Straight answers about what ScribeLens measures, what it costs, and — just as importantly — what a detection result cannot tell you.
Paste text or upload a PDF, DOCX, TXT, or Markdown file. Extraction happens in your browser, and a single scan analyses up to 40,000 words — a full thesis chapter or long research paper in one pass, with no chunking.
You do not need an account to try it: a guest check covers up to 2,000 words per scan and 3,500 words per day. A free account raises that to 35,000 words per month, where one token covers one word scanned.
Every sentence is classified human-like, mixed, or AI-like, and highlighted in the document. The overall verdict is graded rather than a bare percentage: it distinguishes likely human-written, mixed signals, a blend of human and AI writing, substantial AI involvement, and strong AI-like signals.
The headline always agrees with the percentages printed beside it. If most sentences read human, the verdict says so — even when document-level style patterns lean the other way, in which case the report explains the disagreement instead of hiding it.
Reference lists and bibliographies are detected and excluded automatically before scoring, along with in-text citation apparatus, DOI and ISSN lines, copyright and licence text, journal running headers, and declaration sections such as Author Contributions, Funding, Conflicts of Interest, Data Availability, and Ethics statements.
This matters because citation formatting is the most formulaic text in any paper by construction, and scoring it inflates AI-like signals on genuine scholarship. You do not need to prepare the document — paste or upload the whole paper.
No. Detection is probabilistic pattern analysis, not proof of authorship. A high score means the writing shares measurable patterns with AI-generated text. Treat it as supporting evidence for review alongside drafts, revision history, and conversation with the author — never as a standalone verdict.
We do not publish a single accuracy figure, because no one number describes detection honestly: performance varies with text length, subject, writing style, and language. Any tool quoting a headline accuracy percentage is overselling what these systems can do. What we do instead is show the sentence-level evidence behind every verdict so you can judge it yourself.
Formal academic prose shares real surface features with AI writing: consistent structure, measured transitions, and even sentence rhythm. That is a known weakness of detectors generally. ScribeLens weighs sentence-level evidence against document-level style signals, so a paper whose sentences read human is not convicted on register alone, and it excludes bibliographies that would otherwise inflate the score.
No. Reference lists, citations, and publisher boilerplate are detected and set aside automatically. They appear in the report for transparency but are never scored and never affect the verdict.
No. We detect patterns common across modern language models rather than fingerprinting a specific one. Any claim to identify the exact model behind a piece of text should be treated sceptically.
No. ScribeLens analyses writing patterns for AI-like characteristics. It does not compare your text against published sources or a plagiarism database.
Yes. Reports export to PDF and DOCX, and the PDF includes a built-in guide so a third party can interpret it without an account. You can also generate a verification link: the report gets a scan ID that resolves to the authentic stored copy, so a printed PDF can be checked for tampering.
Your analysis is kept in your own browser session so a refresh does not lose it, and it is cleared when the tab closes. A copy is only stored on our side if you explicitly create a verification link, and those expire automatically after 180 days.
Run any text and inspect the sentence-level signals behind the verdict. Free, no signup required.