AI text detection in dentistry

This dataset contains the derived data and analysis files used to benchmark six AI-content detectors on a corpus of 120 full-length dentistry manuscripts.

Overview.
This dataset contains the detector scores used in the International Dental Journal submission by Garcovich et al. (2025). It includes per-document outputs (range 0–1) for six AI-content detectors applied to 120 full-length dentistry manuscripts.

File.
Resultados detectores.xlsx — one workbook with 4 sheets (30 documents per sheet):

CONTROLLO = Human-written articles (n=30)

GPT 4.5 = GPT-4.5 generated texts (n=30)

GPT4o = GPT-4o generated texts (n=30)

DEEPSEEK = DeepSeek-R2 generated texts (n=30)

Columns (each sheet).
[ID], Aidetector, Copyleaks, DetectingAI, GPTZero, Originality, Turnitin.
Values are reported exactly as returned by each tool on a 0–1 scale (not percentages). Higher values generally indicate a higher likelihood of “AI-generated” according to each detector’s own scale; no additional thresholding has been applied in this file.

Notes.

No copyrighted full texts are shared; only derived numeric outputs.

Detector versions and run dates are available upon request / will be added in a later version.

See manuscript for methodological details (ROC/AUC, thresholds and analyses).