Collection: Our Blends

Traditional blending is an art guided by experience, intuition, and taste. At its best, it produces extraordinary results. But intuition has limits. A skilled blender can consciously hold perhaps a dozen variables in mind. A large language model trained on the intersection of botanical pharmacology, sensory science, and consumer research can hold millions.

Our approach was to use AI to do what it does best: analyse at scale. We fed our models with peer-reviewed research on botanical compounds, traditional medicine texts from five cultures, flavour pairing data, and blend ratings. We then ran iterative simulations asking: for a target cognitive state (deep focus, neural recovery, metabolic activation, creative uplift), which ingredient profiles come closest? What is the mechanistic explanation for why they work? And how should they be positioned and communicated?

The blends in ChatGPTea v1.0 are not synthetic — they are real teas from real plants. What AI contributed was the analytical framework: the why behind each selection, the precise mapping between compound profiles and target states, and the honest claim that these are not just nice teas — they are the right teas, for the right moment, backed by the best evidence we have.