Word-choice pattern
Watermarking schemes work by nudging a model toward some words over other, equally good ones. The check reads the sequence of choices across a passage and scores how far it drifts from what unmarked writing looks like.
AI WATERMARK CHECK
AI watermark checker: see what is in your own text
Some AI models now add an invisible statistical marker to the text they generate. You cannot see it, and nobody asked you whether you wanted it there. ClearScript checks your draft for supported signals, shows you which passages carry them, and leaves the decision with you.
Supported signals and eligible text only. Results vary by model and by draft. A clean result means the check found nothing it can look for, not that nothing is there.
WORKFLOW PREVIEW
The way people approach written communication has changed.
Extracting the token sequence
Scoring every word choice
Matching known watermark patterns
87% match to a watermarked word pattern
Some models pick their words on a hidden pattern. Retyping the sentence or deleting characters leaves it exactly where it is.
Z-SCORE 6.41 / THRESHOLD 4.00
Pattern reads differently. 9% match left.
Rebuilt in your own rhythm, every change yours to approve. No promise about what any other tool will say.
Z-SCORE 0.87 / THRESHOLD 4.00
0 words
Excellent
Rated 4.8/5 by 12,000+ students and professionals
Benchmarked against
What is an AI text watermark?
An AI text watermark is a statistical pattern that some language models deliberately embed in the words they choose, so that software holding the right key can later estimate whether that text came from the model. It is invisible when you read it, and it survives ordinary copying and retyping.
- It is not a hidden character, a tracking pixel or metadata. Stripping formatting or pasting into a plain text editor does not touch it.
- It lives in word choice across a passage, so short snippets carry far less signal than long ones.
- Different providers use different schemes, and most are not public. No tool can check for a scheme it has never seen.
- Editing a passage substantially changes the pattern, so two drafts of the same paragraph can read very differently.
Used by students from 3,500+ universities and institutions
How we check your draft
You get a passage by passage read of your draft, with every score attached to the text it came from.
Give it a paragraph, not a sentence
Signal builds up over length. A paragraph gives the check something to work with, and a single sentence usually does not. Nothing here is certain either: you get a score per passage to reason about, not a verdict.
Your flagged passages, in context
The passages that scored highest, in context, with the score attached. It reads the draft in front of it rather than a permanent record, so editing a passage changes what comes back. Nothing is changed and nothing is sent anywhere until you ask for it, and looking at your own draft is not a commitment to doing anything about it.
Rewrite it in your own voice
ClearScript can rewrite a flagged passage so it carries your rhythm rather than the model's, and it shows you every proposed change side by side. You approve each one. What we will not do is tell you a rewrite makes text undetectable, because nobody can honestly promise that. The final text is yours, and so is the call on whether to use it.
What a watermark check can and cannot tell you
Any tool that skips this section is selling you something. Here is where the limits actually are.
What it can doWhat it cannot do
Coverage
What it can do
Check eligible text against the watermarking signals we support, and tell you how strongly each passage reads.
What it cannot do
Check every model or every scheme. Providers do not publish most of them, and several change without notice. A clean result means we found nothing we can check for, not that nothing is there.
Certainty
What it can do
Give you a score you can reason about, per passage, with the text in front of you.
What it cannot do
Prove that a watermark was present, or prove that one was removed. A score is evidence to weigh, never proof of authorship in either direction.
After a rewrite
What it can do
Rewrite a flagged passage so it carries your rhythm and sentence shape, and show you exactly what changed.
What it cannot do
Guarantee any outcome with any third party. We do not test against private systems and we will not tell you a rewrite makes text undetectable, because nobody can honestly promise that.
Know what is in your writing before you send it
Most of the writing people do with AI help is ordinary and legitimate. A cover letter drafted at midnight, a project update, a first pass at a literature review you then rewrite properly, an email in a language you are still learning. In none of those cases did anyone agree to have a permanent marker embedded in their sentences, and in none of them is the marker relevant to the person reading the finished work.
