Anglicism checker for text and web pages

An anglicism is a word borrowed from English that has a perfectly good native equivalent in the target language. We flag well-attested loans documented in major reference works, and we always offer 1–3 native alternatives so you can decide. Naturalised borrowings (e.g. Russian “biznes”, German “Computer”) are penalised much less than bare anglicisms — and not at all in lax mode.

In Link mode the server fetches the page once for you (needed because of CORS restrictions). Tokenization, anglicism matching and scoring run in your browser via WebAssembly.

The dictionaries start from Wiktionary's lists of English borrowings and are curated by hand. Each entry carries a severity, category, naturalization flag and a list of acceptable domains, so you can tune the result to your context. Many entries name a reference work where the usage can be checked (Gramota.ru, Duden, OQLF, RAE and others). You can ignore any word per page or globally.

Domain awareness. Each anglicism declares the domains in which it is acceptable. In IT context “server” is treated as standard vocabulary; in marketing or general writing the same word may be flagged. Switch the Domain selector to test how a word behaves across contexts.

The share of your text that consists of loanwords the tool recognizes, weighted by how established each one is — and the second half is what makes the number usable. A borrowing that has been in the language for a century and has no native equivalent is not the same problem as one that arrived last year and displaced a word everyone knows, so counting them equally would produce a figure that punishes ordinary prose. The strictness setting moves that line: at the loosest, only recent and avoidable borrowings count; at the strictest, anything with an English origin does. Change it and the score changes, which is the point — a single number without a stated threshold means nothing, and here you can see which threshold you are being judged against.

For now the tool ships with curated lists per language. We will open a community submission queue as the tool moves out of beta. In the meantime, the per-word ignore list and per-domain whitelist mean you can adapt the tool to your exact needs.