M
Medrecords AI
Medrecords AI turns medical records into the chronologies, summaries and reports that injury, malpractice, disability and claims work runs on. It reads PDFs, scans, handwriting and DICOM imaging, sets aside duplicate pages before billing, flags missing records and files that mix patients, and builds a dated chronology in which every event links to its source page or imaging slice. Record Q&A answers questions about the file with a citation on each sentence.
Reports are drafted in the customer's own template and letterhead across more than twenty types, among them IME reports, demand letters, deposition summaries, life care plans and Medicare set aside drafts. On workers' compensation and IME cases, case chat can also cite California workers' compensation law and published decisions. Named buyers include plaintiff and defense law firms, legal nurse consultants, IME and QME physicians, expert witnesses, insurers, third party administrators and government claims programs.
Pricing is published per deduplicated page, and an annual on premises license is quoted separately. The company is AI Health Studio LLC, doing business as Medrecords AI, of Sheridan, Wyoming. It was founded in 2024, is run by its two founders and takes no venture funding.
Alternatives to Medrecords AI
Other products that do the same kind of work as Medrecords AI, and what each one does.
- WWisedocsWisedocs is an AI medical record review platform for claims work, sold to insurance carriers, third party administrators, independent medical evaluators and law firms.
- EEvenUpEvenUp is an AI platform built only for plaintiff personal injury firms, which it calls a Claims Intelligence Platform.
- NNovoNovo drafts the two documents that carry a personal injury claim through the stage before litigation: the medical chronology and the demand letter.
- PPattern DataPattern Data is an AI case evaluation platform built for mass tort litigation, sold to plaintiff law firms, defense firms, settlement administrators and special masters.
- NNeosNeos is cloud case management software built for plaintiff personal injury firms.
- EEveEve is an AI platform for plaintiff law firms, covering the case from intake to resolution. It handles intake and case evaluation, medical overviews and chronologies, demand letters, complaints, discovery in both directions, and case value analysis.
Capability grades
All 15 axes, graded from public sources on the date shown. Hover a grade to see what the letter means on that axis.
AI Centrality
How much of the product is actually AI. Whether the machine learning is the mechanism the buyer is paying for or a feature layered onto conventional software, and whether the vendor is specific about which is which.
Models do the work that is sold. Reading scans and handwriting, extracting events, building the chronology, answering questions about the file and drafting every report all run on model calls, and the customer pays by the page processed. Take the models away and what remains is a document store with a duplicate page check.
Citation Accuracy and Hallucination Disclosure
Whether the vendor publishes measured accuracy on citations and assertions, grounds output to primary sources, and says plainly what its system does when it does not know. Legal has a documented public record of fabricated citations reaching filed briefs, so an untested claim of accuracy is not evidence.
Grounding is described in detail, and no accuracy figure is published. Every chronology entry and answer links to its source page or DICOM slice. Two checks are said to run on every extraction: a mechanical match of each fact against the text of its page, and a second model checking the first. Q&A sentences are checked against the pages they cite before the user sees them, and report claims are labeled Supported, Unsupported, Contradicted or Not checked.
A speed figure is given, about 1,000 pages an hour on live cases. No accuracy rate, test set or error analysis is published, and buyers are pointed to a free first case on their own file instead.
Autonomy and Oversight Model
What the system decides on its own, what a lawyer must approve, and whether the vendor documents where the review point sits. A tool that drafts under review and a tool that files without one are different products and different risks.
The control structure is published, and the professional is written in as the last step. Output comes in three modes, one of which returns only verbatim record text. Fields carry a confidence score, low confidence readings are flagged to the reviewer instead of resolved, and report claims that fail verification are marked and can be rewritten. Some uses are ruled out by name. Imaging tools must not be used for diagnosis, and section 9.4 of the terms says benchmark ranges from comparable matters are not a valuation and must not be presented to a client, court or regulator as one.
Case merit scoring and standard of care findings are labeled provisional and need professional sign off. Section 9.2 requires a licensed or qualified professional on the customer's side to verify output before it is filed, served or relied on, and section 16.1 puts the cost of relying on unverified output on the customer.
