AI Assistants Show Brand Preference Patterns in 37,000-Run Audit
An analysis involving approximately 37,000 production runs across four configurations of AI models and 215 commercial prompts spanning 19 industries indicates that AI assistants, such as ChatGPT and Claude, function as recommendation engines rather than search engines, directly suggesting brands. This evaluation utilizes a reference catalog of 533 brands categorized into five tiers of prominence (from L1 category leaders to L5 regional players) based on external authority lists. Although L1 brands are featured in almost every relevant retrieval, they secure only 25-41% of recommendation slots, suggesting that differentiation is more critical than visibility. The study emphasizes that effective marketing to AI necessitates a focus on positioning, content, and product alignment beyond mere discoverability.
Key facts
- 37,000 production runs audited
- Four model configurations tested
- 215 commercially-framed prompts used
- 19 sectors covered
- 533-brand reference catalog stratified into five tiers
- L1 brands appear in nearly every retrieval but win 25-41% of slots
- L2 challengers have different failure modes
- AI assistants function as recommendation engines, not search engines
Entities
—