Glossary

What is an AI hallucination?

Also known as AI hallucination, confabulation, AI error

Definition

A hallucination is when an AI engine states something false as if it were true, such as wrong prices, opening hours, services or history for a business. The answer sounds confident, which makes the error easy to miss.

Key points

A hallucination is a confident but false statement in an AI answer.
Some come from pure invention, others from real pages that are wrong or out of date.
Finding the source of the wrong detail is usually the first step to fixing it.

What does it mean in practice?

A veterinary practice closed its second branch last year. ChatGPT still lists both branches, and Gemini says the practice offers out-of-hours cover it has never offered. Both errors can send customers to the wrong place.

Some errors are pure invention. Others come from real pages that are out of date or wrong, such as an old directory listing. OpenAI describes hallucinations as plausible but false statements made by language models. It links them partly to training that rewards guessing over admitting doubt.

Why does it matter for AI search?

A wrong answer about your business can cost enquiries without you knowing. Buyers rarely check what an AI engine told them, and they rarely tell you where they heard it.

The fix usually starts with finding the source. If an engine cites an out-of-date page, updating that page is the most direct route. If it names no source, look for pages online that repeat the wrong detail.

Answer accuracy in Axiom GEO checks each claim in an answer against an approved fact sheet of the business. For engines that name no source, a source finder shows the pages that state the wrong detail. Our guide on when AI gets your business wrong walks through the steps.

These terms connect to hallucinations.

Grounding, which reduces made-up answers but does not stop them.
NAP consistency, keeping your core details the same everywhere.
Entity, how engines tell your business apart from similar ones.
Citation, the page an engine points to as its source.

Common questions

Why does ChatGPT give wrong information about my business?

Usually because it found or learned something wrong. An old directory listing, an out-of-date page or a similarly named business can all feed wrong details into an answer. Models also guess when they are unsure. Finding the page that states the wrong detail is the best place to start.

Can I get an AI engine to correct a wrong answer?

Not directly, but you can change what it finds and learns. Fix the wrong detail on your own site and on the pages the engine cites, and keep your facts consistent across listings. Engines that search the web can pick up changes fairly quickly. Answers from training take longer to change.

Reviewed 26 September 2026

All terms A to Z