What's Wrong? · Automatic QA

Every delivery checked, in under a minute.

Drop a bilingual file, or let it run inside your TMS. Deterministic rules catch the certain errors instantly, an AI judge re-reads only the risky segments, and your master file is never touched.

Two layers, one verdict

Rules first. The model only where it earns its keep.

Most QA cost comes from sending everything to a model. We do the opposite: the deterministic engine settles the certain errors for free, and the AI judge reads only what the rules flag as risky.

Layer 1: the rule engine

28 deterministic checks: tags and placeholders, numbers and codes, protected brand names, Do-Not-Translate, URL localization, punctuation, spacing, length, consistency, and RTL script. Instant, exact, zero AI cost.

Layer 2: the AI judge

Re-reads only flagged segments for what rules cannot see: reversed meaning, dropped content, wrong register. It writes a suggested fix with its evidence, and it never edits your file.

A human closes the loop

Every finding is ignore-or-fix reviewable, segment by segment, with keyboard-speed triage. The report goes to your client only after a person has made the call.

The report

Every issue, with the evidence.

Findings arrive with the rule that caught them, the exact span highlighted in source and target, the surrounding segments for context, and a suggested fix you can paste straight into your CAT tool.

  • Profiles per account and language pair

    Brand locks, DNT lists, custom rules in plain language, and custom checklist import. Nothing hardcoded to one client.

  • Runs where the work happens

    Inside Crowdin, from a watched Dropbox folder, through the API, or a straight file drop. Exports as HTML, CSV, JSON, or standards-based JSON.

See it on your own content

Run it on a real delivery.

Tell us your stack and your language pairs. We will set you up with a workspace and a first verdict in under a minute.