Localization Engine Suggestions now come from the corrections human reviewers make, not only from low AI-review scores.

When a reviewer edits a translation whether in-house or through an external provider the localization engine compares the change against what it first produced. When the same fix recurs across jobs, or a correction clearly sets a term, it proposes the glossary, instruction, or brand-voice edit behind it. Apply writes the change into the engine, and the next translation runs with the fix in place.

Suggestions can now be approved or dismissed in bulk select a set and act on all of them at once, instead of one card at a time.

Full changelog: lingo.dev/changelog

AI reviews got more accurate: false positives dropped by 59.7%.

A false positive is a review that marks a correct translation as failing fewer of them means fewer alerts to check by hand, and review scores that can be trusted. We measured the drop on our internal benchmark. The biggest change: when a review doesn't apply to a given string, it no longer counts against the score, so scores reflect only the reviews that genuinely apply.

The reviewer setup form now shows example instructions as a placeholder for each review type.

Also this week:

lingo push --estimate reports what a translation run will cost before it runs.

The CLI uploads the source bytes so the server plans the exact delta only the keys and locales that are new or changed against what's already localized then prints the estimate and exits without translating anything.

Run it before a large or first-time translation to size the job, then run lingo push to translate for real. The number is an estimate, not a quote: it reflects the planned delta at the time it runs.

The async CLI now handles more file formats end to end: TypeScript locale modules, YAML, Gettext PO, Flutter ARB, Android strings, and Xcode .strings, String Catalogs (.xcstrings), and .stringsdict. Point a files[] entry at the source file and every lingo push runs each one through your engine's full pipeline. Translatable text is translated; everything structural is preserved keys, IDs, metadata, placeholders, and code.

Full changelog: https://lingo.dev/en/changelog

Two new built-in AI Reviewers ship: a glossary reviewer and an instructions reviewer. Toggle on the glossary reviewer and every configured glossary term is automatically checked on every translation. Toggle on the instructions reviewer and every configured instruction is checked the same way. Open any translation in the dashboard log and see pass/fail per term, pass/fail per instruction, with the reviewer's reasoning on failure. One toggle replaces authoring an AI Reviewer for every single criterion.

Also shipped

• Per-engine AI Reviewer assignment. Attach different reviewers to different engines instead of one set for the org.

• Organization-level timezone setting. Reports display in the timezone you configure for your org.

Full changelog: lingo.dev/changelog

Audit logs are now available an append-only record of every state-changing action in your organization.

Each change to engines, glossaries, instructions, brand voices, AI reviewers, API keys, members, and roles is captured with the actor, action, target, and request context. The log lives in the dashboard and filters by user, action category, target, and date so you can see exactly who created, updated, or removed a glossary item three weeks ago, even if the actor is no longer on the team.

Also this week:

The new Lingo.dev CLI now runs the localization pipeline pre-edit, human review, AI review, rephrase, back-translation, each toggled on the engine.

Toggle stages on the engine's Pipeline tab, or override per job through pipelineConfig. Pre-edit cleans typos in the source before they reach any locale. The core localization step applies your glossary, brand voice, and instructions. Human review pauses the run until your team or an external provider signs off. AI review reconciles a reviewer's edits back to engine config. Rephrase rewrites for marketing copy. Back-translation flags semantic drift.

A run that includes a human-review stage can wait days before it lands, so the CLI splits a job into lingo push and lingo pull. pull re-attaches to the most recent push and downloads outputs from any machine, including after the original terminal was closed.

The localization pipeline wraps the core translate step with optional stages, each independently toggleable per engine and overridable per job:

• Pre-localization AI edit. An AI agent cleans the source payload before translation, so a single source error doesn't propagate across every target locale.

• Post-localization human edit. Sends the translation to a qualified human translator. The job pauses on a webhook until the edit returns.

• Post-localization AI review. Reconciles the human output against the engine's glossary, brand voice, and instructions.

• Back-translation check. Translates the final output back into the source locale and compares. The agent flags semantic drift by severity and auto-adjusts on major or critical drift.

Turn on what you need, leave the rest off.

The Lingo.dev GitHub App runs continuous localization on any repository, straight from git.

Install it from your organization settings, commit a .lingo/config.json that points at your localization engine, and every push translates changed source files through your engine and opens a pull request with the localized files. Each run executes on the async localization pipeline, so translations pass through the stages configured on your engine pre-edit, AI review, human review, back-translation. It localizes JSON, JSONC, Markdown, MDX, Markdoc, and OpenAPI files, and you can scope runs from a pull request with /lingo translate.

Also this week:

API keys now come in two types: Personal and Service.

Personal keys behave like before they inherit their owner's role and per-engine grants. Service keys carry their own role and per-engine scope, so a production worker isn't tied to any one teammate. A service key can be roleless and reach only the engines listed on it.

Also this week: