AI Models Favor Neon and Upstash Over Incumbents in Database Recommendations

A weekly experiment tested how four major AI models — ChatGPT, Claude, Gemini, and Perplexity — respond to common database-related queries a real buyer might ask. Neon, a serverless Postgres provider, was recommended first in 14 of 20 conversations, while Upstash and Turso also outranked traditional incumbents in their respective sub-categories. However, in areas like vector search, object storage, and real-time analytics, established names such as Pinecone, Amazon S3, and ClickHouse dominated, leaving newer entrants like Tigris and Tinybird virtually invisible. The analysis concluded that AI recommendations reflect training data rather than product quality, meaning companies absent from independent online writing are unlikely to appear regardless of their technical merits. The findings suggest that visibility in AI outputs depends on content presence at the time of model training, not on feature superiority.