AI hallucination
An AI hallucination is a confident but false or invented statement produced by an AI model, such as a made-up fact, a wrong product detail or a non-existent source.
AI hallucination is the term for answers in which a language model states something untrue as if it were fact. Examples include inventing a quote, giving the wrong price or founding year for a company, attributing a product feature to the wrong brand, or citing a study that does not exist. The model is not lying on purpose. It generates the most likely sounding text, and sometimes that text is wrong.
Why hallucinations happen
- Gaps in training data. If a model saw little information about a topic or brand, it fills the gap with plausible guesses.
- Outdated knowledge. Without a web search, a model only knows what it learned before its cutoff date.
- Conflicting sources. When different sites give different information, the model may mix them up.
- Poor retrieval. In AI search, the system can pick an irrelevant or low-quality page and summarise it faithfully, which still produces a wrong answer.
- Misreading context. A model can attribute a statement from one part of a page to the wrong subject.
Grounding answers in retrieved sources reduces hallucinations considerably, but does not remove them.
Why it matters for brands
When an AI assistant tells a potential customer that your product lacks a feature it actually has, or quotes old prices, you lose business without knowing it. Hallucinations about brands often come from thin or inconsistent information online. The less clear, consistent material there is, the more the model has to guess.
How to reduce hallucinations about your brand
- Publish clear, factual pages about your company, products, prices and policies, and keep them current.
- Use consistent names and descriptions across your website, profiles, directories and press material.
- Add structured data for your organisation and products so key facts are machine-readable.
- Correct outdated information on third-party sites, such as old reviews, directories and articles.
- Earn coverage on trusted publications that describe your brand accurately, since these are likely sources for both training and retrieval.
- Monitor key prompts regularly to spot errors early.
Common mistakes
Assuming that a correct homepage is enough, while dozens of outdated pages elsewhere say something different. Also, do not use AI answers as a source for your own content without checking. Copying a hallucinated fact into an article spreads the error further and can damage trust in your site.
Questions about AI hallucination
What is an AI hallucination?
It is a false or invented statement that an AI model presents as fact, such as a wrong date, a made-up source or incorrect product information.
Can I correct what ChatGPT says about my brand?
You cannot edit the answer directly, but you can publish clear, consistent information and correct outdated sources, which improves what AI tools retrieve and learn.
Does grounding prevent hallucinations?
It reduces them by basing answers on retrieved sources, but a model can still misread or mix up those sources.