---
title: "What Is Generative AI? — Thoughtful Comics"
description: "How AI makes new content from learned patterns, and why a polished draft still needs checking."
source: "https://thoughtfulrobots.ai/thoughtful-comics/what-is-generative-ai"
---

# What Is Generative AI?

How AI makes new content from learned patterns, and why a polished draft still needs checking.

![A four-panel comic. Panel 1. The reader: “Did the AI find these somewhere, or did it make them?” The robot: “Generative AI created new outputs from patterns it learned during training.” Panel 2. The reader: “How can one system create so many kinds of content?” The robot: “A generative model predicts an output that fits your prompt and the patterns in its training.” Panel 3. The reader: “If the output is original and polished, is it reliable?” The robot: “Not automatically. Generation can create useful drafts and convincing mistakes, including hallucinations.” Panel 4. The reader: “I will use it to create options, then check facts and apply my judgment.” The robot: “That is a strong human-AI workflow.” Glossary. Generative AI: AI that creates new content such as text, images, audio, video, or code. Generative model: A model trained to produce new outputs from learned patterns. Prompt: The request or instruction given to an AI system. Hallucination: An AI output that sounds plausible but is unsupported, inaccurate, or invented.](/images/comics/what-is-generative-ai/strip.webp)

Click to enlarge ⤢

## The takeaways

- 01 Generative AI produces new content rather than only retrieving existing content.

- 02 Different generative models work with different kinds of media.

- 03 A generated answer still needs review, especially when it makes factual claims.

- 04 Generate quickly, then verify and refine before the output is used.

- Reader Did the AI find these somewhere, or did it make them? Robot Generative AI created new outputs from patterns it learned during training. Takeaway: Generative AI produces new content rather than only retrieving existing content.

- Reader How can one system create so many kinds of content? Robot A generative model predicts an output that fits your prompt and the patterns in its training. Takeaway: Different generative models work with different kinds of media.

- Reader If the output is original and polished, is it reliable? Robot Not automatically. Generation can create useful drafts and convincing mistakes, including hallucinations. Takeaway: A generated answer still needs review, especially when it makes factual claims.

- Reader I will use it to create options, then check facts and apply my judgment. Robot That is a strong human-AI workflow. Takeaway: Generate quickly, then verify and refine before the output is used.