AI hallucination
An AI hallucination is when an AI tool states something false or invented while sounding completely confident, presenting a wrong fact, a fake citation, or a made-up source as if it were true. For students using general chatbots to study, it is the main reason answers can quietly mislead.
Drafted with AI by the StudyPDF team. We check dates, numbers and research claims against official sources and read pages on a sample basis. How we write these pages
In more detail
Large language models predict likely-sounding text, not verified facts. When a model does not actually know something, it often guesses rather than admitting the gap, so the answer can read as polished and certain while being wrong. The confidence is the trap. A hallucinated explanation looks exactly like a correct one.
A common misconception is that hallucinations only happen on obscure topics. They also show up on basic material: a wrong date, a misattributed quote, a citation to a paper that does not exist. If the answer comes with no source you can check, you have no way to tell a real fact from a fluent guess.
Example
A student asks a general chatbot to explain the steps of the Krebs cycle before an exam. The bot lists eight steps confidently, but swaps two intermediates and invents an enzyme name. The student memorizes it, then loses marks on the exam for an error the AI introduced.
How StudyPDF does this
Bo answers only from the material you upload, with a citation to the exact PDF page, video second, or source. When something is not in your material, Bo tells you instead of inventing an answer, which is what keeps hallucinations out of your study.
See also
- TermCited answer
- TermRetrieval-augmented generation (RAG)
- TermCourse-scoped AI
- TermStudy agent
- ComparisonNotebookLM vs ChatGPT: which is better for studying?
- AlternativesThe best NotebookLM alternatives in 2026
- ComparisonClaude vs ChatGPT: which is better for studying?
- ComparisonChatGPT vs Gemini: which is better for studying?
- ComparisonPerplexity vs ChatGPT: which is better for studying?
- ComparisonGrok vs ChatGPT: which is better for studying?
Common questions
Why do AI chatbots make up facts so confidently?
They are trained to predict plausible text, not to know the truth, so when they lack a real answer they guess rather than say they do not know. The output sounds fluent and certain because that is how the training data reads. Confidence is not a sign of accuracy.
How can I tell if an AI answer is a hallucination?
Check for a real, verifiable source. A hallucinated answer usually sounds sure but offers no citation you can open, or points to something that does not exist. Cross-check any fact against your actual course material before you trust it.
Put it to work on your own course
Upload your lectures and ask Bo. Cited answers, plus flashcards, exams, and more built from your material.