What backs our claims

Every fact in a chronology is linked to the page it was quoted from.

Each entry carries the page it came from, and the link opens that page inside the record set we deliver. In every run we have measured, no entry was delivered without one.

Zero uncited facts delivered, in every measured run.

The same measured runs. A statement that cannot be matched back to a page is held out of the chronology.

Missing pages are reported as a provable lower bound: “at least N pages did not arrive”.

The finding is arithmetic on the production’s own stamp sequence and the documents’ own page numbering. Every page we could not read is subtracted from it before it is stated.

We do not retrieve records, so we have no reason to call a set complete.

Nothing we sell fetches a record from a provider, and we hold no connection to a health information exchange. The day we retrieve a record for a client, this sentence comes off the site.

Every check is tested against an adversary before its number is quoted.

Before a check is trusted, someone builds the input designed to make it report the comfortable answer. The check is watched failing on it. A check whose adversarial result is not on this page may not have its number quoted on our site.

The set comes back ordered by service date, Bates-stamped, and indexed one row per page: number, date, provider, and the page of your production it came from.

Pages with no readable date keep their order at the end of the set. The report says how many. The index and the ordering are in the downloadable sample case.

The sample case you can download is exactly what a client receives.

It is the unedited set of files from one real run: the same six files a client gets, in the same shapes. Every file was hash-compared before and after zipping. Every page link was checked to resolve inside the extracted copy.

There is no real patient data anywhere in the sample.

It was generated from a synthetic record set built for testing.

A page we cannot read is listed as unreadable and reported to you.

Unreadable and refused pages go into their own list in the completeness report. They also go into a review queue before anything ships. Such a page is netted out of the missing-page count.

No record is sent to an AI model.

The chronology is built by rule-based software. Scanned pages are read by Tesseract, an open-source text-recognition program running in our own hosting. An AI model is added only under a signed business associate agreement with zero data retention, and this page changes before it is.

Security reports go to security@claracord.com, which is also the contact in our security.txt.

The address is live and accepts mail from outside senders, and /.well-known/security.txt names it. The file and this row change together or not at all.

What you type into the request form is not shared or sold, and our privacy page names every field it asks for.

It posts to this site’s own server and is stored there. The notification we receive says only that something arrived.

The workers’-compensation deliverable is clerical: it contains no summary or opinion.

Claracord organizes records. They are not summarized for a med-legal report. The deliverable is not for use in QME or AME report preparation (California Labor Code section 4628).

We work only with licensed firms and licensed claims professionals, and the license number you give is checked before any file moves.

The request form will not submit without a state and a license number. A state bar number, or a claims or adjuster license. The citation audit is ordered through that same door, on the same published tiers. No record set is accepted until a person has checked it.

The worked example on our resource pages is the same case as the finding on the home page.

59 stamped sheets, 57 pages received, 50 carrying a legible stamp. Our build fails if those numbers ever disagree between the home page, the resource page and the report.

The missing-page check runs on every record set we organize.

Every record set gets one, alongside the chronology. Neither is optional.

Every figure we quote from somebody else carries the date we checked it, and a vendor who publishes no price is given no number.

Ours prints the month it was verified, and a weekly watch re- reads the sources. Where a vendor posts no rate we describe how they sell.

A chronology is not an expert opinion, and it does not satisfy a certificate-of-merit or affidavit-of-merit requirement.

Those requirements call for review by a qualified expert, and nothing we deliver is that review. What we deliver is the record in order, each entry cited to its page.

Every page removed as a duplicate comes back with your set, stamped with what it duplicated and why. A borderline page is kept and flagged rather than removed.

The removed pages are delivered alongside the clean set, each one showing what it matched. The close calls are kept. The removed-page document is in the sample case on our home page.

No trackers, no cookies, nothing third-party.

This site sets no cookie, runs no analytics script, and loads nothing from any other domain. It is checked before every deploy.

You can download the sample without telling us your name.

It is a plain file on this site. No form in front of it, no account, no email address, and nothing about you sent anywhere.

Every sentence on this site maps to something real, and this page is generated from that map.

The map is a table in this site’s own source repository, and this page is rendered from it. Our build fails if a claim on the site and its row in that table disagree. It fails too if a row is added with no wording for this page.

Human-verified: a person reviews the work before it ships.

No chronology leaves without a person going through it. The reviewer console logs every review action.

