AEO Content AI: free AI content detector AEO Content AI: free AI content detector
Improve my score

Cheap AI content is easy to spot.
Find out how yours reads.

Paste text, a page address or a file. See who wrote it, how it reads, and whether AI answers like ChatGPT will cite it.

Score your writing

Examples:

Free · No sign-up · Nothing you paste is stored

0 / 30,000 characters

26,577 real AI-visibility audits run by the same engine that scores this page.

  • Shopify
  • HubSpot
  • Figma
  • Stripe
  • Slack
  • Airbnb
  • Semrush
  • Zapier
  • Webflow
  • Intercom
  • Dropbox
  • Zendesk

  • Marked in your own text

    Every stock phrase is highlighted where it sits, not listed somewhere below. You see the sentence you have to rewrite, not a word you have to hunt for.

  • Seven signals, with their weights

    Sentence rhythm, stock phrases, point of view, short-sentence mix, active voice, facts versus interpretation, vocabulary range. Every weight is on the page.

  • Same text, same score

    No model runs anywhere in it, so nothing drifts between two runs. Edit a paragraph, check again, and the difference in the number is the difference you made.

  • Text, a page address or a file

    Paste up to 30,000 characters, drop in any public URL, or upload a .txt, .md or .csv. Whichever you give it, the measurement is the same.

  • Nothing you paste is stored

    The text is measured and dropped. No account, no sign-up, no copy of your draft sitting on a server waiting to be leaked.

  • A result you can send

    One link carries the score, the verdict and the full breakdown - useful for settling an argument with a writer or a client without a screenshot.

What makes this different

Reading human is half the job.

Writing can read completely human and still be invisible to ChatGPT. So every check answers both questions: how human the writing reads, and whether an answer engine can lift and cite it. Two numbers, never blended into one.

Human voice →

Human, but invisible

A real voice - but the structure gives AI answer engines nothing they can lift and cite.

Cited & trusted

Reads human and gives answer engines clean, citable material. This is the quadrant to stay in.

Cheap AI slop

Reads machine-generic AND gives answer engines nothing to cite. Readers bounce, engines skip it.

Optimized, but soulless

Well-structured for machines, but it reads like nobody wrote it. Readers discount it, and increasingly so do engines.

Answer-engine readiness →
  • The vertical axis is how distinctly human the writing reads, across seven weighted components including a stock-phrase lexicon.
  • The horizontal axis is whether an answer engine can lift and cite the page. Every check scores it too, with AEO Rank, the engine behind our site audits.
  • The score is deterministic the same input always returns the same score, so a re-check after an edit is directly comparable.

How it was trained

Trained to know machine writing. Tested on writing it never saw.

The who-wrote-it call comes from a model trained for this one job. It learned from both sides, human and machine, and it had to prove itself on writing it had never seen before it was allowed near yours.

9,963 documents written by people, kept out of training and used to test the model before release.

  • Human writing from before ChatGPT

    News stories, forum answers, company blogs and public agency pages, every one written by a person before AI text was everywhere.

  • Machine writing from many models

    Text from ChatGPT, GPT-4, Llama 2 and ten open models, plus AI drafts that were rewritten to sound more human.

  • Tested across 17 kinds of writing

    Before release it read thousands of documents people wrote that it had never seen. The bar for calling AI is set so high that almost none of them reached it.

Methodology

Everything the detector measures, and what each part is worth

The score is a deterministic measurement, which is why the same input always returns the same number - and why we can show you the weights instead of a badge. The who-wrote-it call is the one place a model reads the text, and its mechanism is spelled out too.

Every one of these measures the text, not the author. The second reading - who wrote it - is a separate call from a trained model, described below, and the two are never blended.

Who wrote it

Was this written by AI, by a person - or is it not clear?

one call, three answers

  1. Passages

    The text is cut into passages of about 300 words; short tails are dropped. Each passage is read on its own.

  2. The classifier

    A model trained on our own AI output and on writing we know people wrote gives every passage a reading, from clearly human to unmistakably machine-written.

