What is generative AI?

The short answer

Generative AI is AI that makes new stuff. You give it a prompt, and it writes the email, draws the image, spits out the code, or answers the question. That is the line between it and older AI, which mostly sorted things into buckets (spam or not spam, cat or dog) or predicted a number. Generative AI does not pick from a list. It produces something that was not there a second ago.

The big examples are the large language models behind ChatGPT, Gemini, Claude, and Perplexity. Same family, different products. And those models are exactly why search stopped being a list of blue links and started being a paragraph that answers you directly.

Generative vs the old kind of AI

For years, most useful AI was what people now call discriminative. It drew lines. Is this transaction fraud? Is this review positive? Which of these ten ads should I show? Useful, quietly everywhere, not flashy.

Generative AI flipped the question. Instead of "which bucket does this belong in," it asks "what comes next." A language model predicts the next word over and over, and somehow that simple trick, scaled up far enough, produces essays, working code, and a surprisingly good explanation of your tax situation. (Nobody fully expected it to work this well. It just did.)

How it actually makes things

Under the hood, a large language model is doing next-token prediction. It read a giant pile of text during training, learned the statistical shape of language, and now predicts the most likely next chunk given everything so far. String enough of those predictions together and you get a coherent answer.

That is also why it sometimes makes things up with total confidence. The model is optimizing for plausible, not for true. A prediction that sounds right can still be wrong, which is the polite way of describing a hallucination. So generative AI is brilliant and a little unreliable, often in the same sentence. Worth remembering before you paste its output into anything that matters.

Modern AI search bolts a retrieval step onto this so the model can answer from live web pages instead of pure memory, which cuts down (does not eliminate) the made-up stuff. If you want the deeper version, our guide to AI search walks through how that works.

Why marketers suddenly care

Here is the thing. Generative AI is what turned search into answers, and that broke the old game.

When someone asks ChatGPT or Google's AI Overviews a question, they often get a written answer and never click a single link. No click means no visit, and a chunk of your traffic just quietly evaporates. The whole point of search shifts from "rank in the list" to "be the source the answer is built from." Getting named in that answer is an AI citation, and earning more of them is the entire job of generative engine optimization.

This is the same muscle as SEO, aimed at a different reader. Your reader is now a model, not a person scanning results. Same fundamentals (clear content, real authority, clean structure), new target. Our AI SEO breakdown covers where the two overlap and where they split.

A few honest notes, since pretending generative AI is magic helps nobody:

• It is confident even when it is wrong. Always sanity-check facts and figures.
• Ask the same question twice and you can get two different answers. It is non-deterministic by design.
• It is great at fluent first drafts and shaky at precise truth. Use it for the former, verify the latter.

Where AI Citation Monitor fits

If generative AI is now answering your customers' questions, the obvious next question is this: does it mention you? Or your competitor?

That is the gap AI Citation Monitor measures. It checks whether ChatGPT, Perplexity, Gemini, Google AI Overviews, and Microsoft Copilot actually cite or recommend your brand, and it reports the result with confidence intervals so you are reading a real signal instead of one lucky run. It tracks five engines today, shows competitor share of voice, tracks which sources the engines pull from, and hands you prescriptive fixes instead of a vague "do better."

There is a free instant check if you just want to see where you stand right now, and plans run from Free at $0 up through Starter ($49), Growth ($129), and Agency ($349, with white-label).

Generative AI changed what an answer looks like. The smart move is to find out whether your answer is the one it gives.

FAQ