AI basics
What are AI hallucinations?
AI hallucinations are confident but false or invented answers from a model. Learn why they happen, how to spot them, and how to reduce them.
An AI hallucination is an answer that sounds confident and plausible but is false, invented, or not supported by any real source. A language model might cite a book that does not exist, state a wrong date, or describe a feature a product never had, all in a fluent, assured tone.
The term is a metaphor. The model is not seeing things; it is generating text that fits a pattern without checking it against reality.
Why do AI hallucinations happen?
A large language model is trained to predict likely next text. It is not a database and has no built-in fact checker. Several causes follow from that:
- Plausible beats true. The model optimizes for text that looks right, so a smooth wrong answer can win over an honest "I don't know".
- Gaps in knowledge. If the topic was rare or missing in training, the model may fill the gap with a guess.
- Outdated information. Facts change after training, and the model does not know what it has not been given.
- Missing context. If the needed detail fell outside the context window, the model may improvise.
- Ambiguous prompts. Vague questions invite the model to assume details.
What do hallucinations look like?
- Invented citations, quotes, links, or statistics.
- Wrong names, dates, or numbers stated confidently.
- Made-up details about a person, product, or event.
- Code that calls functions that do not exist.
- Correct-sounding explanations that contain a subtle logical error.
How can you spot a hallucination?
- Be extra careful with specifics: numbers, names, quotes, and links.
- Check claims against a reliable primary source.
- Ask the model to show how it reached an answer, then verify those steps.
- Rephrase the question and see whether the answer changes. Big changes are a warning sign.
- Treat very polished answers on obscure topics with suspicion.
How can you reduce hallucinations?
You cannot eliminate them, but you can lower the odds:
- Provide the source material in the prompt and ask the model to answer only from it.
- Allow uncertainty. Say "if you are not sure, say so" in your instructions.
- Be specific about what you want. See how to write good prompts.
- Keep tasks focused instead of asking for many things at once.
- Verify anything that carries real consequences, such as health, legal, or financial decisions.
Are hallucinations a bug that will be fixed?
They have become less common as models improve, but they come from how language models work, so no one can promise they will disappear. The safe habit is to treat AI output as a first draft. For the technical term in context, see Wikipedia on hallucination in AI).
Common mistakes
- Copying AI-generated citations without checking they exist.
- Assuming confidence means correctness.
- Asking for facts the model could not possibly know, like private details.
Hallucinations and Based
Based is an AI chatbot, so like any language-model assistant it can occasionally be wrong. Give it the relevant text in your message when accuracy matters, and double-check important answers. You can learn how it works in the AI chat docs or see its summaries feature, where you can supply the text to be summarized directly.
Frequently asked questions
What is an AI hallucination?
It is a confident-sounding answer from an AI that is false or made up, such as an invented source or a wrong fact.
Why do AI models hallucinate?
Language models predict plausible text rather than check facts, so when they lack information they can generate a believable guess.
How do I stop an AI from hallucinating?
You cannot fully stop it, but you can reduce it by supplying the source text, asking it to admit uncertainty, and verifying important claims.
Can AI hallucinations be dangerous?
They can mislead, especially on medical, legal, or financial topics, so check high-stakes answers against trusted sources.
Keep learning
- AI basics
What is a large language model (LLM)?
A large language model is an AI trained on huge amounts of text to predict and generate language. Learn how LLMs work, what they do well and where they fail.
- AI basics
What is a context window in AI?
A context window is the maximum amount of text an AI model can consider at once. Learn what tokens are, what happens when it fills up, and how to work within it.
- AI basics
How to write good AI prompts
A good AI prompt states the task, context, format and constraints clearly. Learn practical steps, examples and common mistakes to get better answers.
- AI basics
What is an AI chatbot?
An AI chatbot is software you talk to in everyday language, powered by a language model. Learn how it works, how it differs from older bots, and its limits.
Try it in your server
Based is an AI Discord bot in 300+ servers. Every slash command also works with the "." prefix.