Operational and Outcome Evidence
Named, dated evidence that the product works in production at real firms or legal departments. Case studies with figures and identified customers count. Unattributed testimonials and launch announcements do not.
No customer is named. The About page says customer names are shared on reference calls under a nondisclosure agreement. Testimonials carry a first name and a role, such as an IME physician or a legal nurse consultant, and the vendor says they come from recorded calls, emails and one G2 review. Company figures are self reported: more than 80 percent of early customers retained, single files above 10,000 pages and cases above 50,000 pages run in production, and a typical case finished in about an hour. None is tied to a named firm or a stated method.
Privilege and Confidentiality Posture
How client confidences are handled: attorney client privilege and work product treatment, segregation of one client matter from another, whether client data trains any model, and what the vendor commits to in writing rather than in marketing.
The commitments a firm would ask for sit in the published agreements. Section 8.2 of the terms and section 4.4 of the DPA prohibit using customer content to train, fine tune or improve any model, the vendor's or a third party's, and the prohibition reaches anonymized and aggregated versions of it. Every AI provider in the path is bound to the same restriction and to zero retention inference under a business associate agreement.
Generated chronologies and reports belong to the customer, and the terms say they typically constitute attorney work product for law firms. Litigation hold and protective order constraints the customer passes on are treated as processing instructions. Access is set case by case, so users see only the cases assigned to them, and retention is set by the customer per case or across the account, with deletion certificates.
UPL and Professional Responsibility Posture
Whether the vendor is clear that it supplies a tool rather than legal advice, who its audience is, and how it addresses unauthorized practice of law, competence and supervision duties, and jurisdiction limits. ABA Formal Opinion 512 is the reference point. Where the advice line is not the duty a product raises, the axis is read through the nearest professional duty it does raise: judicial conduct rules and the reviewing duty for products sold only to courts, and the duty to bill for time actually spent for products that draft time entries.
The advice line is drawn plainly and tied to professional duty. Section 9.3 of the terms says the service is not the practice of law or medicine and creates no attorney client relationship. Section 9.2 says the duties of independent judgment that attorneys and physicians owe cannot be delegated, and requires a licensed or qualified professional to verify output before use. Section 2 says the service is not directed to consumers or to anyone under 18.
What is not stated is where it may be used. No country or state limits are named, and the buyer list includes disability and veterans representatives and legal nurse consultants who are not lawyers, held to the same verification duty.
AI Governance and Bias Disclosure
Published governance over model behavior: who owns it inside the vendor, what is tested before release, and what is disclosed about disparate output across matter types, parties, or populations.
Scope limits are published, and a governance program is not. The vendor states what the product never decides, labels its riskier features provisional and routes AI requests only to approved models. The About page names the CTO as leading the review pipeline, an engineering role rather than accountability for how the models behave. No release testing regime, evaluation results or bias review is published, on a product whose summaries feed injury, disability and malpractice decisions.
AI Safety and Data Stewardship
Retention, deletion, access control, and what happens to prompts and documents after they are processed. Whether the vendor states its subprocessors and its incident practice, or leaves the buyer to assume.
Retention, deletion, access, subprocessors and incident handling are all published in specific terms. Customers set retention per case or for the whole account, with 180 days, 365 days, six years or no automatic deletion offered in settings. They can place legal holds and delete a document, case or workspace at any time. Deleted content leaves production systems and backups within 30 days, and a deletion certificate follows within 48 hours of a termination purge, or on request for a single case.
Access is case by case, and every access to protected health information is logged, kept six years and exportable. The dated subprocessor list carries 30 days notice of changes with a right to object, and breach notice is due within 72 hours of confirmation.
AI Liability and Recourse
What the vendor stands behind contractually when its output is wrong. Indemnities, caps, carve outs, insurance, and whether any of it is published or only reachable through a negotiated agreement.
Liability follows a standard limitation clause, and output errors sit with the customer. The service is provided as is, with no warranty that output is accurate, complete or reliable without professional verification. Each party's total liability is capped at the fees paid in the preceding 12 months, and consequential damages are excluded. The vendor's indemnity covers only claims that the service infringes US intellectual property.
