An AI writing detector that tells you what it found

What it actually looks for

Phrases that carry no information. "It's worth noting", "in today's fast-paced world", "in conclusion", "at the end of the day", "cabe destacar", "en conclusión". A phrase made to be generic is almost never something a person wrote on purpose.

Vocabulary that belongs in a press release. "delve", "leverage", "seamless", "robust", "landscape", "holistic", "synergy", "paradigm", "myriad", "sinergia", "holístico".

Structure that gives it away. "Not just X, but Y". "It's not about X, it's about Y". "More than X, it's Y". A rule of three. A rhetorical question followed by a one-word answer. A chain of em dashes. Hedging that never commits. The assistant sign-off.

Rhythm. Three parallel items. Repeated clauses of near-equal length. Uniform paragraph length. Bolded lead-ins on every bullet. Headers that promise one thing and the paragraph delivers another.

Register. Paragraphs that read as if two different people wrote them. False intimacy ("we get it", "sound familiar"). Stacked hedges.

What is missing. Texts with no concrete nouns read as machine-written even when every word is fine. No names, no places, no dates, no quantities, no prices. Humans reach for specifics because specifics are what people remember, and a text with none of them has a different kind of emptiness.

What the score means

The score is not a verdict about a person. It is a measurement of how many of the patterns above the text contains, weighted so that the expensive ones count more.

• 0 to 10. Varied prose, concrete nouns, no scaffolding. This is what most human writing scores.
• 15 to 25. Clean and competent. The only tells are structural.
• 40 to 70. Empty intensifiers, scaffolding or personification are present.
• 85 to 100. Marketing slop with no concrete nouns.

Texts under 20 words are not scored, because there is not enough text to measure.

What it does not do

• It does not tell you who wrote a text. There is no authorship test here, and the page will not pretend otherwise.
• It does not check facts, sources or plagiarism. It works on the writing.
• A low score is not a certificate. It means the patterns this tool looks for were not found in that text. It does not mean the text is good.
• A high score does not mean a machine wrote it. It means the text reads that way.

How to use it

1. Paste the text. Up to 5,000 characters per run on the free tier.
2. Read the score and the breakdown. Hits are split by category, so you can see whether the problem is vocabulary, phrases or shape.
3. Rewrite as human and compare. The rewrite comes back with the score before, the score after and the number of hits removed.

The surface pass is instant. The deep read, which is what finds the structural tells and the missing concrete nouns, takes up to about 40 seconds on dense text.

What the free tier includes

• 10 scores and 10 rewrites per day
• Up to 5,000 characters per run
• No account, no API key, no credit card
• Spanish and English rule packs, plus a French phrase and vocabulary list

Questions people ask

Can it tell me if my essay was written by AI? No, and it does not try. It measures the patterns in the text, which is why you can use it on your own writing. Anyone can paste a text in.

How accurate is it? It is accurate about the patterns it checks for, because it reports each one with the exact phrase that triggered it. It is not a probability that a text came from a model.

Does it work on documents? On any text up to 5,000 characters per run on the free tier, 40,000 on paid plans. Paste a section at a time for a long document.

Which languages? Spanish and English, plus a French phrase and vocabulary list.

Is there an API? Yes, on the Team plan only. The pricing page has the numbers.

Check a text you are unsure about

Take the paragraph you keep rewriting and paste it. Read the score, then read the marks. That is the whole tool.