Three checks in one upload
Text, slide by slide
All slide text, speaker notes and PDF pages go through our AI text detector. Headers, footers, page numbers, pages repeated by animation steps, and code, commands or formulas (which detectors cannot judge) are left out first, and every slide or page gets its own score.
Pictures
Embedded pictures are checked for AI generation. Logos and backgrounds repeated on every slide are skipped.
AI labels and credentials
We read the AIGC labels Chinese AI services must add (GB 45438-2025), Content Credentials (C2PA) from OpenAI, Adobe and Google, and Stable Diffusion parameters.
Generator traces
Files built by AI assistants with code libraries (python-pptx, PptxGenJS, python-docx, ReportLab) or exported by AI slide services keep recognisable properties.
Questions about document detection
Which files can I check?
PowerPoint (.pptx, .ppsx), PDF and Word (.docx). Old .ppt and .doc files need to be saved as .pptx / .docx first, or exported to PDF. Scanned PDFs without selectable text are checked for pictures and file data only.
Can you tell that a deck was made with ChatGPT, Gamma or Kimi?
Often, yes. When ChatGPT or Claude create a PowerPoint file they usually build it with a code library, which leaves its name in the file properties. AI slide services leave their name or an "AI generated" label, and Chinese services must embed an AIGC label. When a deck was retyped or re-saved by hand these traces can disappear, so the text check matters most.
Why is the text human but a picture AI?
They are checked separately. Many people write their own slides and add AI-generated illustrations, so each picture gets its own score.
Is my file stored?
No. The file is read in memory to extract its text and pictures and is discarded after the check. Signed-in users only keep the score and file name in their history.