Skip to content
Academic AI detection

Academic AI Detector for Research Papers

Check your research paper or manuscript for patterns associated with AI-written text. Paste text or upload a PDF or DOCX, review sentence-level classifications, and download a highlighted report. Results support review and do not prove authorship.

  • Paste or PDF/DOCX ≤ 20 MB
  • Per-sentence probability
  • Free plan included
app/ai-tools/detector
manuscript-draft.pdf
Download PDF

1. Introduction

The rapid advancement of artificial intelligence has fundamentally transformed the landscape of academic research. Furthermore, the integration of machine learning methodologies facilitates unprecedented analytical capabilities. Our lab measured the effect across three cohorts between 2021 and 2023.

This paradigm shift necessitates a comprehensive reevaluation of established frameworks.

2. Method

We recruited 214 undergraduate students across three cohorts and collected their written assignments over two semesters. Each submission was anonymised before analysis, and two raters coded every essay independently. Moreover, the proposed framework leverages state-of-the-art techniques to ensure robust and reliable outcomes.

3. Results

Scores in the second cohort rose by a third between the first and the last assignment, while the other two cohorts stayed close to their starting point. We return to this difference in the discussion, where we compare the three groups' writing habits week by week.

Table 2 lists the mean score of every assignment by cohort, with the number of essays behind each value. Missing essays were left out rather than imputed, so the later columns rest on slightly fewer submissions than the first.

overviewsentences

Detection Groups

1

AI-generated

788 words

61.4%

Writing that carries the signature of an AI model.

2

AI-paraphrased

234 words

18.2%

AI text revised by a person or a tool. Not included in the AI score.

The AI score counts AI-generated text only.

four-stage pipeline

Extract
Detect
Similarity
Report
3-class
Human · AI-generated · AI-paraphrased
≤ 20 MB
PDF & DOCX, or paste raw text
Per sentence
Individual probability on every line
Background
Long documents run as tracked jobs

// sentence-level evidence

See which passages need a closer look

Review individual sentence classifications instead of relying only on the overall score.

AI-generated

The model identifies patterns associated with generated text.

AI-paraphrased

The model identifies patterns associated with AI-paraphrased text.

Human

The model classifies the sentence as human-written; this is still an estimate.

AI-generated3AI-paraphrased2

Click a sentence to copy it and find it in the PDF.

AI-paraphrased97% · 12 words

It is imperative to acknowledge the multifaceted implications of these emergent technologies.

1. Introduction

The rapid advancement of artificial intelligence has fundamentally transformed the landscape of academic research. Furthermore, the integration of machine learning methodologies facilitates unprecedented analytical capabilities. Our lab measured the effect across three cohorts between 2021 and 2023.

This paradigm shift necessitates a comprehensive reevaluation of established frameworks.

2. Method

We recruited 214 undergraduate students across three cohorts and collected their written assignments over two semesters. Each submission was anonymised before analysis, and two raters coded every essay independently. Moreover, the proposed framework leverages state-of-the-art techniques to ensure robust and reliable outcomes.

3. Results

Scores in the second cohort rose by a third between the first and the last assignment, while the other two cohorts stayed close to their starting point. We return to this difference in the discussion, where we compare the three groups' writing habits week by week.

Table 2 lists the mean score of every assignment by cohort, with the number of essays behind each value. Missing essays were left out rather than imputed, so the later columns rest on slightly fewer submissions than the first.

// the AI score

How the AI score is calculated

The AI score is the share of your document's words that the model classifies as AI-generated. It summarises the text for review; it does not establish who wrote the document.

Example · 5,000-word manuscript

4,500 words counted500 reference words · left out
  • AI-generated675
  • AI-paraphrased270
  • Other analysed text3,150
  • Not analysed405
AI score=675 AI-generated words÷4,500 words=15%AI-paraphrased is reported on its own: 270 ÷ 4,500 = 6%
Counted in the score

AI-generated

Sentences the model classifies as AI-generated. Their words, as a share of the document, are the AI score.

Shown separately

AI-paraphrased

AI text revised by a person or a tool. It gets its own percentage and is not included in the AI score.

Not scored

Not analysed

Headings, tables, captions and equations are not classified and count as not AI. The reference list is left out of the total.

There are no pass or fail bands: the score's colour moves gradually from green to red as the share grows.

// how it runs

From your document to a report

Upload a supported document or paste your text, then follow the analysis and review the report.

  1. 01

    Extract

    We extract supported text and split it into sentences.

    manuscript-draft.pdf

    1.80 MB • 1,284 words

  2. 02

    Detect

    The model classifies and scores those sentences.

  3. 03

    Similarity

    When enabled, the analysis looks for overlap with published sources. See Similarity Check.

    Primary Sources

    • 1

      Learning analytics and AI-assisted writing (OpenAlex 2023)

      8.4% match — ~108 words

    • 2

      AI in higher education: a systematic review (Crossref 2022)

      6.1% match — ~78 words

  4. 04

    Report

    Review the result and download the highlighted report.

    manuscript-draft - AI Report.pdf

    Highlighted PDF

// FAQ

Academic AI Detection: Common Questions

How to review a research manuscript, interpret sentence-level results and understand the tool’s limits before submission.

  • The model reviews each sentence and estimates whether it resembles human-written, AI-generated or AI-paraphrased text. Those results contribute to the document score.

// start detecting

Check a draft with the AI detector

Start with a document you want to review and inspect the sentence-level results.