For security incidents, the vendor is liable only for a material failure to keep the safeguards listed in the DPA. No insurance position is published.
Practice Systems Integration Depth
How deeply the product reaches into the systems legal work already lives in: document management such as iManage and NetDocuments, Word and Outlook, contract lifecycle management, matter management, e-billing, and court filing systems.
An API exists, and no practice system is named. The API page describes creating cases and uploading files over REST, then pulling structured output, including summaries, timelines, providers, missing records and citations, by polling. Credentials and endpoint documentation are released after a technical review and a signed agreement. The surfaces disagree on push: the home page mentions pushing output to a CRM and the terms list webhooks, while the API page says outbound webhooks are a roadmap item. No case management, document management or claims system is named on any page read.
Deployment Model and Data Residency
Where the software runs and where the data sits. Multi tenant cloud, single tenant, private deployment, on premises, and whether region of residence is a published option or an enterprise conversation.
Both deployment options and where data sits are published. The self serve cloud stores records on AWS in US data centers, us-east-1 with us-west-2 for disaster recovery. The DPA says customer content is processed only in the US, and the model providers on the subprocessor list are US based. Enterprise customers can run the platform in their own infrastructure or private cloud on AWS or Azure, with an air gapped option and no telemetry or callbacks. In that mode no subprocessor receives customer content unless the customer turns it on.
Security Certifications and Trust Center
Independent attestation a buyer can pull without a sales call: SOC 2, ISO 27001, penetration test summaries, a trust center with current reports and named scope rather than a badge image.
No independent attestation is held, and the vendor says so itself. Marketing copy describes the company as SOC 2 and HIPAA compliant. The security page offers SOC 2 and HIPAA readiness reports dated 3 November 2026, free to download with no form, and describes them as output of continuous control monitoring in Sprinto, not an auditor's attestation. No auditor, report type or audit period is named. Penetration test summaries and the full evidence package are offered under a nondisclosure agreement on request. The DPA adds that the vendor uses the word certified only when a certificate exists.
Model Supply Chain Disclosure
Which models sit underneath, whose they are, where they run, and whether the vendor commits to telling customers when that changes. A legal buyer inherits every dependency it cannot see.
Providers are named, and most models are not. Every AI request goes through OpenRouter, which the subprocessor list says sends most requests to open source models, with Anthropic and OpenAI as backup providers reached only through OpenRouter. All three sit under business associate agreements with zero retention and no training terms, and processing stays in the US. Which open source models carry the main workload is not stated.
The 30 day change notice applies to subprocessors, so a switch between models behind the same gateway would not trigger it as written.
Commercial Transparency
Whether a buyer can learn what this costs without entering a sales process: published rates, the unit being charged, what sits behind an enterprise tier, and what implementation adds.
Self serve pricing is complete, and the enterprise license is quoted. Pages cost 10 cents each after deduplication, falling to 8 cents at 100,000 credits and 5 cents at 500,000. Reports cost $10 each, there is no setup fee, seat charge or minimum, and credits do not expire. Duplicate pages are not billed, and the first case is free for seven days up to 1,000 pages. The annual on premises license, with uncapped pages and a one time deployment fee, is priced by quote with no figure published.
Firm and Practice Coverage
Who the product is actually built for. AmLaw, midlaw, small firm and solo, in house departments, government and courts, and which practice areas are supported rather than merely claimed.
Coverage is described case type by case type, and the boundary is drawn around tasks rather than buyers. Solution pages cover plaintiff and defense personal injury, medical malpractice, mass tort, nursing home neglect, workers' compensation, long term disability, Social Security and veterans disability, and IME and QME work. Insurers, third party administrators, hospital risk teams, settlement administrators and government claims programs are named alongside law firms.
The stated limits are about the work: the product is not an EHR, a clinical scribe or a records retrieval service, and it does not decide standard of care, causation or case value. Firm size and in house legal use are not addressed.
7 public documents
The public pages on file for Medrecords AI, with the recorded signals each one supports and the date it was last read. Open any of them and check the reading against the record.
