Invoice Validation Checklist: 12 Checks Before You Pay

Every invoice that lands in your inbox carries a question: is this right? The vendor name, the amount, the VAT, the bank details, the line items. Each field is a potential error or, worse, a potential fraud vector. Between October 2013 and December 2023, the FBI's Internet Crime Complaint Center recorded over 305,000 business email compromise incidents with exposed losses exceeding $55 billion globally. Many of those losses started with a single invoice that nobody checked properly.

A structured invoice validation checklist catches problems before they reach the payment queue. The twelve checks below cover the essentials. For each one, we explain what to look for manually and how the same check works when automated.

1. Vendor identity verification

Manual check: Compare the vendor name, address, and contact details against your approved vendor master list. If the vendor is new, independently verify their business registration and phone number before processing.

Automated: AI extraction pulls the vendor name from the invoice and flags first-time suppliers automatically. Zerentry's AI document processing identifies vendor details from any layout without templates, so unfamiliar names surface immediately rather than slipping through in a stack of familiar ones.

2. Purchase order matching

Manual check: Locate the PO number on the invoice. Open your purchase order system and confirm the PO exists, is still open, and matches the vendor and approximate amount on the invoice.

Automated: Three-way matching systems compare the invoice against the purchase order and the goods receipt simultaneously. When an invoice arrives without a corresponding PO, it gets flagged before anyone approves it. No PO, no payment.

3. Arithmetic validation

Manual check: Multiply each line item's unit price by its quantity. Sum the line items. Check that the subtotal matches. Calculate the expected VAT. Confirm the total equals subtotal plus VAT.

This sounds simple, but it is tedious at volume. One transposed digit, one mistyped decimal, and the error passes through. The average AP team processes invoices at a cost of $15 to $40 each manually, partly because checks like this take time, and partly because errors slip past and require corrections later.

Automated: AI extraction captures every line item, quantity, unit price, subtotal, VAT, and total as structured data. Arithmetic validation runs instantly: if the line items do not multiply out to the stated subtotal, or if the VAT does not equal the subtotal multiplied by the stated rate, the invoice gets flagged. No manual calculator work required.

4. Duplicate invoice detection

Manual check: Search your records for the same invoice number from the same vendor. Check for invoices with the same amount and date from the same supplier, even if the invoice number differs slightly.

Duplicate invoices are the simplest form of invoice fraud. Sometimes intentional, sometimes a genuine re-send because the supplier has not received payment. Either way, paying twice costs real money.

Automated: Systems that extract and store invoice numbers as structured data flag duplicates before they enter the approval queue. What used to be a memory test across hundreds of records becomes a simple database lookup.

5. VAT and tax accuracy

Manual check: Confirm the VAT registration number is present and formatted correctly for the supplier's country. Verify the VAT rate matches the expected rate for that product or service category. Check that the VAT amount actually equals the subtotal multiplied by the stated rate.

Fraudulent invoices often get VAT wrong because the person creating the invoice does not know the correct rate or fabricates a registration number.

Automated: Zerentry's AI extraction pulls subtotal, VAT amount and rate, and total from each document and returns a confidence score for every field. Arithmetic mismatches between subtotal, VAT, and total surface immediately rather than hiding until reconciliation.

6. Bank detail confirmation

Manual check: Compare the payment details on the invoice against the bank details you have on file for this vendor. If anything has changed (new account number, new bank, different country), do not update your records based on the invoice alone. Call the vendor using a phone number you already have on file to confirm.

Bank detail changes are the signature move of business email compromise. An attacker sends an email that appears to come from a known supplier, requesting that future payments go to a “new” account. One unchecked change redirects every future payment.

Automated: Automated systems compare incoming bank details against stored vendor records and flag any deviation. The system cannot make the confirmation call for you, but it ensures you never miss a change.

7. Invoice date and payment terms

Manual check: Confirm the invoice date is reasonable. An invoice dated in the future or more than 90 days in the past is unusual. Check the payment terms match your agreement with the vendor (net 30, net 60, etc.) and calculate the actual due date.

Automated: AI extraction captures dates and payment terms as structured fields. Rules flag invoices with dates outside expected ranges or terms that differ from the vendor's standard agreement.

8. Line-item review

Manual check: Read each line item. Does the description match what was ordered? Are the quantities reasonable? Are unit prices consistent with prior invoices from this vendor?