There is a straightforward argument for being able to see it: you cannot make an informed decision about text you cannot inspect. ClearScript exists to put that inspection in your hands, the same way a grammar checker puts your own mistakes in front of you before anyone else sees them. What you do next is your call, and it should be made with the rules that apply to you in mind.
Built for everyone who writes with AI
Hand in writing that reads like you
- Paste your draft. See which lines look AI-written
- Rewrite them in your voice without rewriting the whole essay
- Keep your argument intact while cleaning up the rhythm
my first draft still sounded like chatgpt, this made it feel like something i would actually hand in
Maya S
Graduate student
Excellent
People come for the score, then stay for the rewrite
See the risky lines, fix the rhythm, leave with a cleaner draft
Rated 4.8/5 by 12,000+ students and professionalsSentence-level proofFinal-pass workflow
5/5
“fixed my draft in 2 minutes”
@maya_s4.5/5
“I thought it would just lower the score, but the draft actually got clearer.”
@jakeT5/5
“i run it before submitting. fast and no drama”
Carlos M4/5
“The sentence view is way more useful than one scary score.”
Prof Chen4.5/5
“Really helpful for cleaning long ChatGPT drafts without changing the point.”
@priya_k5/5
“4 stars because I still review everything, but it saves me a lot of time.”
Dr ParkUse it to review your own work
Do not misrepresent authorship, hide AI use where disclosure is required, or break the rules of your school, your workplace or your platform. Do not use AI where it is prohibited. If your school, your employer or your publisher requires you to disclose AI use, this tool does not change that obligation and is not a way around it. Knowing what is in your draft is exactly what lets you be accurate about it. The final draft is yours, and so is the responsibility for it.
Questions about AI watermarking
Straight answers, including the ones that do not flatter us.
What is an AI watermark in text?
It is a statistical pattern deliberately embedded in a model's word choices, designed so that whoever holds the key can later estimate whether the text came from that model. It is invisible to a reader and it is not stored as metadata, so copying, retyping or changing file format leaves it untouched.
Which AI models add a watermark?
Several major providers have published research on text watermarking and some have said they apply it to certain outputs. What is actually switched on, for which models, and at what strength is not fully public and it changes. We check the signals we support and we say so rather than implying broader coverage.
Can ClearScript detect every watermark?
No. ClearScript checks eligible text for supported signals only. A clean result means the check found nothing it is able to look for, which is not the same as proving nothing is there. Anyone claiming complete coverage of unpublished schemes is guessing.
Does deleting characters or retyping remove a watermark?
No. The pattern lives in which words were chosen, not in the characters themselves, so retyping the same sentences reproduces the same choices. Substantially rewriting a passage is a different matter, because that changes the choices.
Is a watermark the same thing as an AI detector score?
No, and the difference matters. A watermark check looks for a marker a provider deliberately put there. An AI detector estimates, without any key, whether writing looks statistically like machine output. Our AI detector is a separate tool and answers a separate question.
Will a rewrite make my text undetectable?
We will not tell you that, because it is not something anyone can honestly promise. A rewrite changes sentence structure and word choice, which changes what any statistical check sees. It is not a guarantee about any specific third-party system, and you should not treat it as one.
Is my text private?
Your text is processed to return the result you asked for, and your drafts and scores stay in your account. We do not publish them and we do not share them with your school or your employer. The privacy policy sets out retention and how AI improvement works, including how to opt out.
Can teachers or editors use this?
Yes, and plenty do. Seeing which passages carry a signal, with the text in front of you, is more useful for a conversation than a single number would be. Please treat it the way you would any other signal: as something to look into, not as a verdict on a person.
Is ClearScript affiliated with any AI provider?
No. ClearScript is an independent writing tool. We are not affiliated with, endorsed by, or partnered with any AI provider or any detection company, and any product names on this page belong to their owners.
Read your own draft the way software reads it
Run the check, see which passages carry a signal, and decide for yourself what happens next.
ClearScript is an independent writing tool and is not affiliated with, endorsed by, or partnered with any AI model provider or detection service. Checks cover supported signals in eligible text only and results vary. Use AI only where it is permitted, and disclose it wherever disclosure is required. All product names and trademarks belong to their respective owners.