How to Spot and Prevent Hallucinations

# How to Spot and Prevent Hallucinations

Now that you understand what hallucinations are and why they happen, this lesson focuses on actionable techniques you can use immediately — in any AI tool — to reduce hallucination risk and catch errors before they matter.

Part 1: Prevention Through Better Prompts

Technique 1: Ground the Model with Source Material

The single most effective way to reduce hallucinations is to give the AI the information it needs rather than asking it to generate from memory.

High hallucination risk: > "Summarize the key findings of the 2024 WHO report on global health."

The model may not have this report in its training data. It will likely fabricate plausible-sounding findings.

Low hallucination risk: > "Here is the executive summary of the 2024 WHO report on global health: [paste text]. Summarize the key findings based ONLY on the text I provided. Do not add information from other sources."

When you provide the source text, the model works as a text transformer rather than a knowledge generator. This dramatically reduces fabrication.

Technique 2: Explicitly Permit "I Don't Know"

Models are trained to be helpful, which makes them reluctant to admit uncertainty. Override this by explicitly giving permission:

This simple instruction measurably reduces confident-sounding fabrications across all major models.