An invoice for exactly $5,000.00 or £10,000.00 with no line items is unusual. Real invoices for goods and services rarely land on perfectly round numbers. When line items are present but the maths does not add up, that is an even stronger signal.

Automated: AI extraction reads line items contextually, understanding that “Facture N°”, “Invoice #”, and “Rechnungsnummer” all mean the same thing. Zerentry's LLM-based engine reads every invoice with human-level accuracy, extracting vendor, amount, VAT and line items without templates or positional rules.

9. Currency confirmation

Manual check: Verify the invoice currency matches your agreement with the vendor. If the currency has changed, confirm with the vendor before processing. Check that amounts are denominated correctly (not accidentally in the wrong currency).

Automated: AI extraction identifies the currency symbol or code on the document and flags mismatches against the vendor's expected currency.

10. Approval authority check

Manual check: Confirm the invoice amount falls within your approval authority. If it exceeds your threshold, route it to the appropriate approver. Watch for invoices deliberately priced just below your approval threshold, which is a known fraud tactic.

Automated: Rules-based routing sends invoices to the right approver based on amount thresholds, department, or vendor category. Invoices that cluster suspiciously close to threshold amounts get flagged for additional scrutiny.

11. Supporting documentation

Manual check: For services: is there a signed contract or statement of work? For goods: is there a delivery receipt or proof of receipt? Missing documentation is a red flag, especially for first-time vendors.

Automated: Three-way matching catches this at the system level. An invoice without a corresponding purchase order and goods receipt fails the match and gets flagged automatically, as described in the invoice processing automation guide.

12. Anomaly and pattern review

Manual check: Step back and look at the full picture. Has this vendor's invoice volume spiked recently? Are there multiple invoices from different vendors for the same service? Is the total spend with any single vendor trending upward without explanation?

Automated: Automated extraction turns every invoice into structured data, making pattern analysis possible across your entire payable history. Trends that would take hours to spot in a spreadsheet become visible at a glance.

The real cost of skipping checks

Every check you skip is a bet that this particular invoice is fine. Most of the time, you win that bet. But 1 in 20 manual entries contain critical errors that lead to problems downstream: wrong totals, misallocated categories, missed VAT. And the fraudulent invoices are specifically designed to look routine.

The manual version of this checklist takes several minutes per invoice. For a business processing a few hundred invoices per month, that adds up to hours of validation work on top of the data entry itself.

From checklist to system

A checklist is only as reliable as the person running it. On the fortieth invoice of the day, around the time you have been typing for two hours straight, check number seven starts to feel optional. That is when errors get through.

The shift from manual to automated validation is not about removing humans from the process. It is about changing the shape of the work. Instead of running twelve checks on every invoice, you review the handful of invoices that failed one or more checks. Four hours of typing becomes ten minutes of checking.

Zerentry's AI extraction pipeline handles checks 1 through 9 automatically: vendor identification, field extraction, arithmetic validation, duplicate detection, VAT accuracy, and line-item parsing. It syncs the structured data directly to Xero, QuickBooks, or Zoho Books. Every field carries a confidence score, so your attention goes to the uncertain values rather than everything.

The remaining checks — approval routing, supporting documentation, and pattern analysis — are process controls that sit on top of the extracted data. With clean, structured invoice data flowing in automatically, building those controls becomes straightforward rather than aspirational.

Start with the checklist. Run it manually for a month if you need to see where the problems are. Then automate the data entry and let the system run the checks for you.

Invoice validation checklist FAQ

How long does manual invoice validation take?

Running all twelve checks manually takes several minutes per invoice. For a business processing 200 to 500 invoices per month, that represents a significant time investment on top of the data entry itself.

Which checks catch the most fraud?

Vendor verification (check 1), duplicate detection (check 4), bank detail confirmation (check 6), and approval threshold review (check 10) are the four most effective fraud controls. These map directly to the most common invoice fraud red flags.

Can AI automate the entire checklist?

AI extraction automates the data-dependent checks: vendor identification, arithmetic, duplicates, VAT, dates, line items, and currency. Approval routing and documentation checks require process rules built on top of the extracted data. The combination of AI extraction and rules-based workflows covers all twelve.

What is three-way matching?

Three-way matching compares the invoice against the purchase order and the goods receipt. All three documents must agree on vendor, quantities, and amounts before the invoice is approved for payment. It is the single most effective control against paying for goods or services you never received.