# Medical Records OCR & Intake Software > Medical records OCR that reads every format — PDFs, faxes, scans, handwriting, and DICOM. Intelligent OCR routing, deduplication, and per-page confidence flags. Canonical page: https://medrecords.ai/product/ocr/ --- [Home](https://medrecords.ai/) ›[Platform](https://medrecords.ai/product/chronology/) ›Medical Records OCR Medical Records OCR & Intake ## Medical records OCR that reads every format. "Nothing skipped, nothing guessed." Medical records OCR and intake starts with one upload: PDFs, faxes, scans, Word files, handwriting, and DICOM discs together. An intelligent routing engine sends each page to the right OCR pipeline, removes duplicates, separates co-mingled claimants, and flags anything low-confidence for your review instead of silently guessing. [Book a demo](https://medrecords.ai/demo/) [Test a file ](https://medrecords.ai/test-a-file/?src=product-ocr) Intelligent routing Reading — p. 1,462 / 1,847 op_report.pdf36 ppFast OCR intake_scan.jpg2 ppLLM OCR nurse_note.png1 ppHandwriting MRI_L-spine.zip24 slicesDICOM Low-confidence flagged, not guessed1,847 → 612 pp · 3 claimants split 4 engines routed per page — fast OCR, LLM OCR, handwriting, vision tables 1,847 → 612 pages deduped on a real intake — co-mingled claimants separated 0 pages silently dropped — every page read, classified, or flagged ### From drop-box to defensible pages. Pipeline 01 Everything in PDFs, scans, photos, ZIPs, discs — or straight from iManage. 02 Routed per page A decision engine picks the right OCR for each page’s condition. 03 Read in parallel Fast OCR, LLM OCR, a handwriting pipeline, and vision tables. 04 Accounted for Every page read, classified, or flagged, with a coverage report. ### Every page accounted for. Intelligent OCR routing A decision engine routes per page: fast OCR for clean text, LLM OCR for degraded scans, a multi-model pipeline for handwriting, automatically. Page segmentation Multi-hundred-page PDFs are split into individual documents with detected boundaries, dates, authors, and facilities — no manual unbundling. Deduplication Duplicate and near-duplicate pages detected with match scores. A 1,847-page file becomes 612 pages of signal. Originals preserved, nothing silently deleted. Co-mingled detection Records from the wrong claimant are detected and separated before they contaminate the chronology or the summary. Native DICOM intake Imaging discs and ZIPs ingest alongside the paper record: studies and series detected automatically, no separate workflow. Coverage report Per case: pages received, processed, included or excluded, with a stated reason for every exclusion. Works with [Medical ChronologyClean pages become a cited timeline](https://medrecords.ai/product/chronology/) [Medical Imaging ReviewDICOM studies ingest alongside documents](https://medrecords.ai/product/imaging/) [Verifiable AI CitationsEvery extraction traceable to its page](https://medrecords.ai/product/citations/) Intake is the first stage of the pipeline — what gets read here is what gets cited everywhere else. ### The intake spec sheet. Universal Intake™ Formats in PDF · TIFF · JPG · DOCX · fax · DICOM One drop zone for the whole production: paper and imaging together. OCR pipelines 3, routed per page Fast OCR, LLM OCR for degraded scans, multi-model handwriting evaluation. Deduplication 1,847 → 612 pages Near-duplicates detected with match scores; originals kept for the record. Uncertainty scoring Per page Low-confidence pages route to a review queue — flagged, never guessed. Scale 10,000+ pages / case Batches of 26,400 pages observed in production. No file-size wall. Accounting Every page, stated Received, processed, included, excluded, with a reason on every exclusion. ### Who runs their intake through it. The messier the production, the bigger the win. See how each team uses intake. [ Law firms Records dumps arrive as one 4,000-page PDF. Intake unbundles it. For law firms ](https://medrecords.ai/solutions/law-firms/) [ Legal nurse consultants The sorting hours disappear; the clinical judgment hours stay billable. For LNCs ](https://medrecords.ai/solutions/legal-nurse-consultants/) [ Insurance carriers Claim files deduped and separated before review time is spent. For carriers ](https://medrecords.ai/solutions/insurance-carriers/) [ TPAs Volume intake with per-page accounting your clients can audit. For TPAs ](https://medrecords.ai/solutions/tpas/) FAQ ### Medical records OCR, answered. PDFs (native and scanned), TIFFs, JPG/PNG images, Word documents, faxes, and native DICOM imaging, folders, discs, or ZIPs. Everything lands in one case file through one drop zone, with no pre-sorting or format conversion on your side. Yes. Handwritten pages route to a multi-model handwriting evaluation pipeline rather than standard OCR. Every page gets an uncertainty score; anything below threshold is flagged for your review instead of being silently guessed. Duplicate and near-duplicate pages are detected with match scores: refaxed records, re-produced exhibits, repeated cover sheets. A 1,847-page production typically reduces to ~612 pages of signal. Originals are preserved, so nothing is deleted from the record. They’re flagged, not guessed. Uncertainty scoring runs per page, and low-confidence pages route to a review queue. The coverage report shows exactly which pages were received, processed, included, or excluded, with a reason for every exclusion. Yes, co-mingled claimant detection catches pages belonging to the wrong patient and quarantines them before they contaminate the chronology, the summary, or anything downstream. No practical one. The platform is built for 10,000+ page records, and single batches of 26,400 pages have been processed in production. Pagination and lazy loading keep large files responsive. ### See your messiest file, organized. Upload one real production and get back a deduplicated, indexed, page-accounted file. Handled under our BAA; never used to train a model. [Book a demo](https://medrecords.ai/demo/) [Test a file ](https://medrecords.ai/test-a-file/?src=product-ocr)