Best Check OCR Software in 2026

The MICR (Magnetic Ink Character Recognition) line is the string of characters printed at the bottom of every check using magnetic ink in E-13B font (in the US) or CMC-7 font (in some international markets). It contains three essential data elements separated by special MICR symbols (transit, on-us, dash, and amount): the ABA routing number (9 digits identifying the paying bank), the account number (variable length, typically 10-12 digits identifying the payer's account), and the check serial number. When checks are processed through the Federal Reserve or clearinghouse systems, the MICR line is read by magnetic readers — not optical scanners — for reliable identification. For software-based check OCR working from images rather than physical checks, accurately extracting these MICR fields is critical because they are the primary identifiers used for deposit processing, reconciliation, duplicate detection, and fraud prevention.

Yes, but accuracy on handwritten fields varies significantly by field type and handwriting quality. The MICR line is always machine-printed, so it is not affected by handwriting. The courtesy amount (numeric amount box) is often handwritten, and modern AI-based OCR achieves 92-97% accuracy on legible handwritten amounts. The legal amount (written-out words) is also often handwritten and is more challenging — accuracy ranges from 85-95% depending on handwriting legibility. The payee name, date, and memo fields are typically handwritten and present the most difficulty, with accuracy ranging from 80-92%. For critical financial processing, the best practice is to extract all fields via AI OCR and then route checks with low-confidence handwritten field extractions to a human review queue. Lido's AI extraction handles handwritten check fields with competitive accuracy across all field types.

Check image quality is governed by the ANSI X9.100-187 standard for check image quality, which specifies minimum requirements for images used in check clearing. The standard requires a minimum resolution of 200 DPI (dots per inch), with 300 DPI recommended for optimal OCR accuracy. Images should be captured in grayscale or bitonal (black and white), with the full check visible including all four corners. Common image quality issues that degrade OCR accuracy include: low resolution from smartphone cameras held too far from the check, motion blur from camera movement, shadow or glare from uneven lighting, skew (check not aligned with the camera), and clipping (edges of the check cut off). The best check OCR platforms include image quality assessment (IQA) that detects these issues and either compensates automatically or rejects the image with guidance for recapture.

Check OCR software extracts the date field from the check image but typically does not enforce date-based business rules — that logic belongs in the downstream processing system. A post-dated check (dated in the future) may or may not be valid depending on the jurisdiction and the parties' agreement. A stale-dated check (typically more than 6 months old, though this varies by jurisdiction and bank policy) may be refused by the paying bank. The OCR software's job is to extract the date accurately so the downstream system can apply the appropriate rules. Some banking-specific platforms like Mitek and OrboGraph include date validation as part of their risk assessment, flagging post-dated and stale-dated checks during the deposit capture process. General-purpose OCR tools like Lido extract the date field accurately and leave the validation to your workflow.

CAR stands for Courtesy Amount Recognition — the extraction of the numeric dollar amount written in the amount box on the right side of the check (e.g., '$1,234.56'). LAR stands for Legal Amount Recognition — the extraction and interpretation of the written-out dollar amount on the line below the payee (e.g., 'One thousand two hundred thirty-four and 56/100 dollars'). Under the Uniform Commercial Code (UCC Section 3-114), when the courtesy amount and legal amount disagree, the legal amount (words) controls. This is why both must be extracted and cross-validated — a mismatch could indicate an error or a fraudulent alteration. The best check OCR tools extract both amounts, convert the legal amount text into a numeric value, and flag any discrepancy between the two for human review.