ALEXANDRIA, VA โ June 21, 2026 โ Alconost has launched QACAT, a translation quality platform that brings automated deterministic checks, AI analysis, and expert linguist review into one workflow. Teams upload their translations or screenshots and get clear reports back, showing what's wrong, how serious each issue is, and how quality is trending across languages and linguists. Instead of quality signals scattered across separate tools and local machines, everything lives in one place.
"AI made translation cheap. It also made quality invisible. QACAT brings quality back into focus, with humans at the centre and automation around them." โ Dmitry Winicki, creator of QACAT and Head of Localization QA at Alconost
QACAT lets teams pick the level of QA they need for each project, from a quick automated pass to a full linguist review.
QACAT supports a wide range of formats: JSON, XLIFF, MQXLIFF, XLSX, Trados packages, and many more. Teams upload their files or screenshots, then choose between automated, AI-driven, linguist review, or a mix. Reviewers can work directly on product screenshots, with built-in OCR reading the text from the interface automatically and auto-filling it into the correction field.
Every project ends with a structured report which includes:
- severity breakdowns, scored per language and per AI engine
- an AI summary highlighting risk areas and what to fix
- performance trends across languages and linguists
Reports are exportable as XLSX.
The platform is built on a clear idea: automation handles the repetitive, rule-based checks, while humans focus on what only humans can do, like nuance, context, and cultural fit. Most layers keep the linguist central, while the AI-driven layer can run at scale with human validation as an option.
What sets QACAT apart is an engine designed not just to detect errors, but to improve. Each QA run follows a loop: detect, validate, improve, calibrate. When a reviewer accepts or rejects a finding, that decision feeds back into the deterministic rules, the AI prompts, and the model strategy, so quality feedback sharpens the system over time. QACAT also tracks linguist performance across projects and reports, bringing quality signals into one shared view.
QACAT offers four QA layers:
- LQT (Linguist Quality Testing) โ professional review inside the real product UI, using screenshots.
- AQA (Automated QA) โ rule-based checks, no AI involved, safe for NDA-bound content.
- LQA (Language Quality Assurance) โ full human evaluation based on MQM, the industry-standard quality framework.
- AIQA (AI-driven QA) โ AI evaluation as the primary QA pass, with optional human validation.
Two further options extend the platform:
- MT Evaluation โ compare different MT and LLM outputs to see which one fits your content best.
- Custom Workflow โ flexible setup for unique pipelines.
The platform is powered by leading AI models including Claude Sonnet 4.6, Claude Haiku 4.5, Gemini 3.1 Pro preview, GPT 5.2, Mistral Large 2512, DeepSeek V3.2, Qwen 3.7 Max, and Grok 4 , with full model choice inside AIQA. The set of available models is updated continuously, so teams always have access to the latest versions.
Native TMS connectors are live, and a REST API connects QACAT to any CAT tool, CMS, or custom pipeline, with the option to trigger checks automatically when new content lands. Webhooks are on the roadmap.
QACAT is available at qacat.alconost.com with a free one-month trial.
About Alconost
Founded in 2004, Alconost is a leading provider of localization services, offering professional translation, localization testing, and language quality assurance services. The company serves clients across gaming, software, mobile apps, and digital content industries worldwide, with offices in multiple countries and a network of over 3,000 linguists.
Alconost Inc. 700 N Fairfax St, Suite 614-C, Alexandria, Virginia 22314, US +1 (571) 357 3985 ยท info@alconost.com
