Decoder. plain-English AI glossary

Fact-Checking

● Core

Also called Fact-Checking

Verifying a claim against trusted evidence to decide whether it’s actually true.

Think of it like

A journalist calling a second and third source before printing — no claim runs on one voice alone.

Example

A system extracts the claim "the company was founded in 1998," searches an authoritative source, and marks it true, false, or unverifiable.

How it actually works

Fact-checking pins a claim to reality, versus grounding, which only checks a claim against the documents you supplied — those documents could themselves be wrong. Automated fact-checking breaks text into discrete claims, retrieves evidence, and judges each one. It’s harder than it sounds: claims are often fuzzy, evidence conflicts, and "true" depends on framing and time. It’s a key defense against misinformation, but not a solved problem.

For product teams

The step that separates "consistent with our docs" from "actually true in the world" — sometimes you need both.

For engineers

Decompose into atomic claims, retrieve external evidence, and classify each; distinct from grounding, which only checks against supplied context.

Related

Read anything AI without the jargon

Look up any term in plain English, or save terms as you read with the free Chrome extension.

Open DecoderAdd to Chrome