Definition
AI that creates new content (text, images, video, code) rather than just classifying or predicting. Powers writing tools, image generators, AI agents, and most consumer AI applications. The output is produced fresh each time from patterns the model learned, which is why two runs of the same request rarely come back word for word identical.
What it looks like in practice
A landscaping company wants a spring aeration offer out to last year's customers. The owner gives the AI three inputs: the service, the price band she is comfortable with, and the fact that most of these customers booked mowing but never lawn treatment. Back comes a subject line, a short email, and a text version trimmed to one message, in about the time it takes to pour a coffee. She cuts a sentence that oversells, swaps 'lush' for 'thicker', and sends it. The draft was not publishable as written. It was also not a blank page at nine at night.
Why it matters
Generative AI is what makes it realistic for a small shop to produce marketing copy, images, and customer replies at a volume that used to mean hiring someone or retaining an agency. The real win is not brilliance; it is that the blank page stops being the bottleneck for the routine content owners never get around to: the reminder email, the seasonal offer, the service page nobody wrote. What it does not supply is judgment about your prices, your capacity, or your reputation. Output quality tracks the instructions and the editing it gets, so the pattern that holds up is fast drafter, human editor.