One click to sound human again
Spotted a passage that reads like a machine wrote it? Fix it in place. The rewriter does not just flag robotic writing, it rewrites the flagged sentences in your own register without changing what they say.
Free to try, no account neededAI Detector for ChatGPT,
AI Detector for ChatGPT,
Claude and Gemini text
Paste any text and see how much of it reads as machine generated, sentence by sentence, with the measurements behind the score.
0 words
Sentence-level report
Every line scored, not one number for the whole document
Eight measured signals
Rhythm, vocabulary, phrasing and predictability, all shown
Built-in rewriter
Fix the flagged passages without leaving the page
Beyond a percentage
Write smarter, not just safer
Go past the detection score: see what was flagged, understand why, and fix it in the same place.
The reasoning, not just a number
Every score comes with the eight signals behind it: sentence rhythm, model vocabulary, stock phrasing, word predictability and more, each with the actual measurement.
Know what reads badly, fix it fast
Stop guessing which part looks generated. The sentence report marks exactly which lines triggered the score, so you edit the two that matter instead of rewriting the whole thing.
The method
How Inkprint OS detects AI writing
Most detectors hand you a number and keep the reasoning to themselves. Every measurement we take is on the screen next to the score.
Reliability scales with length. Below 60 words we mark the result low confidence instead of guessing; above roughly 300 words the signals are stable.
Signal analysis
We measure predictable word choice, even sentence rhythm and structural repetition. These patterns survive a pass through a paraphraser or a grammar tool, which is where most detection breaks down.
Context validation
Signals are weighed against the document as a whole rather than judged line by line, and the thresholds are set to keep false positives down on formal, academic and non-native writing.
Confidence report
You get a score, a stated confidence level, sentence-level highlights showing exactly what triggered detection, and a plain reading of which parts need attention and which do not.
Who it is for
Detection that holds up in academic and professional work
Check your work before you submit
Worried your assignment reads as AI written even though you wrote it? Run it through first. See which sentences look generated and fix them before anyone else sees them.
- Check any draft for generated text before submission
- See the score and the exact sentences behind it
- Rewrite flagged passages without changing your argument
Read a submission with context
A percentage on its own settles nothing. See which sentences were flagged and why, so a conversation with a student starts from something concrete rather than a number.
- Scan any submission for generated text
- Get a sentence-by-sentence breakdown you can save
- See the measured signals, not a black-box verdict
Publish knowing what you are publishing
Check an article, newsletter or script for generated text before it goes out. Find the sections that read as machine written and put them back into your own voice.
- Check articles, posts and scripts before publishing
- Rewrite marked sections in one click
- Keep a scan history for everything you commissioned
Three steps
How to check a text with Inkprint OS
From paste to a clean rewrite in about a minute.
Paste or upload your text
Paste it straight in, or upload a plain text file. Pro handles up to 15,000 words in a single pass.
Run the scan
Hit Detect AI and the result lands in about a second. No queue, no account needed for your first scans.
Read the report and fix what is flagged
Work through the sentence report to see what triggered detection, rewrite the flagged parts with the rewriter, then scan again to confirm.
A score is where the conversation starts, not where it ends
Inkprint OS exists to help people produce and publish work they can stand behind. No detector is proof of anything on its own, and we will never present one as if it were. Before you act on a result, check the rules your institution or organisation sets for disclosing AI assistance.
The full toolkit
Everything that comes with the account
AI Rewriter
Turn flagged passages back into natural writing without losing your meaning or your tone.
Sentence report
Every sentence scored and highlighted, so you know precisely what to change.
Scan history
Every scan kept on your account with its score, so you can show your work later.
Plagiarism checker
soonCopied and paraphrased source matching, alongside the AI score. In development.
FAQ
Questions people ask first
Still stuck? Write to us.
What is an AI detector?
It is a tool that reads a piece of text and estimates how likely it is that a language model produced it. It does not search a database and it does not look for copied sources. It looks at how the text itself is built: which words were chosen, how the sentences are shaped, and how predictable the whole thing is.
How does it actually work?
Three families of measurement, combined into one score:
Word predictability. Language models pick the likely next word. People wander off it. We measure how much of the text is drawn from the most common vocabulary in the language.
Sentence rhythm. People mix four-word fragments with forty-word sentences without thinking about it. Generated prose settles into a narrow band. This is the heaviest weighted signal in the score.
Linguistic markers. Stock phrasing, bolted-on transitions, expanded contractions and the vocabulary models over-produce, from “delve” to “robust” to “it is important to note”.
No detector is fully accurate, and every one of them is wrong in both directions sometimes.
How accurate is it?
Accuracy depends almost entirely on length. Under 60 words no detector on the market is reliable, and we mark those results as low confidence rather than pretending otherwise. Above roughly 300 words of continuous prose the signals become stable. We publish the measured value of every signal alongside the score precisely so you can judge the result rather than take it on faith.
Will it flag text I wrote myself?
It can. Formal, structured writing scores higher than casual writing, and people writing in a second language are flagged more often than native speakers across every tool in this category. That is a documented limitation of the whole approach, not a quirk of ours. Never use a score on its own to make a decision about a person.
Does it catch text that went through a paraphraser?
Partly. Paraphrasers change vocabulary, which weakens the lexical signals, but they rarely change sentence rhythm, and rhythm carries the most weight in our score. Text that a person genuinely rewrote is a different matter, and it should read as human, because at that point it is.
What is the difference between an AI detector and a plagiarism checker?
A plagiarism checker compares your text against a database of existing sources to find copied or paraphrased material. An AI detector looks only at the internal patterns of the text to estimate whether a model generated it. They answer two different questions, and a document can fail one and pass the other.
Do you keep the text I scan?
Scans made without an account are not stored at all. If you are signed in we keep the score, the word count and a short excerpt so your history is useful, and nothing else. You can delete any scan from your dashboard.
Can I use it without creating an account?
Yes. Anonymous scans run up to 300 words, three times a day, with the full score and signal breakdown. A free account raises that to 1,000 words per scan and 5,000 words a month.