Phase 1 · Building with LLMs · Context & retrieval
Hallucinations and grounding
Why models make things up, and how to design against it.
Open this lesson Create a free account
Why it matters
A model that invents a statistic in a blog post, a discount in a customer reply or a "clinically proven" claim in an ad creates real liability: advertising law expects claims to be substantiated, and customers screenshot everything. You can't make a model incapable of making things up, but you can design prompts and checks that catch most of it before anyone sees it.
What you'll be able to do
- Explain why language models produce confident, false statements
- Recognize which kinds of output carry the most risk
- Apply prompt techniques that reduce hallucination: grounding, permission to abstain, quotes first
- Write code that flags numbers and quotes an answer can't back up
Before this lesson
Next in this phase: Tool use