Can models see product nuance?
Genesis 01 tests whether vision systems understand brand, season, similarity, and uncertainty.
We study how models see products, form preferences, route shoppers, and turn recommendations into commerce outcomes.
AI shopping is not just a new interface. It is a recommendation layer with taste, shortcuts, and commercial consequences. Caeliai studies that layer so brands can understand what is happening before the market treats it as obvious.
Genesis 01 tests whether vision systems understand brand, season, similarity, and uncertainty.
Genesis 02 maps repeated model taste: the designers, objects, and narratives systems converge on.
Genesis 03 follows recommendations into official pages, third-party leaks, and missing purchase paths.
Genesis is the research layer behind Caeliai. Each study looks at a different part of machine-mediated discovery: what models perceive, what they prefer, and where their recommendations send shoppers.
A foundational benchmark comparing CLIP, SigLIP, and DINOv2 across runway images, uncertainty detection, and collection coherence.
A multi-model look at aesthetic consensus, canonical fashion objects, and how preference clusters emerge across current AI systems.
A field study of ChatGPT and Gemini shopping conversations, measuring official wins, third-party leaks, and no-PDP losses after a brand is recommended.
The advisory page turns this research into a practical score for ecommerce brands. The proof is not that Caeliai sells scoring. The proof is that Caeliai has already done the research.
Recommendation systems can be observed, tested, and compared.
AI answers create new forms of shelf space before search clicks happen.
Recommendations only matter commercially if they lead shoppers to usable buying paths.