Medrecords AI for QME record files, described plainly
Short answer: Medrecords AI is AI medical record review software for physicians: it turns a record file into a page-cited chronology and summary, and drafts reports from your notes. For a QME, it helps with the slow part of record review on large files. It does not count pages for your verification, sign anything, or bill the payer.
Limits: Not legal or billing advice. California rules are quoted as read on 2026-11-19 from dir.ca.gov and leginfo.legislature.ca.gov; verify the current text there before you bill, sign or serve, and ask counsel about your case.
Medrecords AI publishes this site. Facts on this page come only from medrecords.ai/pricing and the company's llms.txt, read on 2026-11-12.
What it does with a QME file
- Reads PDFs, scans and handwriting (OCR) and builds a chronology and summary where every line cites its source page.
- Detects exact duplicate pages and doesn't charge for them, so you can see how many pages repeat across productions. See deduplication.
- Adds Bates numbers, redacts, and exports hyperlinked files.
- Flags records that look missing from the file.
- Answers questions about the file with page citations.
- Drafts a report in your template from your exam notes or dictation, for you to edit, verify and sign. See report drafting and the QME report checklist.
Typical turnaround is about 1 hour per case.
What it costs
| Item | Price |
|---|---|
| Per deduplicated page, to start | 10 cents |
| Per page at 100,000 credits | 8 cents |
| Per page at 500,000 credits | 5 cents |
| A report | $10 |
| Trial | 7 days, 1 case, up to 1,000 pages, no card, after you book a demo |
From medrecords.ai/pricing.md, read 2026-11-12. No subscription or seats; credits don't expire.
For scale: on the synthetic file in our worked example, the 970 declared pages minus 140 exact duplicates would be 830 billed pages, or $83 at 10 cents. That is an illustration of the published rate, not a quote.
Security
SOC 2 and HIPAA controls, a BAA signed at signup, and a contractual commitment that customer records never train a model. An Enterprise plan runs inside your own infrastructure for practices that can't send records to a cloud service; it's quoted on a call.
What it doesn't do
- It doesn't sign your page verification. The 9795(c) verification is yours, under penalty of perjury. Use its duplicate and page data as input, then count and sign yourself.
- It doesn't decide your duplicate policy. It shows duplicates; whether you bill them is your call, and the regulation is silent.
- It doesn't judge declarations. Whether a sender's 9793(n) declaration is complete is for you and the parties.
- It doesn't bill or collect. No ML code submission, explanation of review handling, second review or independent bill review.
- It doesn't form opinions. Causation, impairment and apportionment are yours to write and sign.
When to pick something else
- Your problem is collections, not review time. A workers' comp billing service handles bills, explanations of review and disputes. Medrecords AI doesn't.
- Your files are mostly under 200 pages. ML-PRR doesn't apply and reading takes little time; software adds less.
- You want a person to prepare the summary. Outsourced record review services use human reviewers. Medrecords AI is software you run.
- Records may not leave your network and you can't wait for an on-prem deployment. The cloud plan won't fit, and Enterprise takes about 4 weeks to deploy.
Try it on a closed file
The fair test is a file you already finished: compare its chronology, duplicate count and record list with your own. Book a demo to start the 7-day trial on 1 case.