The deterministic text-rules engine needs no model and works offline after the first load. For loose layouts, the AI route reuses the on-device model behind Ask this PDF, or your own key under the same consent. Scanned pages are OCR’d inside the run while the allowance covers them. The result is extracted-fields.xlsx and extracted-fields.csv in one ZIP — nothing uploaded, one daily job for the whole batch.
How it works
- Stage the PDFs and open Extract fields → CSV
- Name the fields — one per line, an optional page label after a |
- Extract and download the spreadsheet — one row per file
Questions people ask
Which documents does it suit?
Documents that share a shape: the same fields printed on each one. Invoices, statements, forms, applications, exam results. Documents that differ completely per file are better served by PDF → Excel, which detects whole tables.
Does it need an AI model?
No. Text rules — label matching with date and amount recognition — run with no model at all. The AI engine is an optional route for layouts where labels move around.
Are my files uploaded?
No. Everything runs in your browser tab; there is no server copy to delete, and the whole batch is one of your daily jobs.