When tofu Flags a Possible Duplicate
For AP Bills, AR Invoices, and Direct Expenses, tofu compares the following details across your extractions:
Invoice Date
Total Amount
Contact
When two or more extractions match on all three, a duplicate warning appears on the affected extractions in the Extractions column.
tofu checks for duplicates against every matching extraction, whether it's still in Draft or already Ready, Published, or Exported. If a new upload matches a bill that's already been verified or sent to your accounting software, the warning appears on the new extraction so you can catch it before it causes a double entry. The earlier, already-settled bill won't carry a badge, since there's nothing left to act on there.
If Auto-Ready is on, an extraction with a duplicate warning stays in Draft until you review it.
How to Review a Possible Duplicate
Click the duplicate warning on the flagged extraction.
tofu shows you which extraction it duplicates, so you can compare the two side by side.
If there is more than one possible duplicate, expand the list to view all of them.
How to Resolve a Possible Duplicate
Review the flagged extraction against the extraction it duplicates, then choose the option that fits:
It is a true duplicate. Delete the redundant extraction, or resolve the underlying source documents if the same file was also uploaded twice.
Both are legitimate. Some documents naturally share the same date, amount, and contact, for example a recurring monthly bill from the same supplier. If both extractions are valid, click Not duplicates on the duplicate set banner to dismiss the warning, or continue your review as normal. The warning clears once the extraction is marked ready.
💡 Tip: The duplicate warning does not block marking an extraction Ready or publishing it. It is there to help you double check before data reaches your accounting software.
Bulk Resolving Duplicate Extractions
If you have a batch of extractions flagged as possible duplicates, you don't have to review them one at a time.
Click Review on the duplicate set banner.
Check the box next to the extraction(s) you want to keep.
Click Keep selected, delete rest.
tofu deletes every extraction in that set you didn't select, keeping only the one(s) you chose.
If the extractions in a set aren't duplicates of each other, click Not duplicates instead. tofu dismisses the warning for that set without deleting anything. If a new extraction later matches the same details, tofu flags it again so you can review it.
The duplicate set also updates as you work. If you mark an extraction in set as Ready, it's removed from the set automatically, and the banner closes once there's nothing left to compare.
Unlike duplicated documents, possible duplicate extractions have no "original." They're just a group of extractions that match on date, amount, and contact, so you're choosing which ones to keep rather than resolving an original against a copy.
Duplicate Extractions vs Duplicated Documents
tofu detects duplicates in two ways:
Duplicated documents. The same file is uploaded more than once. These are labeled Duplicated in the Source Documents panel.
Possible duplicate extractions. Different files produce extractions with matching invoice date, total amount, and contact.
📘 Learn more: How to Handle Duplicated Documents in tofu
Best Practices
Review the duplicate warning before marking the extraction Ready, since it is hidden once the extraction is Ready.
When deleting a duplicate, keep the extraction with the better source document or the more complete data.
If duplicates keep appearing from the same channel, check whether the same folder or inbox is connected twice in your upload integrations.