generative AI (GenAI)
Generative AI is a type of artificial intelligence that creates new content upon your request: text, images, audio, video, or source code.
In plain terms
Generative AI is a program that does not just search or classify but creates new content. You describe what you want, and within seconds receive a letter, an image, a song, or a working code snippet.
Imagine it as an apprentice who has read countless books and seen countless images. It has learned how a good sentence or a sunset is constructed and assembles something new from that. It rarely copies word for word, but it may return a learned text or image almost unchanged.
Earlier AI systems mostly decided or predicted: is this email spam, what will traffic be tomorrow. Generative AI, by contrast, creates. This is why it became an everyday tool in offices, schools and homes within a few years.
You need to know that it does not understand the world as you do. It generates what seems probable based on learned patterns, so it can be confidently wrong.
How it works
- Training: the model processes vast amounts of text, images, or audio and learns the patterns within them, such as which word typically follows another.
- Request: you enter an instruction, or prompt. This can be a question, a task, or even an uploaded image or document.
- Generation: the model builds the answer step by step. For text, it repeatedly selects the next word fragment from probable continuations. For images, it starts from random noise and gradually cleans it into an image.
- Randomness: the generation process includes a degree of intentional randomness, so you generally do not get exactly the same result twice for the same request.
- Refinement: if you do not like the result, refine your request, and the model will create a new one considering the previous conversation.
An everyday example
You ask ChatGPT, Claude, or Gemini to write a polite complaint letter to your service provider, and within half a minute the finished text is ready. The same happens when your phone's photo editor removes strangers from the background and draws fitting details in their place.
Who is affected
A common misconception
Limits and risks
- It can make mistakes and invent facts (this is called hallucination), so always verify important data from a reliable source.
- It can be used to create deceptively lifelike fake images, audio, and video, so an image, audio recording, or video alone no longer proves that something actually happened.
- Do not enter sensitive personal, health, or corporate data until you know how the provider handles it and what it uses it for.
- The copyright status of training data and the legal protection of AI-generated works remain disputed in many countries.
Similar, but not the same
- artificial intelligence – Artificial intelligence is the broader umbrella term, while generative AI is the part of it that creates new content.
- large language model – The large language model is the type of generative AI that works primarily with text, while generative AI also includes systems that generate images, audio, and video.
- AI-agent – Generative AI creates content based on your request, while an AI agent builds on this to independently complete a task in multiple steps.
Related concepts
large language model, image generator, diffusion model
foundation model, prompt, hallucination, deepfake, chatbot, AI-agent
A brief history
Two important precursors to today's generative AI are the Generative Adversarial Network (GAN) introduced in 2014 and the transformer architecture published in 2017. In 2022, first image generators (such as DALL-E 2 and Stable Diffusion), and then in November, ChatGPT, made the technology known to the general public, and the term became widely used at that time.
Where things stand
By the end of 2026, generative AI has become an everyday tool: it is built into search engines, office software and phones. The focus of development is shifting from mere content creation to agents that carry out tasks independently. The EU AI Act has been in force since August 2024, and its rules apply in phases. Obligations for general-purpose AI models have applied since August 2025.
In the news
- Mashreq adds Clari5 generative AI to strengthen fraud risk managementNov 3, 2026 · Money
- Stakeholders offer competing views on how FDA should regulate generative AI in medtechNov 3, 2026 · Health
- Hungarian user Vera Mérő sought weight-loss help and received life-threatening advice from an AI chatbotNov 3, 2026 · Health
- Meta ad revenue growth outpaces Google Search as nearly 10 million advertisers use its AI toolsNov 2, 2026 · Marketing
- New York City schools and South Portland move to restrict generative AI for younger studentsNov 2, 2026 · Education & Family
- Dutch lower house of parliament moves to ban generative AI for primary school pupilsNov 2, 2026 · Education & Family
- Dutch lower house wants to ban generative AI in primary schoolsOct 29, 2026 · Education & Family
- Only 35% of open source projects have a GenAI policy as AI-generated code floods inOct 26, 2026 · Code
Also known as
generative AI · generative AI · generative AI · GenAI · gen AI · generative AI · generative artificial intelligence · generative AI
Sources
This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.