What GenPolish does, and what it refuses to do

The list of what we change is public

The engine sorts every edit into nine change categories, and each one is defined in writing. Seven are found by matching a turn of phrase. Two — repeated shapes and the reflex to turn prose into a list — can only be seen in the shape of the text, so they are named without being rewritten.

Four things are preserved by default: numbers, URLs, citations and Markdown. The engine does not rewrite an argument.

There is no proprietary score you cannot inspect. A score you cannot argue with is a score you cannot check, and the taxonomy is published for exactly that reason.

We do not sell detector evasion

Rewriting text does not prove authorship, and it does not guarantee any detector result. We do not sell a way around detectors, and we do not claim one.

The register score describes how machine-like a text reads against our published taxonomy. That is all it describes. It is not a probability that a machine wrote something, and it should not be used as one.

What a scanner can see, and what it cannot

Our tools recognise eight invisible characters. They are real, and they do turn up in text that has passed through a model, an editor, a PDF or a web page. Finding one points to a text-handling step, not to an author.

A watermark is a different thing. It is a statistical bias applied while a model picks its words, and no scanner can read it from the text alone. We do not claim to detect or remove one.

The watermark database records what each of five providers actually does, what the evidence is, and when it was last checked.

Five tools that never send your text

Five utilities inspect, clean, compare or count text entirely in your browser. They need no account, and they are excluded from measurement even if you accept analytics elsewhere.

The AI rewriter is the exception, and it is explicit: it sends a draft to the server when you ask for a rewrite, and only then.

Every claim here carries its source

Each provider record cites primary sources and carries the date it was last checked. Each article cites what it relies on, in links that are followed rather than devalued.

Four tests make that policy enforceable rather than aspirational. A character named in the prose must exist in the engine. A claim that a provider marks its text must carry a primary source. A date must be a real one. And the prose may not contain a first-party research claim, because we have not run one.

That last test exists because of a mistake. Text ported from the original mockup asserted a study of forty thousand drafts that never happened, along with several figures drawn from it. All of it has been removed, from the articles, from the cards, from the home page and from the engine.

Nobody is named on this site

The articles are signed by the organisation, and that is deliberate. The ported mockup signed them with a person who does not exist; declaring that name as an author in structured data would have asserted her existence to a search engine.

The same name appeared under a testimonial on the home page until recently. It has gone for the same reason. We would rather have no social proof than invented social proof.

What is not connected yet

Several parts of the product are previews. Saved documents, document history and usage records are not stored. API key management issues no usable key. Paid offers are configured separately and are off unless explicitly enabled, and the rewriting provider is off by default too.

We list these because a page called About is where a reader decides whether to trust the rest. A preview described as a feature is the same kind of claim as an invented customer.

Who publishes this

GenPolish is published by Bahri Labs, a société par actions simplifiée registered in France on 28 August 2026. Its registered office is at 47 rue Vivienne, 75002 Paris.

Registration number 109 283 416, R.C.S. Paris. VAT number FR57109283416.

Two details are missing from that list, and their absence is not an oversight: the share capital and the name of the publication director are not yet available from the national register. They will be added here rather than estimated.