Generative AI creates text, images, audio, video, code, or other content by producing likely structures learned during training and conditioned by current input.
A product may add instructions, retrieval, tools, filters, and human review around the model. The generated result is an output, not evidence by itself.
Example
A drafting assistant
Given a topic, audience, and notes, it can generate a new email draft rather than retrieve one stored email unchanged.
Important limit
New does not mean true or original in every sense
Outputs may be inaccurate, derivative, unsafe, or unsupported; legal and authorship questions depend on source, use, and jurisdiction.
Interactive model
Separate before deciding
Selected layerGeneration
Produces a new arrangement or continuation.
Read the complete text alternative
GenerationProduces a new arrangement or continuation.
RetrievalFinds existing material; products may combine both.
Useful distinctions
Separate the questions before judging
Generation
Produces a new arrangement or continuation.
Retrieval
Finds existing material; products may combine both.