Limitations of the AI document reader

What the AI can’t do yet and what to check manually

SafeKeep’s AI document reader is a powerful tool, but it’s important to know where it has limits. Understanding what the AI can and can’t do helps ensure your information stays accurate and useful.

It only works with selected document types

AI extraction only runs when you choose a predefined document type during upload. This is because SafeKeep has mapped specific fields for each supported type. If you upload a custom or unclassified document, AI will not run.

It can sometimes make mistakes

While our AI partner Google Gemini is highly advanced, it’s not perfect. Sometimes fields may be left blank, misread, or slightly incorrect. You should always double-check the fields after auto-fill runs and make changes where needed.

Look for the AI icon next to any field that was filled by AI. You can edit or delete these fields at any time.

It doesn’t extract from ZIP files

If you upload a ZIP file, the AI won’t be able to read or extract its contents. We recommend uploading supported files such as PDFs, JPEGs, or PNGs directly to enable auto-fill.

It only fills mapped fields

AI won’t guess or create custom information. It only extracts details that have been predefined by the SafeKeep team for that document type. If you want to add extra information, you can do so manually using custom fields.

It doesn’t handle handwriting or poor-quality scans

If a document is hard to read, handwritten, blurry, or cropped, the AI may not extract the correct information. For best results, always upload clear, high-quality scans.

Summary

The AI document reader is designed to speed up your life admin, but it’s still important to review and update fields where needed. For now, AI helps with supported document types only and cannot work with unstructured uploads or custom documents. We’re working every day to improve these capabilities and expand what’s possible.