AI text detection · Humanize rewriting
Bypass AI does both, and neither is an afterthought. Scan text and images for AI-like patterns and get the full signal breakdown with a stated confidence level — then rewrite whatever reads like a machine, at the intensity you choose.
Two free scans a day. Minimum 300 words, and short text never costs you a scan.
A likelihood score, a confidence level telling you what that score is actually worth, and every signal behind it.
Rewrites the passages that read as machine-generated into natural phrasing, with your meaning left intact.
AI-writing signals · originality check
300 words minimum. Below that we decline the scan rather than guess, and it costs you nothing. Includes an AI-originality check for work you are about to submit.
Generated visuals
JPG, PNG or WebP. The report names what moved the score — texture, edges, lighting, background consistency.
Short clip check — in development
Video scanning is not available yet. When it ships, uploading a file or pasting a link will return the same likelihood and confidence read the other scans give you.
The score is an average of five measurable writing properties. You get to see each one — which is the difference between a verdict you have to accept and a result you can argue with.
Predictable phrasing
How closely each word follows the statistically likely next word. Language models optimise for this; people rarely do.
Sentence uniformity
Variation in sentence length and structure. Human writing is uneven; generated prose tends to settle into a rhythm.
Specificity
Concrete detail versus safe generality. Models hedge toward the general when they have nothing particular to say.
Tone consistency
Whether register holds steady throughout. Unnaturally even tone across a long piece is itself a signal.
Repetition
Recurring constructions and transitions. Certain connective habits show up far more often in generated text.
This report reflects writing-pattern signals and is intended as guidance, not a final judgment of authorship.
Confidence is not the same thing as the AI score. It answers a different question: how much should you trust this particular result? A 91% with low confidence deserves less weight than a 62% with high confidence — and no other detector tells you which one you're holding.
High confidence
“Clear writing patterns found.”
600+ words, a score well away from the middle, at least three core signals pointing the same way, and clean paragraph structure.
Medium confidence
“Some patterns are clear, but results may vary.”
300–599 words, or a score in the 60–74% / 26–40% range, or template-shaped content like email and marketing copy.
Low confidence
“The writing patterns are not clear enough for a strong result.”
A score between 41% and 59%, contradictory signals, broken formatting or OCR noise, or text that is mostly headings and lists.
When confidence is low, the app says so on the result screen rather than presenting the number as though it were settled.
Detection report
AI likelihood
Some patterns are clear, but results may vary.
Pattern distribution
Under 300 words there is not enough signal for a meaningful result, and every detector that returns a number anyway is guessing. Bypass AI blocks the scan instead.
The input shows 184 / 300 words until you clear the threshold, then switches to 326 words · Ready to scan. No ambiguity about whether you can proceed.
Short input is handled as input validation, not a failed scan. Your daily quota is untouched, and so is your subscription allowance.
Short-text false positives are where detectors do the most damage. Declining to answer is the more useful answer.
A detection score is only useful if you can act on it. Humanize rewrites the passages reading as machine-generated into natural phrasing, at the intensity you choose, keeping your meaning intact.
Light
Breaks up the most uniform sentences and varies connectives. Your voice stays almost entirely intact.
Medium
Restructures paragraphs, varies sentence length and replaces the most predictable phrasing. The usual choice.
Strong
A thorough rewrite of structure and word choice. Read it before you use it — the meaning is preserved, the phrasing is not yours.
Humanize is for tightening your own drafts. It is not a way to defeat an academic-integrity policy that applies to you, and using it that way is against our terms.
Scans and rewrites share a single searchable log. Reopen any report, compare a before and after, or delete a record for good.
Humanize rewrite
Just now
Text detection
2h ago
Image detection
Yesterday
Plagiarism check
Monday
Free covers occasional checks. Paid covers a term's worth of them.
$0
Start free
$4.99/ week
Choose weekly
$9.99/ month
Choose monthly
$79.99/ year
Choose yearly
Billed through the App Store or Google Play. Cancel any time in your store account — cancellation takes effect at the end of the current period.
What detectors actually measure, where they break down, and how much a number is really worth.
Full signal breakdown, a stated confidence level, and a rewrite button for when you need one. Two free scans a day to try it.
iOS and Android.
