What is Hallucination (AI)?
A hallucination is when an AI tool produces confident, fluent output that is factually wrong or entirely invented.
Hallucination is the industry term for an AI system stating something untrue as if it were established fact — a fake statistic, a court case that never happened, a citation to a non-existent paper, or an API method a library does not have. It happens because language models are optimised to produce plausible text, not verified text. When the training data is thin on a topic, the model still generates something in the right shape. Hallucinations are reduced, never fully eliminated, by grounding the model in real sources with retrieval, asking for citations, keeping questions inside the tool's actual knowledge base, and having a human check anything that carries legal, medical, or financial risk. When you compare AI tools, treat visible sourcing as a feature, not a nicety: a tool that links every claim to a document lets you audit it in seconds, while a tool that outputs polished prose with no provenance shifts the entire verification burden onto you.
Also known as
- AI hallucination
- confabulation
- made-up facts
Related tools
- Perplexity AI — AI search engine that gives direct answers with cited sources.
- ChatGPT — The most versatile AI assistant for writing, research, and brainstorming.
- Claude — Thoughtful, safety-focused AI assistant with excellent reasoning.
- Google Gemini — Google's multimodal AI with real-time information access.
- Grammarly — Fix grammar, improve clarity, and adjust tone across everything you write.
Related terms
- RAG (Retrieval-Augmented Generation) — RAG is a technique where an AI tool searches your documents first, then writes an answer grounded in what it found.
- Large Language Model (LLM) — An LLM is an AI model trained on huge amounts of text that predicts language well enough to write, summarise, translate, and answer questions.
- Generative AI — Generative AI is any AI that creates new content — text, images, audio, video, or code — rather than only analysing existing data.
- Context Window — The context window is the maximum amount of text, measured in tokens, an AI model can consider at one time.
Category: AI Concepts