Check fraud detection combines check-image review, duplicate detection, account activity analysis, and, where available, verification against the issuer’s records. Banks and credit unions use these signals to identify suspicious items, prioritize investigations, and decide what action to take within their operational and legal requirements.

A check can contain a real account number and still be fraudulent. Criminals may alter a stolen check, manufacture a counterfeit using its details, or deposit an unchanged check with a forged endorsement. The FBI and U.S. Postal Inspection Service describe all three in their mail theft-related check fraud advisory.

For fraud managers, the challenge is connecting evidence quickly enough to act while avoiding unnecessary disruption to legitimate payments.

Where does check fraud detection happen?

Detection can take place when a check is deposited and when it reaches the institution responsible for paying it.

The bank of first deposit receives the check from the depositor. Its relationship with that customer provides context for evaluating the deposit. The paying bank holds the account on which the check was drawn and can compare the item with information about the check writer.

These institutions have different information. A workflow should account for what each can verify, rather than assume every reviewer sees both sides of the transaction.

What checks does a detection process perform?

1. Read the check and examine its image

Image review looks for inconsistencies or signs of alteration. Optical character recognition, or OCR, extracts text for comparison with other records. Image-analysis tools can also help flag suspicious visual features. Respondents to the Federal Reserve Financial Services 2026 Risk Officer Survey identified AI image analysis, Positive Pay, and manual review among their approaches to reducing check fraud losses.

Image quality matters. A blurred scan or background interference can affect recognition. Needham Bank’s Payee Positive Pay guidelines explain how these issues can create false payee mismatches. A reading error needs investigation; it does not establish fraud.

2. Look for duplicate presentment

Duplicate detection checks whether an item has already been presented. A check deposited through a mobile app might later be presented at a branch, accidentally or deliberately.

The NCUA’s remote deposit capture guidance recommends duplicate detection across multiple days, points of presentment, and deposit applications within the institution.

Fraud managers should confirm which channels a system actually covers and how much history it retains. Detecting a duplicate inside one institution does not automatically establish visibility into deposits at other institutions.

3. Compare the item with issued-check records

For enrolled business customers, Positive Pay compares presented checks with a file of checks the customer issued. Standard comparisons commonly include check number and amount. Payee Positive Pay adds the recipient’s name.

That distinction matters when a criminal changes the payee but leaves the amount and check number intact. Commerce Bank’s Positive Pay explanation describes how payee verification addresses this gap.

Mismatches become exceptions for review. The process depends on accurate issue records and timely decisions, so managers should check both the matching controls and the exception workflow.

4. Put the transaction in account context

Behavioral analysis considers whether activity fits the account’s history. Relevant signals include an unusual check deposit, a sudden change in deposit activity, or rapid withdrawals after deposits.

FinCEN’s check fraud alert identifies these patterns and cautions that no single red flag determines whether activity is suspicious.

For example, an unusually large deposit may have a legitimate explanation. The same deposit followed by unexpected attempts to move the money out warrants a closer look. Reviewers need the sequence of activity and the customer’s context, not just a dollar threshold.

5. Prioritize review and decide what happens next

Rules can flag defined conditions. Risk scores can combine signals to help prioritize the review queue. Neither should be treated as proof that a check is fraudulent.

A useful alert should show why the item was flagged and give the investigator access to the supporting image, transaction details, and relevant account history. The next step might be further verification, customer contact, or an action permitted by the institution’s procedures and applicable requirements.

Speed must be evaluated at the workflow level. An immediate alert offers limited protection if the information arrives too late or nobody owns the review. Funds-availability obligations and return deadlines also constrain what an institution can do; a fraud score does not override them.

How should fraud managers evaluate performance?

Start with outcomes and review quality, not alert volume alone. Useful measures include:

  • Confirmed fraud losses and recoveries, separated from amounts merely flagged.
  • The proportion of reviewed alerts confirmed as fraud.
  • Time from detection to investigation and action.
  • Legitimate transactions delayed or incorrectly challenged.
  • Fraud discovered after an item passed the initial controls.

Use those findings to adjust thresholds, address missing data, and improve staff guidance. A practical starting point is to trace a recently investigated check from deposit through final disposition. Identify which signal mattered, when it became available, and whether the team could act on it. That exercise can expose gaps a list of software features will miss.

For more educational insights on fraud prevention, explore REDi’s fraud prevention blog.