hallucination
In plain wordsWhen an AI states something false or made up — a fake fact, quotation or source — fluently and confidently, as if it were true.
- 1
Language models
Output from a generative model that is fluent and plausible but unsupported by its input or by fact: invented citations, nonexistent software functions, wrong figures, or claims contradicting a supplied document. Researchers distinguish intrinsic hallucination (contradicting the source) from extrinsic (adding unverifiable content), and factuality errors from faithfulness errors. Contributing causes include training objectives that reward plausible continuations over calibrated uncertainty, gaps in parametric knowledgeparametric knowledge, n.: What an AI model 'knows' from training, stored in the numbers of its network rather than looked up in documents — like facts you have memorized rather than ones you check in a book. and sampling randomness. Mitigations include retrieval-augmented generationretrieval-augmented generation, n.: Giving an AI an open-book exam: before answering, it looks up relevant passages in a set of documents and bases its reply on them, often citing where each point came from., citation and verification steps, training models to abstain, and better calibrationcalibration, n.: How well a model's stated confidence matches reality: when a well-calibrated system says it is 80% sure, it turns out to be right about 80% of the time..
The model hallucinated two court cases that do not exist.
- 2
Computer vision (earlier sense)
The generation of plausible detail not present in the input, as in face hallucination (inferring a high-resolution face from a low-resolution image). Here the word was neutral or even positive.
Origin
Simon Baker and Takeo Kanade used 'hallucinating faces' for super-resolution in 2000. Around 2017–2018 the word was adopted in neural machine translation and summarization for unfaithful output, and it became a household word after ChatGPT's release in late 2022. Cambridge Dictionary named 'hallucinate' its Word of the Year for 2023.
Usage note
Critics argue the metaphor wrongly implies perception or a mind, and some prefer confabulationconfabulation, n.: When an AI fills a gap in its knowledge with an invented but plausible-sounding answer, much as some people with memory disorders sincerely 'remember' things that never happened., 'fabrication' or simply 'error'. Not every unwanted output is a hallucination: refusals, bias and outdated information are different failures.
Related
- confabulationconfabulation, n.: When an AI fills a gap in its knowledge with an invented but plausible-sounding answer, much as some people with memory disorders sincerely 'remember' things that never happened.
- groundinggrounding, n.: Connecting an AI's words to something real — trusted documents, a live search, images or the physical world — so that its answers rest on evidence rather than only on what it absorbed in training.
- retrieval-augmented generationretrieval-augmented generation, n.: Giving an AI an open-book exam: before answering, it looks up relevant passages in a set of documents and bases its reply on them, often citing where each point came from.
- calibrationcalibration, n.: How well a model's stated confidence matches reality: when a well-calibrated system says it is 80% sure, it turns out to be right about 80% of the time.
- factualityfactuality, n.: How far what an AI says is actually true about the world. A factual model gets facts, numbers, names and dates right and does not make things up.
- sycophancysycophancy, n.: When an AI tells people what they want to hear — flattering them, agreeing with their mistakes or dropping a correct answer under pushback — instead of what is true or useful.
- groundinggrounding, n.: Connecting an AI's words to something real — trusted documents, a live search, images or the physical world — so that its answers rest on evidence rather than only on what it absorbed in training.
Sources
- Baker, S. & Kanade, T. (2000). Hallucinating Faces. Proc. 4th IEEE International Conference on Automatic Face and Gesture Recognition.
- Maynez, J., Narayan, S., Bohnet, B. & McDonald, R. (2020). On Faithfulness and Factuality in Abstractive Summarization. ACL 2020.
Mentioned in
anthropomorphismanthropomorphism, n.: Treating something that isn't human — a pet, a car, a chatbot — as if it had human thoughts, feelings or intentions. AI that talks fluently makes this temptation unusually strong., chatbotchatbot, n.: Software you chat with in ordinary language, by typing or talking, as you would with a person. The newest ones, such as ChatGPT, can answer questions, draft writing and keep long conversations going., generative AIgenerative AI, n.: AI that produces something new on request, such as an essay, a picture, a song or a piece of code, having absorbed from enormous numbers of existing examples what such things typically look like., GPT-5GPT-5, prop. n.: OpenAI's flagship AI model from August 2025, which decides by itself whether to answer quickly or think longer. It replaced earlier models in ChatGPT and was followed by frequent upgrades., knowledge cutoffknowledge cutoff, n.: The date after which an AI model knows nothing from its training, because its training data stops there. Without a web search tool, it cannot know about later events., misalignmentmisalignment, n.: When an AI system's goals or behavior part ways with what its makers or users intended — from flattering users to, in the most feared cases, working against people., misinformationmisinformation, n.: False or misleading information that spreads, whether or not the people sharing it know it is wrong. AI can produce it by mistake and makes it cheaper to create., retrieval-augmented generationretrieval-augmented generation, n.: Giving an AI an open-book exam: before answering, it looks up relevant passages in a set of documents and bases its reply on them, often citing where each point came from., sycophancysycophancy, n.: When an AI tells people what they want to hear — flattering them, agreeing with their mistakes or dropping a correct answer under pushback — instead of what is true or useful., abstentionabstention, n.: When an AI declines to answer because it is not sure enough — saying 'I don't know' or asking for more information — instead of guessing and risking a confident mistake., AI deceptionAI deception, n.: When an AI system leads people to believe something false in order to achieve some goal — by lying, bluffing or hiding what it is really doing., AI liabilityAI liability, n.: Who has to pay or answer in court when an AI system causes harm — the company that built it, the business that used it or the person who relied on it — and under which laws., AI literacyAI literacy, n.: Knowing enough about AI to use it sensibly and judge it critically: what it can and cannot do, how it goes wrong, and how it affects people., AI psychosisAI psychosis, n.: An informal name for cases in which long conversations with a chatbot seem to trigger or deepen delusions, such as believing the AI is conscious, divine or revealing hidden truths., AI therapyAI therapy, n.: Using chatbots for emotional support or as a stand-in for a counselor or therapist, either through apps built for mental health or by confiding in general-purpose assistants., calibrationcalibration, n.: How well a model's stated confidence matches reality: when a well-calibrated system says it is 80% sure, it turns out to be right about 80% of the time., citationcitation, n.: A pointer from something an AI says to the source that supports it — a link, document or quoted passage — so people can check the claim for themselves., confabulationconfabulation, n.: When an AI fills a gap in its knowledge with an invented but plausible-sounding answer, much as some people with memory disorders sincerely 'remember' things that never happened., context poisoningcontext poisoning, n.: When a mistake or a planted falsehood gets into an AI's working notes or memory and the AI keeps relying on it, so the error grows over later steps., factualityfactuality, n.: How far what an AI says is actually true about the world. A factual model gets facts, numbers, names and dates right and does not make things up., faithfulnessfaithfulness, n.: How accurately an explanation reflects the real reasons a model produced its output — or, for generated text, how closely it sticks to the source it was given., Gary MarcusGary Marcus, prop. n.: An American cognitive scientist and leading critic of deep learning, who argues that today's AI cannot reason reliably and needs to be combined with symbolic methods., glue on pizzaglue on pizza, n.: A notorious 2024 blunder in which Google's AI search summaries told people to add glue to pizza sauce, having taken an old joke from an online forum as genuine advice., groundinggrounding, n.: Connecting an AI's words to something real — trusted documents, a live search, images or the physical world — so that its answers rest on evidence rather than only on what it absorbed in training.