For years, growth strategy revolved around brand gravity.
Build authority. Own the narrative. Rank the domain.
If the brand was strong enough, products followed.
AI search breaks that logic.
After analyzing 30 days of large language model (LLM) responses across the leading AI systems – ChatGPT, Claude, Gemini, Perplexity, and Llama—a structural shift becomes impossible to ignore:
LLMs think in products.
Marketers still think in brands.
That mismatch is already reshaping how consumers discover, evaluate, and decide what to buy, and most companies are not prepared for it.
AI Isn’t “Assisting” Discovery. It Is Discovery.
This past season, AI-assisted purchase behavior grew by 805%. That number isn’t interesting because it’s large. It’s interesting because it reflects a change in where decisions are being formed.
Consumers aren’t just using AI to research what they already plan to buy.
They’re asking AI what they should buy.
And when they do, the answers look nothing like search results.
Search retrieves.
AI decides.
What AI Sees When It Looks at Your Category
To understand how AI evaluates products, we tracked repeated prompts across models in a single category:
- Best skate shoes for beginners in 2025
- Top rated skate shoes for durability and comfort
- Top skate shoes for streetwear style
Three intents:
- Informational
- Functional
- Identity-driven
Across all three, the result was consistent:
Products appeared first. Brands appeared second – if at all.
Top Product Mentions Across LLMs (30 Days)
| Product | Brand | Mentions |
|---|---|---|
| Old Skool | Vans | 98 |
| Marana | Etnies | 58 |
| Busenitz | Adidas | 50 |
| SB Dunk Low | Nike | 44 |
| Superstar | Adidas | 42 |
These aren’t new releases.
They aren’t brand campaigns.
They’re specific products with cultural memory.
What’s missing is just as important as what’s present:
- Brand positioning
- Brand storytelling
- Brand preference
AI doesn’t reward what brands say about themselves.
It reflects what the world has said about their products.
Brand Presence vs Brand Visibility: The AI KPI Most Teams Miss
AI introduces a new performance split most leadership teams don’t track yet:
- Presence = how often a brand is mentioned
- Visibility = how often it’s recommended or ranked at the top
Those are not the same thing.
In our dataset:
- Nike and Adidas showed high presence but near-zero visibility
- Etnies dominated visibility almost entirely through one product: the Marana
A brand can be everywhere—and still not be chosen.
This is the first real fracture between traditional SEO logic and AI discovery logic.
In search, authority wins.
In AI, relevance wins—and relevance is product-specific.
Here is the 30‑day snapshot of brand performance:
| Brand | Presence | Δ Presence | Visibility | Δ Visibility | #1 Counts | Top Model |
|---|---|---|---|---|---|---|
| Vans | 66.7% | +32.7% | 65.4% | -0.97% | 5 | Llama4 |
| Etnies | 34.0% | +0.8% | 66.4% | +2.9% | 3 | Perplexity |
| Nike | 66.7% | +32.7% | 3.3% | -63.0% | 2 | Llama4 |
| Adidas | 66.7% | +32.7% | 3.6% | -62.7% | 2 | Llama4 |
Why LLMs Default to Products, Not Brands
Three forces drive this behavior.
1. AI compresses choice into a decision
Search offers options.
AI offers conclusions.
When a system must recommend “the best,” it doesn’t reach for an abstract brand promise. It reaches for a concrete artifact.
The product becomes the atomic unit of trust.
2. Training data favors community memory, not brand messaging
LLMs are trained heavily on:
- Reviews
- Forums
- YouTube
- Subreddits
- Wikis
- Cultural discourse
These aren’t brand-controlled environments.
That’s why:
- Old Skools outperform newer Vans tech models
- Busenitz shoes beat newer Adidas lines
- SB Dunks surface for style, not performance
- The Marana dominates durability because of years of outsole lore
Your brand narrative only exists in AI if the community has reinforced it—through the product.
3. Each model has its own bias map
Different LLMs consistently favor different products:
- Perplexity heavily favors Marana for durability
- Llama elevates Vans across most prompts
- Claude leans into cultural relevance
- GPT-4 is balanced, but still defaults to legacy classics
There is no single “AI algorithm” to optimize for.
AI discovery is now model-based market share.
The Strategic Implication: Your Hero Product Is Your Brand in AI
Inside an LLM, brands don’t exist as identities.
They exist as examples.
For each brand in our dataset, one product defined everything:
- Vans → Old Skool
- Etnies → Marana
- Nike → SB Dunk Low
- Adidas → Busenitz / Superstar
If that product performs well, the brand appears credible.
If it doesn’t, the brand becomes invisible—regardless of market power.
This is the most important shift CMOs need to internalize:
Your best product doesn’t support your brand.
It becomes your brand inside AI systems.
What Leadership Teams Should Act on Now
1. Product pages matter more than brand pages
LLMs pull from specificity:
- Attributes
- Use cases
- Comparisons
- Cultural context
Generic category pages don’t survive synthesis.
2. Identify-and protect-your hero product
Every brand has one product AI trusts more than the rest.
If you don’t know what yours is, AI will decide it for you.
3. Presence without visibility is a hidden revenue leak
Being mentioned without being recommended is worse than not being mentioned at all. It creates false confidence.
4. Model-level analytics are no longer optional
Just as teams segment by Google vs Amazon, they now need visibility by:
- GPT
- Claude
- Perplexity
- Gemini
- Llama
Each rewards different signals.
5. AI discovery is already shaping conversion paths
An 805% shift doesn’t signal experimentation.
It signals restructuring.
Brands that wait for attribution clarity will arrive too late.
The Bottom Line
Consumers aren’t browsing anymore.
They’re asking.
And the systems answering them don’t care about brand equity, taglines, or campaigns.
They care about:
- Products
- Proof
- Cultural memory
- Repeated consensus
AI favors products over brands.
Narratives over names.
Utility over identity.
The companies that understand this now won’t just adapt to AI discovery.
They’ll own it.
More Data on LLM Search Visibility: