Writers and students
Check a draft before submission and decide which passages deserve another read.
What you get back
One reading. Every line marked.
Example reading
It is important to note that effective time management is a crucial skill. By prioritizing tasks, individuals can significantly enhance productivity. My own trick is a kitchen timer and a door that closes. Setting clear goals helps professionals maintain focus.
A reading, with how sure we are
We tell you how sure we are.
High
Plain words
Long text in a calibrated language. The reading says what it sees, for example "Reads as AI-written".
Medium
A lean, not a call
Shorter or mixed text. The reading says "Likely" and the marks show where it comes from.
Low
The reason, shown
Very short text, or a kind of writing we measure weakly. You see why, not only a number.
6 tools, one box
Check it. Then fix it.
AI detectorOne figure with our confidence rating and a mark on every sentence.OpenAI humanizerRewrite flat, machine-sounding text so it reads like a person.OpenGrammar checkerEvery issue named and explained, fixed in one click.OpenParaphraserSay the same thing in different words, in eight modes.OpenSummarizerKey sentences, a paragraph or bullets at the length you set.OpenWord counterWords, characters, sentences and reading time as you type.Open
Measured, not promised
We publish how often we are wrong.
- False flags on human writing, per language
- Detection rate per kind of text
- The date and size of every run
13 languages
Each one fitted on its own writing.
- English
- Français
- Español
- Português
- Deutsch
- Italiano
- Русский
- हिन्दी
- العربية
- Türkçe
- Polski
- Nederlands
- Bahasa Indonesia
Start here
How to check your text
Bring a complete passage from an essay, article or draft you want to review.
Paste your passage
Choose a continuous section of your draft and paste it into the tool.
Select Check text
Run the check, then read the headline under Result. Open Why this reading to explore the context behind the result.
Review the relevant passages
Use Most machine-like sentences to focus your review. Reread those passages alongside the surrounding argument before deciding what to revise.
Purpose and limits
What an AI detector can tell you
An AI detector estimates whether a passage reads more like human or machine writing. Blurwrite uses trained models to assess each sentence in the context of the passage. It interprets those patterns through a curve fitted to human and machine writing in the text's language, including writing by non-native speakers.
It is not a source search or a test of authorship. It does not look for banned words, watermarks or database matches, and it cannot name the model behind a text. Human writing can be flagged and AI writing can be missed.
Method and measurement
How the AI detector reads your writing
Blurwrite examines how predictable the writing is, sentence by sentence. More expected wording points toward AI writing; more surprising choices point toward human writing.
Each supported language has its own calibration curve, fitted on human and machine writing in that language.
Blurwrite publishes measured error rates by language on its accuracy page. It also measures false positives on writing by non-native speakers, giving you evidence relevant to the text you want to review.
Worked example
What the sentence marks look like
It is important to note that effective communication remains a cornerstone of professional success. Furthermore, by prioritizing clarity, individuals can significantly enhance their overall performance. My manager still keeps a paper notebook, and half the team teases her about it. Clear writing tends to save everyone time in the long run. The meeting ran long because nobody wanted to be the one who left first.
Reading your result
What the score, range and confidence mean
Start with “Our confidence,” which is labeled High, Medium or Low. Then read the range and the figure at its midpoint. The figure describes the passage as a whole. A low or middling percentage does not tell you how many sentences AI wrote.
The range reflects how readings vary across overlapping pieces of your text. Longer, even passages tend to have narrower ranges. Short passages or a change in register can widen them, so similar figures can carry different confidence ratings. Very short text gets no reading. Short text receives Low confidence because there is less context to assess.
Finally, review the headline and sentence marks. A “Mixed” headline appears only when the marks point toward both human and AI writing. “Most machine-like sentences” directs you to wording worth reviewing. “Why this reading” adds the length band and an independent model's reading. For English, it can also show whether the text reads as AI-edited or humanized.
Testing and errors
How we measure AI detector accuracy
Accuracy depends on the language, the writing and the test. We measure how much machine writing a test recognizes and how much human writing it flags in error. These rates answer different questions, so a headline accuracy figure cannot replace them. On the accuracy page, we publish dates and sample sizes, separating calibration results from tests on writing kept out of calibration. Those tests include language learners, older model generations, other model families and machine continuations of human openings.
Writing style matters too: in our English tests, argumentative essays are flagged more often than encyclopedia prose. A language goes live only when its false-flag rate meets our publication rule. Results that fail the rule are published and labeled. We also identify weaker areas, including very short text and older model generations. Measurements are repeated when the model or threshold changes. Even with this testing, a reading can be wrong and should not decide a person's authorship.
For your writing workflow
Review essays, submissions and articles
Teachers
Use a check to focus a discussion about an essay. Ask the student to explain their argument, source choices and revision process alongside the submitted text.
Editors and content teams
Review a contributor's article before publishing. Pair the reading with your source checks, editorial notes and questions for the writer.
Different review questions
AI detector or plagiarism checker?
