1. You choose a document
The browser grants this page temporary access to the PDF or image you select. PDF Guard validates the type and size, then reads the bytes into browser memory. It does not learn a permanent local path and does not submit the file to a processing endpoint.
2. The local PDF engine reads it
A modern PDF worker extracts embedded text and renders pages when required. Dynamic code evaluation is disabled. If a page has little readable text and OCR is enabled, the browser paints it to a temporary canvas.
3. Local OCR handles clear scans
The English recognition worker, WebAssembly core and trained data are served from the same website and execute on the device. No scanned page image is sent to a remote OCR or language-model provider.
4. Explainable checks create a report
Local code compares extracted text with your checklist, identifies likely uppercase abbreviations and surfaces common labelled details. The report shows found, unclear and not-detected signals instead of a hidden confidence score.
5. You verify and decide
Review important fields against the original. PDF Guard cannot determine authenticity, correctness, medical meaning or regulatory compliance. Download the local report only if useful and store it according to the document's sensitivity.