-
medrecords.ai/dpa3 signals
Prompt and Output Retention, Ethical Walls and Matter Segregation, Third Party Request and Subpoena Notice
Read Nov 15, 2026
-
medrecords.ai/changelog2 signals
Primary Law Corpus Provenance, Court Disclosure Support
Read Nov 15, 2026
-
medrecords.ai/terms2 signals
Client Data in Training, Bar Guidance Alignment
Read Nov 15, 2026
-
medrecords.ai1 signal
Refusal and Uncertainty Behavior
Read Nov 15, 2026
-
medrecords.ai/security1 signal
Outside Counsel Guideline Readiness
Read Nov 15, 2026
-
Billing and Fee Posture
Read Nov 15, 2026
-
Fabricated Citation Record
Read Nov 15, 2026
$0.10 per page
- You pay for each page of medical records the tool reads.
- A page costs 10 cents, and the price drops to 5 cents if you buy a lot.
- Duplicate pages are free, and there is no setup fee or monthly minimum.
- A finished report costs $10.
- Big customers who run it on their own servers get a custom yearly price.
Two plans are published. Self Service is metered by the page: one credit buys one deduplicated page at 10 cents, falling to 8 cents at 100,000 credits and 5 cents at 500,000. A drafted report costs $10, or 100 credits. There is no setup fee, seat charge or minimum, credits never expire, and exact duplicate pages are detected before billing and not charged. The worked examples on the pricing page are 1,847 pages for $184.70 and 40 pages for $4.
Enterprise On Prem is an annual license with uncapped pages and no seat fees, priced by quote, plus a one time deployment fee; from the second year only the license is charged. Both plans carry every capability, and nothing is held back for Enterprise.
Implementation: None on Self Service. Enterprise On Prem adds a one time deployment fee, with no figure published, and takes about four weeks from scoping to go live.
Confidentiality and data terms: A business associate agreement is in place before the first upload on every plan, and the DPA is built into Self Service signup. On Enterprise On Prem the BAA or DPA is available on request.
Note: The free trial covers one case of up to 1,000 pages for seven days. Self Service access opens within one business day of a demo, so the route to buy runs through a call even though the rates are public.
Legal Signals
What each signal meansA signal records what public sources say on the date shown. It is not a grade and it is not a recommendation. Where a signal reads Not addressed, it means the index did not locate the material in public sources on that date, which is a statement about disclosure rather than about the product.
Client Data in Training
Can material a lawyer puts into this product be used to train a model?
The published terms prohibit training on customer content. Not a policy page, the agreement.
Section 8.2 of the terms and section 4.4 of the DPA prohibit using customer content to train, fine tune or improve any model, the vendor's or a third party's, and the prohibition reaches anonymized and aggregated versions of it. Every AI provider in the processing path is bound in writing to the same restriction. The vendor keeps a right to use service data, meaning usage telemetry, in aggregated form, and the terms state that right is not a license to strip identifiers from customer content and use the result.
Prompt and Output Retention
How long does the product keep what a lawyer typed, and can that be set to zero?
The customer controls the retention window, by product configuration or by contractual instruction, but zero retention is not stated as available.
Section 5.1 of the DPA makes retention a customer setting, per case or across the account, with legal holds and deletion of any document, case or workspace at any time. Settings offer 180 days, 365 days, six years or no automatic deletion, and without a setting content is kept while the case stays active. Q&A history is not addressed apart from case content, and the model providers run zero retention inference.
Ethical Walls and Matter Segregation
Does retrieval respect the firm’s ethical walls, or can the model read across them?
The product maintains its own permission model, documented, requiring the firm to keep it aligned.
The DPA describes case level authorization in which users and staff see only the cases assigned to them, with role based permissions, least privilege, access reviews and every access logged. The home page FAQ says case level permissions are in place and full role based access control is coming soon. Nothing describes conflict screening across matters inside a firm, or access taken from a document management system.
Third Party Request and Subpoena Notice
If someone subpoenas the vendor for a firm’s data, does the firm hear about it first?
Terms commit to notice where lawfully permitted. No transparency report located.
