Document Not Read Correctly
First, Check the Original
Open the preview and try to read the field that doesn't match. If it's blurry, cropped, or has a shadow over it, there's no question — the original is flawed.
Typical Causes and What to Do
| Symptom | Cause | Solution |
|---|---|---|
| Confused values everywhere | Photo taken at an angle | Take a straight-on photo |
| Missing footer | Cropped edge | Photograph the entire document |
| Incorrect numbers | Low resolution, OCR off | Scan at 300 DPI, turn on AI + OCR |
| Swapped supplier and customer | Your IČO (Company Identification Number) missing in settings | Add company identity |
| Wrong document type | Model misidentified | Process again with the correct type |
| Error and label Document Not Recognized | File is not a document or is unreadable | Upload a better original |
When the Original is Fine
- Correct the data in the review. If the type is correct and only some values are off, it's quicker and free.
- Select Process Again and choose the type manually. Automatic selection sometimes misses, especially for documents that are partly one thing and partly another. If the model selected the type, the first new processing is free.
- For repeat suppliers with a non-standard layout, consider a custom schema with field descriptions. The description serves as a guide for the model on where to look.
Send It to Us
Documents that fail are our most valuable feedback. Please send the document identifier; recognition improves from unsuccessful cases.