An AI detector for publishers, editors and magazines

IsThisAiWritten is a free AI detector for publishers, editors and literary magazines. It screens submissions of any length, from a 2,000-word story to a full manuscript, without uploading anyone's unpublished work: paste or open a submission below. Use it for triage: a flag should open a conversation with the author, never close one.

What to expect from a stack of submissions

Like any detector, this one can be wrong in both directions, so treat a result as a reason to look closer, not as proof. A policy that rejects on a flag would turn some real writers away and still let some AI-written work through. A flag alone should never decide a submission, and a score is no substitute for an editor's judgment. Our methodology page explains how it is tested.

Used for triage, the detector helps an editorial team decide which submissions get a closer second read, alongside everything else your readers notice.

How publishers can screen submissions with an AI detector

1. State your AI policy in your submission guidelines, including that you may use a detector.
2. Paste or open each submission in the box above. Chapters are found from headings, so a full manuscript gets a chapter-by-chapter view.
3. Check the invisible-character panel and note anything it found.
4. Set aside any submission that comes back AI or Mixed for a second read, not for rejection.
5. On the second read, look at the passages the sentence map marks highest and judge the work on its merits.
6. If you're still concerned, write to the author and ask about their process before you decide, explaining what you saw.

Hidden characters in submitted text

Zero-width characters and look-alike letters, such as a Cyrillic letter standing in for a Latin one, can turn up in submissions. Some arrive when text is copied out of web pages or other software; some tools insert them. The cleaner makes them visible and lets you copy a cleaned version. Their presence alone proves nothing about who wrote a piece, but it is worth knowing they are there before you run any check or paste the text into your own systems.

Confidentiality for unpublished work

Authors trust you with work that isn't public yet. Because the analysis runs in your browser, manuscripts are never uploaded to us or to a third party, and no copy of the text remains once you close the tab. That matters if your agreements with authors limit who can see a manuscript. Your IT team can confirm the behaviour in the browser's Network tab; the private AI detector page explains what to look for.

Narrative fingerprint and chapter view for fiction

For fiction, the report includes a narrative fingerprint: counts of features such as dialogue share, smell imagery and narrator commentary, shown beside human and AI base rates published in the StoryScope research. It is descriptive and never used in the verdict, but it can give an editor something concrete to discuss. For long manuscripts, the chapter view shows whether one section scores well above the rest. Details are on the long documents page.

Submissions in other languages

The detector reads English only. Submissions in other languages, and translations not yet in English, come back Inconclusive.