Does this sound like you,
or like ChatGPT?
Paste anything you wrote — or anything a tool wrote for you. We'll score it, show you exactly which words and sentence patterns are giving it away, and tell you what to fix. No sign-up, and nothing you paste leaves your browser.
Buzzwords cost you more in a cold email than in a blog post — we weight the score accordingly.
40 words minimum for an accurate score
How this works
This isn't a “detector” that claims to know whether an AI wrote your text — nobody can promise that reliably, and we're not going to pretend otherwise. What it actually does is score your text against the same three checks Ozigi runs on its own output before it ships:
Words like leverage, robust, and seamlessly that sound impressive but say nothing specific. One or two are normal. A cluster of them is a tell.
Phrases like “navigate the complexities of” or “in today's fast-paced world” — formulaic transitions that show up constantly in AI output and almost never in text someone typed themselves.
Patterns like “it's not just X, it's Y,” bolded mini-headers in a list, or a paragraph that opens with “Moreover.” These matter more than word choice, so they count three times as much as a single flagged word.
Your score is a density measurement — how much of this shows up relative to how long your text is — not a lie detector. Use it as a second pair of eyes, the way you'd use a grammar checker: it catches patterns, you decide what's actually wrong.
An AI writing checker that tells you which words to change
Most AI detectors hand you a percentage and no reasoning, which is useless for the thing you actually want to do next — fix the draft. This one shows its work: every flagged span is underlined in your own text, and tapping it tells you what the pattern is and why a reader clocks it.
It's the same check Ozigi runs on its own generated output before that output reaches a user. More than 400 banned words, phrases, essay-transition openers, engagement-bait closers, and structural patterns — contrast framing, bold-label listicles, em-dash overuse, three sentences in a row opening on the same word — scored as a density against your word count, then weighted for the kind of writing you picked.
The whole thing runs in your browser. The word lists ship with the page, the scan is plain pattern matching, and there's no API call behind the button — so nothing you paste is uploaded, logged, or saved, and you can check a client's draft or an unreleased launch email without thinking twice about it.
To use it: pick what kind of writing it is, paste at least 40 words, and hit check. Switch the context afterwards to see how the same text scores as a cold email versus a blog post — the reader's patience is different, so the weighting is too.
Frequently asked questions
Is this actually accurate?+
It catches the vocabulary and sentence patterns that make writing read as AI-generated to a human reader. It can't tell you with certainty whether a model wrote something — nothing can, reliably. It's a mirror, not a verdict.
Does a low score mean I definitely used AI?+
No. Plenty of people write "In today's fast-paced world" without a language model's help — usually because they learned to write in a context (corporate comms, agency copy) that trained the same habits into them. The score describes the writing, not its author.
Can I use this on someone else's writing?+
Yes — that's most of how people use it. Paste a competitor's landing page, an old email of yours, a draft someone sent you for review, anything.
Is my text stored anywhere?+
No. The whole check runs in your browser — the word lists and pattern rules ship with the page, so nothing you paste is uploaded, logged, or saved. Close the tab and it's gone.
Why does the same text score differently for a cold email than a blog post?+
Because the reader's patience is different. A cold email gets three seconds and one whiff of template kills it, so buzzwords are weighted 30% heavier there. A blog post is read by someone who already chose to be there, so it's weighted 15% lighter.
Where does the word list come from?+
It's the same banned lexicon Ozigi runs against its own generated output before shipping it to a user — more than 400 words, phrases, openers, and structural patterns, plus a set of widely-documented AI tells added for this page.