Generative AI refers to models trained to sample from a learned distribution of content — they don't just label a photo as "kitchen", they create new kitchens. The two dominant families today are diffusion models (used for images) and large language models (used for text). Both work by learning a compressed representation of a huge training set and then sampling from it on demand.
In interior design, generative AI is what makes "show me this room in Scandinavian style" possible. The model has internalized millions of design photos and learned what features cluster together — pale woods + neutral textiles + minimal hardware — so when you ask for that style applied to your room, it produces an output that's consistent with the pattern.
Generative AI is not infallible. It can invent details that don't exist in your room, hallucinate brand names, or ignore parts of the prompt. The skill of using it well is constraining the generation: locking the room geometry, providing reference images, narrowing the prompt to specifics the model handles reliably.