A map of similar things
Imagine a map where “puppy” sits near “young dog,” while “raincoat” is farther away. An AI model turns each item into a list of numbers that acts like its position on this map. That list is an embedding.
Software compares these number lists to find related documents for AI answers or suggest similar products.
Similar does not mean correct
The model can miss context, so nearby items are not necessarily interchangeable, accurate, or useful. Results also vary by model and language. Embeddings can retain clues about sensitive source data, so apps need to protect them.