Hallucination
A fluent, confident answer that is plain wrong: an invented citation, function, statistic or fact.
Nothing has broken when a model hallucinates. It produced the most plausible continuation, and a made-up reference that looks like a real one is very plausible. Smooth wording says nothing about accuracy.
The cases that catch people are specific and easy to check: library functions that do not exist, papers with real-sounding authors, legal citations, version numbers, prices. If the model states a number or a name, verify it.
You can lower the rate. Supply the source material so it does not lean on training data. Ask it to flag what it is unsure about. Give it a way to answer with nothing: a prompt that demands three findings gets three findings whether or not they exist. Retrieval, covered under rag, goes after the same problem from the data side.
Tagged
Where this comes up
Prompts, configs and tutorials in the library that touch this term.
- GuideUniversal
Turn a messy task into a prompt in five minutes
Most weak prompts come from a task nobody fully described. A five-step routine to fix that before you type, with a worked example.
- GuideUniversal
Summaries you can trust
A summary is a claim about a document, so it can be wrong in the same ways as any claim. Four habits keep it honest.
- GuideUniversal
Briefing a coding assistant: the first message
A coding assistant works from what you tell it and what it can see. Here is what to put in the first message so you spend less time correcting it.
- GuideUniversal
How to check what an AI tells you
A few checks that catch most wrong answers: which claims to verify, how to ask for sources, and why a confident tone tells you nothing.
- CheatsheetChatGPT
ChatGPT commands and keyboard shortcuts
The commands and shortcuts the interface never shows you, in one table. Mostly keyboard, mostly undocumented.
Tested on GPT-5 · May 2026
- GuideUniversal
How to write a prompt that actually works
A short checklist to run any prompt through before you send it: who, what, shape and what to leave out. Those four things separate a specific prompt…