Terminology

Some markets require specific wording. A word that is fine in one country can be forbidden by compliance in another, and a brand may standardize on one term across all its content. Terminology lets you define those rules once per language, and i1n enforces them on every AI translation.

Rules apply everywhere translations happen: the dashboard, i1n push --translate from the CLI, and MCP / AI-agent flows. Available on Business and Enterprise plans.

Owners and admins configure rules in the dashboard under Settings, AI Context, Terminology. Each language gets term pairs (a term to avoid and what to use instead, with optional variants for inflected forms) plus free-text instructions for rules that are not a 1:1 word swap.

A rule defined for a base language code (for example "es") applies to every regional variant (es_ar, es_mx, es_co). A rule for a specific locale applies only there.

The AI receives your rules with every translation into that language, scoped so other languages keep their natural wording. Every result is then double-checked: if a forbidden term appears anyway, the translation is applied but flagged with a Terminology badge and lands in Needs Review, where the sidebar counter turns red until it is resolved.

Hand-written text is covered too. If someone types a forbidden term (dashboard edit, new key, or CLI push), it is never rewritten automatically, but it gets the same flag so nothing ships unreviewed. Values a human verified are not re-flagged unless they change.

Languages with active rules use their own private translation cache, so an old translation created before the rules can never come back with the wrong term. Fresh results are cached under the current rule set.

When rules change, use Retranslate to rebuild the whole language in one click with the current rules. See the Retranslate section in AI Translation for details.