A chatbot gives a plausible journal title and URL for a paper that does not exist.
A generated answer can sound specific and confident without being grounded in a source. Hallucination is commonly used for invented content, rather than every possible error. A calculation mistake or poor wording is not necessarily the same phenomenon. IBM explains the distinction.
Check the parts that could change your decision: names, dates, figures, quotations and links. Open an alleged source and inspect the relevant passage. A citation-looking string is not proof that a document exists or says what the model claims.
Give the model the material you want it to use and ask it to identify missing information. That can help structure review, but “do not hallucinate” is not a guarantee. For exact calculations, use a calculation tool and check its inputs. Keep source evidence alongside the generated explanation.
Sources
Primary references checked 11 October 2026. This explanation is not a benchmark of a particular model or computer.