What is generative AI? Text, images, video and music made by machines
Generative AI creates new text, images, audio, video and code. Here's how it works, what it's used for, the main tools, and the big arguments about copyright and jobs.
First published · AI Digested
Generative AI is artificial intelligence that creates new content: text, images, audio, music, video or computer code. Type a request in plain English, called a prompt, and it produces something that didn't exist before.
It's the kind of AI that made the subject front-page news. ChatGPT, image generators like Midjourney, and video tools like Sora and Veo are all generative AI.
How is it different from other AI?
For most of its history, AI has been used to analyse things. Is this email spam? Is this payment fraud? What will this house sell for? Which film will you enjoy next?
Generative AI makes things. Instead of labelling a photo of a dog, it can draw a new one.
How it works, briefly
All generative AI follows the same basic recipe: learn the patterns in a huge amount of existing content, then produce new content that follows those patterns.
- Text generators (large language models) learn from vast amounts of writing and generate text one small piece at a time, each time predicting what's likely to come next. More on how they work.
- Image generators are mostly diffusion models. They learn by watching millions of pictures get gradually covered in random noise, and learning to reverse the process. To make a new image, they start with pure noise and remove it step by step, guided by your description.
- Video, music and voice tools use similar ideas, extended over time.
It doesn't copy and paste from its training data. It produces something new, though sometimes uncomfortably close to things it has seen.
What people use it for
Writing: drafting emails, letters, reports and posts; rewording things more clearly; translating; summarising long documents. See write emails with AI and summarise documents with AI.
Learning: explaining a hard topic at your level, quizzing you, or working through a problem step by step. See learn anything faster with AI.
Images and design: illustrations, mock-ups, product photos, logos, social media graphics. See make images with AI.
Coding: writing, explaining and fixing software. Programmers were among the earliest heavy users.
Audio and video: voiceovers, music, dubbing into other languages, short video clips.
The main tools
| What you want to make | Popular tools |
|---|---|
| Text and conversation | ChatGPT, Claude, Gemini, Microsoft Copilot, Meta AI, Le Chat |
| Images | ChatGPT, Gemini, Midjourney, Adobe Firefly, Stable Diffusion |
| Video | Sora, Veo, Runway |
| Music | Suno, Udio |
| Voices | ElevenLabs |
| Code | GitHub Copilot, Claude Code, Cursor, OpenAI Codex |
See our guide to the main AI models for who makes what.
The problems
It makes things up. Text generators can state false information confidently. Always check facts that matter. Why AI makes things up.
Copyright and consent. These systems learned from the work of millions of writers, artists, photographers and musicians, mostly without permission or payment. Authors, artists, news organisations, Getty Images and record labels have all brought lawsuits. Some cases have been settled and some licensing deals have been signed, but many key questions are still being decided by courts and governments.
Deepfakes and scams. Realistic fake images, voices and videos make fraud and misinformation easier. How to spot AI scams and deepfakes.
Bias. Output reflects patterns in the training data, including stereotypes.
Jobs. Writers, illustrators, translators, voice actors and customer service staff have been among the first to feel the effects. Will AI take my job?
Energy use. Training and running large models uses a lot of electricity and water for cooling data centres.
The bottom line
Generative AI is best thought of as a very fast, tireless assistant for first drafts, ideas and variations. It's at its best when a person who knows the subject guides it, checks the result and takes responsibility for what's finally used.
Keep going
Common questions
Is AI-generated content copyrighted?
It depends on the country and on how much a human contributed. In the US, the Copyright Office has said purely AI-generated material can't be copyrighted, though work where a person made meaningful creative choices may be. UK law has a special rule for 'computer-generated works', but how it applies to modern AI is being reviewed. If it matters for your business, get legal advice.
Can people tell if something was made by AI?
Not reliably. AI detectors make frequent mistakes, including flagging human writing as AI. Some companies add invisible watermarks or 'content credentials' to their output, but these aren't universal and can be removed.
Is it OK to use generative AI at work or school?
Check the rules first. Many employers and schools now have AI policies. Be open about using it, never paste in confidential information without permission, and check everything it produces.
Spotted a mistake or something out of date? Tell us and we will fix it. Read how we research and write our articles.
Read next
How ChatGPT works: large language models explained simply
How do chatbots like ChatGPT, Claude and Gemini actually work? A plain-English explanation of large language models: tokens, training, prediction and why they sometimes get things wrong.
Updated 22 November 2026
How to make images with AI: a beginner's guide
Create pictures, illustrations, logos and edited photos with AI tools like ChatGPT, Gemini, Midjourney and Adobe Firefly. Step-by-step, with prompts and the rules you need to know.
Updated 22 November 2026
What is artificial intelligence? A plain-English explanation
What AI actually is, how it learns, what it can and can't do, and why it suddenly seems to be everywhere. No jargon, no maths.
Updated 22 November 2026