Generative AI is the branch of artificial intelligence that creates new content — text, images, audio, video or code — instead of only classifying or predicting.

How a chatbot writes

Chatbots are powered by large language models (LLMs). During training they read enormous amounts of text and learn to predict the next piece of a sentence. At its core, answering your question is a long series of these predictions, one small piece (a token) at a time. Additional training with human feedback teaches the model to follow instructions and be helpful.

How an image generator draws

Most image generators use diffusion models. They learn to turn random noise into an image step by step, guided by your text description. Video and audio generators use related ideas.

What generative AI is good at

  • Drafting and rewriting text
  • Summarizing and explaining
  • Brainstorming ideas
  • Translating
  • Writing and explaining code
  • Creating illustrations and mock-ups

What to watch out for

  • Hallucinations. Models can state false information confidently. Check facts that matter.
  • Out-of-date knowledge. A model only knows what was in its training data unless it can search the web or read your documents.
  • Bias. Models can reflect stereotypes present in their training data.
  • Privacy. Don’t paste confidential information into tools whose data policies you haven’t checked.
  • Copyright. Rules on AI-generated content and on training data are still being decided in courts.

A good mental model

Think of generative AI as a very fast, very well-read assistant that never says “I’m not sure” unless you ask it to. Excellent for first drafts; your judgment is still needed for the final version.

Everyday examples

  • Writing: drafting emails, summarizing articles, rewriting text in a different tone.
  • Images: creating illustrations, editing photos, generating product mock-ups.
  • Audio: text-to-speech voices, music generation, translating speech.
  • Video: short clips from text descriptions, automatic subtitles and dubbing.
  • Code: suggesting lines of code, explaining errors, building simple apps from descriptions.
  • Search and research: answering questions with citations from the web.

Many of these features are built into tools people already use — office software, phones, photo apps and search engines — sometimes without being labelled as “generative AI”.

How to try it safely

  1. Start with low-stakes tasks, like drafting a message or planning a meal.
  2. Check facts for anything important; generative AI can be confidently wrong.
  3. Don’t share sensitive information such as passwords, financial details or private documents unless you trust the service’s policies.
  4. Be transparent when you use AI-generated content in school or work, following the rules that apply.
  5. Respect others’ rights: avoid generating images of real people in misleading or harmful ways.

Generative vs. other AI

Not all AI is generative. Traditional machine learning predicts or classifies — spam filters, fraud detection, recommendations. Generative AI creates new content. Many products combine both: a shopping app might use predictive models to recommend products and a generative model to answer questions about them.

Key terms in one minute

  • Model: the trained system that generates content.
  • Prompt: your instruction or question.
  • Token: a chunk of text the model reads and writes, roughly part of a word.
  • Hallucination: a confident but false output.
  • Multimodal: able to handle several types of input or output, such as text and images.
  • Fine-tuning: further training a model for a specific task or style.

Next, read what machine learning is, how diffusion models generate images and seven myths about AI.

Key takeaways

  • Generative AI creates new text, images, audio, video and code from patterns in training data.
  • Chatbots predict text token by token; image generators turn noise into pictures.
  • It is useful for drafting, explaining, creating and summarizing.
  • It can be wrong, biased or misused, so verify important outputs.
  • Start with low-stakes tasks and protect sensitive information.

Frequently asked questions

Is ChatGPT generative AI?

Yes. ChatGPT is a chatbot built on large language models, a type of generative AI that produces text and, in newer versions, images and audio.

Does generative AI copy existing content?

It generally creates new combinations based on learned patterns, but it can sometimes reproduce parts of training data, which raises copyright questions.

Do I need to pay to use generative AI?

Many tools offer free versions with limits; paid plans add more powerful models and features.