Questions
Including the ones where the answer is no. If a question you have is not here, ask us and we will add it.
Not reliably enough to base a decision on. They disagree with each other on identical text, they change without notice, and they flag genuine human writing regularly.
The bias is documented: Stanford researchers found detectors flag writing by non-native English speakers far more often than writing by native speakers. Plain, well-structured business prose — exactly what good commercial writing looks like — is also disproportionately flagged.
They can be a rough signal across a large sample. They are not evidence about any single document.
We do not promise that, and we will not build the product around it. Wordcheck is not sold as a detector-bypass tool and must not be resold or described as one.
What it optimises for is writing that is genuinely better: varied sentence rhythm, real specifics, a consistent voice. That tends to score well on detectors as a side effect, because those are the properties detectors look for. It also survives the update after next, which a bypass trick does not.
Do not reply with a competing detector score. Sending one concedes that a detector is the arbiter of your work, and you have then agreed to be judged by a tool you do not control.
Answer with evidence you actually hold: the brief, the research trail, the version history, the specifics only a person could have known, and an offer for the writer to talk it through. There is a full guide to this, including an email you can adapt.
No, not for being AI-assisted. Google's stated position is that it rewards helpful, reliable, people-first content however it was produced.
What its spam policies target is content generated at scale primarily to manipulate rankings. Careful work with a person behind it is not the target; thin, mass-produced pages are.
There is no general UK law forcing you to volunteer it, but a misleading answer when asked is a different matter — and being found out later costs far more than saying so upfront.
The practical move is to decide your position once, write it down, and be able to send it in thirty seconds. The disclosure policy guide has the three honest positions and a template to copy.
Most of it, yes. What makes two writers sound different is largely countable: sentence length variation, contraction rate, punctuation habits, how often specifics appear, how sentences open.
Vocabulary is what people think voice is, and it matters least. A banned-words list will not save you if the rhythm is wrong. The full breakdown is here.
Replace adjectives with rules. "Confident but approachable" changes nobody's writing; "always use contractions, address the reader as you, vary sentence length deliberately" does.
Then split your review in two: conformance to the brief, which anyone can check before it reaches you, and judgement about whether the argument is any good, which is the part that actually needs you.
Eight to ten pieces is usually enough for stable patterns. Fewer than five and you are mostly measuring one writer on one day.
If the samples disagree with each other, that is a useful result in itself — it means the voice is not yet agreed internally, and no amount of measurement settles an editorial argument for you.
It detects register first, so a formal, sourced piece is not penalised for having no contractions the way a blog post would be.
That reweighting is what stops it producing nonsense on legal, technical or academic clients.
Web checking sends short exact phrases from your text to a search provider to look for matches. Deep checking sends the whole document to Copyleaks, which compares it against its own database.
Web checking is included in your monthly allowance. Deep checking costs us money per document, so it runs on credits.
No. It means nothing was found, which is not the same thing.
A check compares your text against sources reachable at the time. It cannot see paywalled work, unpublished work, printed material, or anything a search index has missed.
One credit covers 250 words of deep originality checking. It is a balance you hold rather than a monthly allowance, and credits do not expire while your account is open.
Every plan includes some to start with. The pricing page has the packs and what they work out at per check.
By default it stays on our own server in London. The scoring, the voice profiles and the built-in rewriting all run there, and no third party sees your text for any of it.
Three optional features do send text out, and you are told at the point you use them: AI rewriting goes to an external model provider, web originality checking sends short phrases to a search provider, and deep checking sends the document to Copyleaks. If you would rather nothing left the server, use the analysis and built-in rewriting only.
There is a guide on what to ask any writing tool about this, including ours.
No. Not ours, not anyone else's, on any plan including the free one. We also do not sell it or share it for marketing.
Your content is stored and processed only to provide the service to you, and that licence ends when you delete the content or close your account.
No. Some originality services do this — it is how they build coverage — which means an unpublished document can end up permanently in a third party's index. We do not.
It is worth asking any checker this question explicitly, because it is rarely prominent in the interface.
Yes, both, yourself, immediately. You can download everything we hold on you as a single file, and delete your account and its contents outright, from inside your account.
Neither needs to go through us. Copies persist in backups until those rotate out, which the privacy policy sets out.
No. The free plan gives you the full analysis engine with no card and no trial clock counting down.
You only talk to us about money if you decide you want a paid plan.
Card payment is not switched on yet, so it is arranged directly: email us, we set the plan up on your account, and we invoice you.
That is a deliberate choice rather than an oversight — we would rather do it by hand than run a half-finished checkout that takes your card details.
The feature stops and tells you. Nothing keeps running and quietly bills you more.
Every metered limit on every plan is enforced in the code, and there is a test asserting it.
Pages like this usually carry FAQ structured data so Google shows an expandable box in the results. Google removed that feature from Search on 7 May 2026, so the markup would do nothing.
The other option, Q&A markup, is reserved for community pages where readers post their own answers. Google names “an FAQ page written by the site itself” as an invalid use, so putting it here would be breaking their rules to chase a result we are not entitled to.
This page is written to be read and quoted instead — by people, and by the assistants a growing number of people ask instead of searching.