Section 6.3 of the DPA commits the vendor to notify the customer within 24 hours of a subpoena, warrant or court order for customer personal data unless legally prohibited, to redirect the requester to the customer where possible and to challenge overbroad demands. The confidentiality section of the terms adds prompt notice and help to contest compelled disclosure where legally permitted. No transparency report is published.
Primary Law Corpus Provenance
Where does the law in this product come from, and does the vendor have the right to use it?
Sources are identified without stating the license or rights basis.
On workers' compensation and IME cases, case chat can cite the California Labor Code, SB 863 and SB 899, Title 8 regulations including MTUS and its Drug List, en banc WCAB decisions and published appellate opinions, each linked. Standard of care answers search state rules first, then federal rules, then PubMed. No license, public domain basis or update cadence is stated for any of these sources, and the rest of the product works from the customer's own records.
Good Law Verification
Does the product tell you when the authority it just cited has been overruled?
No located public material addresses whether authority is checked for subsequent history.
Case chat can cite published WCAB and appellate decisions, and nothing says whether they are checked for later treatment. For regulations, a question about care on a past date finds the amended rule and the answer notes that the text is a later version, and retracted papers are marked in the clinical literature search. The changelog, the AI information page, the home page, the law firm page and the terms were read.
Refusal and Uncertainty Behavior
What does the product do when the answer is not in the corpus?
The vendor describes refusal or abstention behavior in public materials.
The vendor states that a finding it cannot cite is not given, that uncited answers are not allowed in Q&A, and that low confidence readings are flagged instead of guessed. Each Q&A sentence is checked against its cited pages before display, and the sample output on the home page labels findings Supported, Not in the record and Unexplained. No published evaluation measures how often the product declines when it should.
Fabricated Citation Record
Does a public court record exist addressing fabricated or hallucinated legal citations in output from this product?
No court order, opinion or disciplinary record addressing fabricated or hallucinated legal citations produced by this product has been located as of the date shown. This is a statement about the public record on that one subject, not a finding about the product, and this signal is not a litigation history.
The AI Hallucination Cases database maintained by Damien Charlotin holds no court order, opinion or disciplinary record naming Medrecords AI or AI Health Studio LLC. This is a statement about the public record on fabricated citations, not a finding about the product.
Bar Guidance Alignment
Has the vendor engaged in public with the ethics opinions its buyers are bound by?
Public materials refer to professional responsibility in general terms without naming guidance.
Section 9.2 of the terms says the duties of independent judgment that attorneys and physicians owe cannot be delegated, and requires a licensed or qualified professional to verify output before filing or reliance. No ABA opinion, state bar opinion or rule of professional conduct is named on the legal documents, the law firm page or the evaluation guide.
Billing and Fee Posture
Does the vendor address what happens to the bill when the work takes an hour instead of six?
Public materials claim time savings without addressing billing or disclosure, and the product sits inside a fee relationship between a lawyer and a client where those savings would change the bill.
The law firm page asks what the product does to a firm's billing model and answers that it turns review hours the firm struggles to collect on into strategy hours, with per page pricing that scales with the matter rather than headcount. That is a realization argument for the firm. Nothing addresses how AI assisted review is recorded or disclosed on a client's bill, or whether per page charges reach the client as a case cost.
Outside Counsel Guideline Readiness
Can a firm get this vendor through a client’s AI clause without a bespoke negotiation?
A subprocessor and model provider list plus client facing disclosure material is published or available without an agreement in place.
A dated subprocessor list names every party that can touch customer content, with its location and terms: AWS for storage, OpenRouter as the AI gateway, and Anthropic and OpenAI as backup model providers. The DPA is published and built into self serve signup, the business associate terms are summarized on a public page, and the security page says it is written to be forwarded to a client audit. The open source models that handle most requests are not named.
Court Disclosure Support
If a judge’s standing order requires an AI disclosure, can the product produce one?
Some elements of the record are available, short of a document level export.
Every line of a chronology or report cites its source page, a verification panel marks each report claim Supported, Unsupported, Contradicted or Not checked, and attested reports keep the professional's signature block on export. Access logs are exportable. Nothing records which model produced a passage, and no export brings the verification results and the model record together for a court or tribunal.