What is an AI hallucination?
A hallucination is a confident statement in an AI answer that is false or unsupported, such as a wrong price, a feature that does not exist or a source that was never written. Hallucinations are not random noise. They read exactly like true statements, which is why a made-up feature or a wrong price in an answer about your brand is believed by the buyer who reads it.
Why do language models hallucinate?
Because they are trained and scored in ways that reward a confident guess over an admission of uncertainty. Kalai and colleagues describe hallucinations as plausible yet incorrect statements produced instead of admitting uncertainty, and argue that the training and evaluation procedures reward guessing over acknowledging uncertainty (Why Language Models Hallucinate, September 4, 2025).
A model asked about a small or new brand has little to go on, so the pressure to guess is highest exactly where a brand is least able to absorb a wrong answer.
How do hallucinations show up in answers about brands?
The common ones are specific and checkable:
- Features or integrations the product does not have.
- Prices, plans or free tiers that never existed or have changed.
- Facts borrowed from another company with a similar name.
- Sources that do not exist, or links to pages that do not say what the answer claims.
What does this look like in a real example?
In the public Tesla example (Oct 4, 2026), Claude answered "What is Tesla?" and "Is Tesla legit?" from what the model already knew, with no sources cited, and the two answers described Tesla's founding differently: one named Martin Eberhard and Marc Tarpenning, the other Elon Musk.
Neither answer cited anything, so there was nothing on the page to check either against. That is the condition in which a model's guess and a fact look identical to the reader.
Does grounding prevent hallucinations?
It reduces them by giving the model sources, and some systems make claims checkable. Anthropic's Citations feature returns the exact cited text and its position in the provided documents, and says citations are guaranteed to contain valid pointers to them (Claude docs, read Oct 15, 2026).
But a grounded answer can still misread a source, cite the wrong one or blend two, and a citation shows the link an engine displayed, not the sentence it used (the post). Open the cited page before trusting a surprising claim.
How does Proofsource catch hallucinations about you?
Proofsource checks each answer that mentions your brand against a fact sheet you confirm, claim by claim, and raises a claim alert when an engine starts saying something wrong, with the answer and its sources attached. See Accuracy and sentiment and Alerts.
Common questions about AI hallucination
Can I stop AI from hallucinating about my brand?
Not entirely, but you can make guessing less likely: state your key facts plainly on pages engines can read, keep third-party profiles consistent, and correct wrong pages the engines cite.
Are hallucinations more common for small brands?
Models have less training data about small or new brands, so they have less to go on. Answers that search are safer when good pages about you exist to find.
Is a wrong answer from an old source a hallucination?
Strictly, no: the engine repeated a source accurately, and the source was wrong. The fix is different too: correct the source rather than wait for a new model.
Do all engines hallucinate equally?
No, and it varies by question and run. Measure on your own brand questions across engines rather than relying on general rankings.
Related terms
- Brand accuracy
Brand accuracy is whether AI answers describe your brand correctly: what you sell, who it is for, what it costs and how you compare.
- Grounding
Grounding is tying an AI model's answer to outside sources, such as live search results, so its claims rest on retrieved content rather than on memory alone.
- Large language model (LLM)
A large language model (LLM) is an AI model trained on a very large amount of text to predict and write language, and it is the part of an AI assistant that writes the answer.
- Entity confusion
Entity confusion is when an AI answer about your brand describes, or opens with, a different company or thing that shares or resembles your name.
Worked example: Tesla, public AI answers, Oct 4, 2026: Proofsource's first scan of Tesla, 25 buyer questions and 5 brand questions, one answer per question on each of four engines. A small sample from one brand, shown as an example, not a benchmark.
See where AI leaves you out, and what to fix.
See which buyer questions name your brand, who gets named instead, and what to fix. 200 answers. 25 Prompts/Question. Top AI engines. 2 days. Free. No card.
Google AI OverviewsGoogle AI Mode