  3. The call

    The document is judged by its most machine-like passage. Written by AI only when one passage is unmistakable - a bar set so high that, on thousands of documents we know people wrote, almost none reached it. Written by a person when no passage even leans AI. Otherwise: not called.

  4. The second model

    A larger language model reads the same passages and reports how predictable the writing is - raw model output is more predictable than human writing, humanized output less - plus the sentences that pull the reading either way. Shown as evidence, never folded into the call.

Treat “not called” as exactly that: inconclusive, not a soft accusation. Paraphrased AI can still land there, and a person can write copy generic enough to lean AI without ever being called it.

AI Content Detector

Does this writing carry the marks of machine prose?

7 weighted components

  • Sentence rhythm 22%

    Whether sentence lengths vary the way natural writing does, or settle into one beat.

  • Stock phrases 20%

    A lexicon of roughly fifty tells that turn up far more in machine prose than in human drafts.

  • Point of view 14%

    Whether anyone is home - first person, direct address, a stated opinion.

  • Short-sentence mix 12%

    Short, punchy sentences among the longer ones.

  • Active voice 12%

    Whether subjects act, or everything happens to everybody.

  • Facts vs. interpretation 10%

    Whether the writing ever says what the facts mean.

  • Vocabulary range 10%

    Words chosen, versus the same handful cycling round.

Every one of these measures the text, not the author. The second reading - who wrote it - is a separate call from a trained model, described below, and the two are never blended.

Questions

Common questions

Does this tell me whether AI wrote the text?

It gives two separate readings and never blends them. How it reads is a deterministic measurement of the writing - sentence rhythm, point of view, vocabulary, stock phrases. Who wrote it is a trained model that makes one of three calls: written by AI, written by a person, or not called. It says AI only when a passage is unmistakably machine-written; when nothing is, it says so instead of guessing. A person can write generic copy and a good AI can write with voice - both readings can be true at once.

How does the "who wrote it" call work, and how sure is it?

The text is cut into passages of about 300 words and each one is read by a classifier we trained on our own AI output and on writing we know people wrote. The document is judged by its most machine-like passage, because long AI drafts often have a machine-written opening and a body that reads human. The bar for an AI call is set high on purpose - on thousands of documents we know people wrote, almost none reached it - and there is a second model behind it that reports how predictable the writing is and which sentences pull the reading either way. Treat "not called" as exactly that: inconclusive, not a soft accusation. Paraphrased AI can still land there.

What does the score actually mean?

Higher is better writing. Seven measurements of the prose itself are weighted into one number: sentence rhythm counts most at 22 percent, stock phrases 20, point of view 14, short-sentence mix and active voice 12 each, facts versus interpretation and vocabulary range 10 each. Open the breakdown in any result to see all seven. The score says nothing about who wrote the text - that is the other reading.

Is it deterministic?

The style score is: no model runs in it, and the same text always returns the same number, which is what makes a re-check after an edit worth anything. The who-wrote-it call comes from a model and is stable for the same text, but it is a call, not a measurement - which is why it is shown as words, never as a percentage.

How much text do I need?

At least 150 words for a reading you can rely on. Below that the ratios swing on a single sentence, and the result says so rather than pretending otherwise. The who-wrote-it call needs at least one full passage of about 300 words to be worth much; on a short text it will more often say "not called".

What languages does it handle?

English only. Every word list and pattern in the style engine is English, and the origin model was trained on English, so a text in another language gets flagged instead of scored.

Do you keep what I paste?

No. The text is sent to the scoring endpoint and to the model that reads it, measured, and never written anywhere. We keep only counts - how many checks, what the calls were, how long they took - never the text.

The people behind this engine

You just saw what is wrong. This is who fixes it.

AEO Content is a content engine for AI search. The same scoring you just ran is what we build against, on every page we write for a client.

  • We find what only you know

    An interview first, not a brief. Your numbers, your cases, the things a model cannot look up - that is what makes a page worth quoting.

  • We write it and publish it

    Long-form articles built to be lifted by answer engines, straight into your CMS. No drafts left sitting in a folder.

  • We watch who quotes you

    ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews, tracked over time - then we fix whatever they still skip.

See how AEO Content works Audit my whole site first

Free audit, no card, no sign-up.