Guide
Glossary
Hallucination
A hallucination is an answer from an AI model that sounds confident and plausible but is false or has no support in any source, such as a case citation that does not exist or a figure that appears in no document. It happens because language models generate likely-sounding text rather than looking facts up. Answering from supplied documents reduces hallucinations, but no current tool removes them. The practical response is verification: check every citation, quotation and number against the original before relying on it, and prefer tools that show where each answer came from.
Also called AI hallucination, confabulation, fabricated citation
Last updated:
Example
In a law firm
A solicitor asks a general AI tool for authorities on a limitation point and gets three case names with neat citations; one does not exist. In NSW Supreme Court proceedings, Practice Note SC Gen 23 requires anyone who uses generative AI in written submissions to verify that every citation exists, is accurate and is relevant, and that check must not be done with an AI tool.
How do you reduce AI hallucinations?
You reduce hallucinations by giving the AI the source material, asking it to answer only from that material, and checking its work. Tools that use retrieval-augmented generation answer from your firm’s own documents and show the passage behind each answer, which makes checking faster. Prompts that tell the model to say when it does not know also help. None of this removes the need for a person to verify citations, quotations and figures before they reach a client, a regulator or a court.
Guides that explain it in context
Secure by design. Set up correctly. Fully managed.
Talk to us before you commit to anything
Start with a free 45-minute discovery call. We look at your systems and priorities, then recommend a first step with a fixed scope, or tell you if we are not the right fit.