> ## Documentation Index
> Fetch the complete documentation index at: https://simpai.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# What is AI hallucination and why prompts cause it

> AI hallucination is when a model confidently generates false information. Learn why vague prompts are the leading cause and how S.I.M.P. helps prevent it.

AI hallucination is when a model confidently generates false information — not because it is lying, but because it genuinely doesn't know what it doesn't know. When your prompt leaves gaps, the model fills them in with plausible-sounding but invented content. S.I.M.P. scores your prompt for hallucination risk and rewrites it to close those gaps before you send it.

## Why hallucinations happen

Language models work by predicting the most statistically likely continuation of text. When your prompt is vague or underspecified, the model has little to anchor its response to. Rather than stopping to ask for clarification, it fills in the blanks — drawing on patterns from its training data that may have nothing to do with your actual intent.

The result is a response that sounds confident and coherent but may contain:

* Facts that were never true
* Dates, names, or figures that are plausible but wrong
* Advice that doesn't apply to your specific situation
* References to sources that don't exist

This is not a flaw that future models will simply eliminate. It is a fundamental consequence of how these systems generate text — and your prompt is one of the most powerful levers you have over it.

## How your prompt contributes to hallucination risk

Three prompt characteristics are most strongly correlated with hallucination:

**Vague scope.** When you ask the model to "tell me about" something without specifying what aspect, depth, or purpose, it decides on your behalf — and its choice may not match yours.

**Missing context.** The model doesn't know who you are, what you already know, or what you're trying to accomplish. Without that context, it makes assumptions, and those assumptions can be wrong.

**No output constraints.** When you don't specify a format, length, or structure, the model picks one. A loosely constrained prompt often leads to a loosely constrained response — one that wanders away from the specific information you needed.

## The spectrum: not all hallucinations are equal

Hallucination is not a single type of failure. You are more likely to catch and correct it if you know what form it takes:

* **Factual errors** — the model states something objectively false (a wrong date, a misattributed quote, a made-up statistic)
* **Off-topic responses** — the model answers a question you didn't ask because your prompt was ambiguous about which question you meant
* **Format issues** — the model structures its response in a way that obscures key information or invents structure that wasn't requested

Factual errors are often the hardest to spot because they blend in with correct information. Off-topic responses and format issues are usually more obvious but still waste your time.

## How S.I.M.P. addresses this

S.I.M.P. analyzes your prompt across four dimensions: missing context, vague scope, ambiguous language, and missing output constraints. It assigns a risk score from 1 to 10 and a label — Low, Moderate, High, or Critical — based on how many of these problems are present and how severe they are.

After scoring, S.I.M.P. rewrites your prompt into a sanitized version and shows you exactly what changed and why, so you can understand what made the original risky and apply the same thinking to future prompts.

<Info>
  No tool can eliminate hallucination entirely. S.I.M.P. reduces risk by making your prompt as unambiguous as possible.
</Info>


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.