Use an AI detector to review writing patterns and a plagiarism checker to compare wording with published sources.
| AI detector | Plagiarism checker | |
|---|---|---|
| What it compares | Writing patterns against learned patterns in human and machine text, interpreted through the language's calibration curve. | Your text against published sources. |
| What a flag means | The wording reads more like AI writing under the detector's method. | Wording matches a source and needs review in context. |
| What it cannot tell you | Who wrote the text, which model wrote it or whether wording matches a published source. | Whether matching wording was written by AI. |
| What to do next | Review confidence, the range and sentence marks alongside drafts and the writer's explanation. | Review the matched source, quotations and attribution. |
Reviewing a flag
What to do if your writing is flagged
If your own writing is flagged, review the reading and gather context before drawing a conclusion.
Reread the marked text
Check the exact passage you submitted, then read the sentence marks in context. Note which wording points toward AI writing rather than treating the headline as a verdict.
Check a longer passage
If you checked a short excerpt, review a longer passage from the same draft. More context can help interpretation, but the reading may still be similar.
Gather your writing records
Collect drafts, version history, notes and sources that show how you developed the text. Use these records to explain your decisions and the sequence of your work.
Explain your process
Discuss how you wrote the passage, using the range and confidence rating to describe uncertainty. Teachers can use the reading to start a conversation, not to decide authorship.
Your review workflow
Choose how checks are kept and continue editing
Anonymous pasted text
Without an account, the text you check is read for the check and not kept.
History you can delete
Signed in, free or paid, checks stay in your history for a year. You can delete one entry or your whole history at any time.
A separate grammar review
Use Blurwrite's grammar checker to review grammar, spelling and punctuation after your content review. Each issue names its category and offers a fix to apply or ignore.
Regular checking options
Tools for repeat checks and review
Batch checking
Every paid plan supports checking several texts at once. You can paste text or upload a file when preparing writing for review.
History and printable reports
Checks made while signed in stay in history until you delete them. Pro adds a printable report with the reading and sentence marks, kept only in your browser.
API and MCP access
The Pro plan includes API and MCP server access for regular checking. See API access for details about these options.
Pricing
Free to start. Fair to grow.
Free
0 USD
One run a day on each tool, on the first 250 words of a detector check. Grammar and the word counter included.
Sign upBasic
6.99 USD / month
Billed yearly. 50,000 AI detection, 10,000 humanizer and 10,000 paraphraser words a month.
Get BasicPlus
9.99 USD / month
Billed yearly. 150,000 AI detection, 25,000 humanizer and 25,000 paraphraser words a month.
Get PlusPro
12.99 USD / month
Billed yearly. 50,000 humanizer and 50,000 paraphraser words a month, reports to print and the API.
Get ProQuestions
Before you paste.
Is there a free AI detector?
Yes, Blurwrite offers one free check a day on the first 250 words. No card is needed. Start with a complete passage you want to review, run the check, and use the result to guide your next edit. You can explore paid options on the pricing page if you need detection for a regular writing workflow.
Does an AI detector work?
Yes, Blurwrite helps you assess whether a passage reads as AI-written or human-written. Its method examines how predictable the writing is rather than matching a list of words. For an informed review, read the result alongside the draft and its context. Consult the accuracy page for measured performance in the language you're checking.
Can I check text from an AI writing assistant?
You can use Blurwrite to review a passage drafted with an AI writing assistant. For a draft from ChatGPT, Gemini or Claude, focus on the wording you submit rather than the model name.
How can students and teachers use an AI detector?
Students and teachers can use Blurwrite to guide a review of a draft or submitted essay. Students can revisit wording and explain their choices; teachers can discuss the result alongside outlines, sources and revision notes. Keep the conversation focused on the passage and the writing process, with the student's reasoning as part of the review.
Which languages can I check?
Blurwrite reads every language listed on the accuracy page, and each one has its own calibration, fitted using human and machine writing in that language. Choose a passage in the language you want to review and keep its original wording intact when you paste it into the tool.
How much text should I paste?
Paste at least 50 words, and choose a longer passage when you have one. From 50 to 99 words, Blurwrite rates the reading Low and notes that the text is short. Longer passages give it more context. A complete introduction or section from your draft makes a useful starting point.
Is anonymously pasted text stored?
Without an account, the text you check is read for the check and not kept. Signed in, free or paid, checks stay in history for a year, and you can delete one entry or the whole history at any time. Keep that in mind when you choose how to work, especially if you're checking a client draft or unpublished material.
Can an AI detector be wrong?
Yes, an AI detector can flag human writing or miss AI writing. In Blurwrite, read the confidence rating, range and sentence marks alongside the passage. Drafts, notes and version history can add context that a reading of writing patterns cannot supply.
Can it find the AI-written parts of a mixed text?
Blurwrite can mark sentences that read more like AI writing, but it cannot prove which parts AI wrote. Every sentence is read in the context of the passage. A “Mixed” headline appears only when the sentence marks show both human-like and machine-like patterns. The overall figure is not a count of AI-written sentences.
Does it recognize paraphrased or humanized AI text?
Rewritten AI text often still reads as AI because Blurwrite was trained with rewritten and paraphrased text labeled as machine writing. For English, “Why this reading” can also show whether a passage reads as AI-edited or humanized.
Which AI models can it recognize?
Blurwrite reads writing patterns across machine text, but it does not identify or name the model that wrote a passage. It learned from varied human writing and machine writing by current models. Tests also cover other model families and older generations. Performance varies, and older model generations are among the weaker areas described in our accuracy measurements.