how it says things rather than what it sells. Every brand that assumed cultural resonance and purchase-decision influence were the same thing.

They were. When humans mediated their own discovery. When you walked into a store and your hands reached for the package that felt like it was made for people like you. When your friend said "you have to try this" and the recommendation carried the full weight of their taste and your trust.

AI mediates differently. It reads the spreadsheet version of your brand. Claims, rankings, specifications, editorial consensus. The version of your brand that survives compression into structured data. Everything else -- the voice, the stance, the vibe, the culture -- is lost in translation. Not because AI is biased against it. Because AI literally cannot see it.

Here is the question I cannot stop thinking about.

If you are a CMO and your brand's primary differentiation is cultural -- voice, attitude, community, the feeling of belonging to something -- what is your AI visibility strategy? Not your SEO strategy. Not your media mix. Your strategy for a world where an increasing share of purchase decisions are mediated by systems that cannot read culture.

Oatly taught a generation of brand builders that personality wins. That being weird and specific and unapologetically yourself is a competitive advantage. And it was. In a world where humans chose. The question is whether it still is in a world where algorithms recommend.