Run the semantic accuracy gate — procedure

Purpose

Use this page to run the Semantic Accuracy Gate on AI-assisted writing: enforce terminology consistency, classify every non-trivial claim under an explicit evidence boundary, and produce a clean final text that removes/rewrites overclaims.

Use this procedure in AI workflows when:

Enforcement (fail-closed):

Choose a mode

Setup

1) Choose (and state) the evidence boundary:

Verify (smoke test)

1) Run with a draft but omit the evidence boundary.

Options

Option 1 — Artifacts-only

Example

Option 2 — Authoritative sources

Example

Option 3 — External verification allowed

Example

Common mistakes