You see the price before anything runs.

The quote comes off the page count when the file arrives. You approve it in the portal before the file is processed.

The first chronology is free: one per firm, up to 500 pages.

The offer is made by the request form on this site and applied in the portal.

The price is published: the first 250 pages are $150, pages 251 to 1,000 are 48¢ each, 1,001 to 2,500 are 36¢, and every page after that is 24¢.

The bands are marginal: a file of 1,001 pages costs 36¢ more than one of 1,000, not a tier more.

An error of ours is corrected and re-issued at no charge, and new records are an update priced on the published tiers.

If the mistake is ours, the corrected version comes back with no invoice attached. A supplemental production is quoted from its page count on the same tiers as the first case.

Every case carries its own invoice.

One invoice per case, itemized against that case’s page count, and never a monthly statement. It arrives with the delivery, which is also when the card is charged.

A demand, chronology or medical summary is checked against the record set it cites: the report states whether each citation is supported by the cited page, and it states no medical opinion and no view on the merits.

It reports whether a citation is supported by the page it cites. Never whether the underlying medical claim is correct, reasonable or worth anything. Its price rides the published tiers, per page of the record set; the document being audited is not charged.

The chronology arrives as a PDF, an Excel workbook and a Word document.

Three files of one chronology, not three chronologies. Each export is stamped as an export of a numbered version of the record. None of them can be edited in our software.

A provider and custodian index, listing the records your set mentions but does not contain.

One page per case: who treated, how many visits, which pages each custodian produced. Also the documents the records refer to that are not in the set. It tells you what your set mentions, not everything that is missing.

Gaps between visits, and the split between records from before and after the injury.

Arithmetic and a fixed legend. The table states the interval between one visit and the next. It states which side of the injury date each record falls on. Short intervals are not listed.

A medication table, as printed.

Drug, dose, prescriber, date and the page it came from, transcribed as the record prints it. Nothing is converted and nothing is inferred. It is not in the sample you can download today.

A billing ledger: every bill, totaled, each charge with its page.

Every charge read off the billing pages in your production, grouped by provider. Each row carries the page it was printed on. The only arithmetic is addition. Nothing in it says what a charge should have been. It is included in every case with a chronology, and it is not in the sample you can download today.

A records-request letter for each custodian with pages missing.

Generated from the completeness finding. Every value is filled verbatim from that report or from what you told us about the case. It is a DRAFT for your letterhead and your signature, and it is never sent by us.

A verify sheet naming the reviewer’s location and what a person checked.

The sheet states what was done to your file, what was not, and who did it, and it names no reviewer. The sample you can download predates the line naming the reviewer’s location.

Workers’ compensation sets carry a dedup certificate: pages received, pages kept, pages removed as identical, pages flagged rather than removed.

Four counts of one file on one sheet, under the same legend the set carries. That legend states what was and was not done to the file. A page that might be unique is flagged and kept, never removed.

Your card is charged only when the work is delivered.

The card is authorized when you approve the quote and captured when the files are in your hands. If we cannot deliver, the authorization is voided and nothing is taken.

Unused prepaid balance is refundable.

If a firm prefers to prepay, the balance is dollars and not credits. It does not expire, and every case is itemized against it in dollars. Whatever is left is returned in full on request.

Records are processed and stored in the United States, on Amazon Web Services in the us-east-1 region.

If it ever moves, this row moves with it, in the same release.

Records are encrypted in transit and at rest.

Object storage and databases are encrypted at rest, and every connection is over TLS.

Each firm’s records are isolated from every other firm’s, and the isolation is tested.

A test suite tries to read one firm’s case as another firm. It has to fail before a release ships.

The subprocessors are Amazon Web Services, Stripe, Cloudflare and Google Workspace.

Hosting, payments, the edge, and mail. The list is closed. A fifth name appears here and on our trust page before it touches a client record.

Reviewers are based in the United States and work under confidentiality agreements.

The same fact prints on the verify sheet of every bundle, read from the reviewer record rather than typed. No reviewer is ever named to a client.

A client account can add a one-time code from an authenticator app at sign-in.

Sign-in is a link we email you, and the code is the second factor on top of it. A firm-level setting that requires the code of everyone at your firm is not built and is not claimed here.

Records and deliverables are retained 90 days after delivery, or from the newest delivery if a case is redelivered.

Portal MSA 1.5-draft, section 6. A case is deleted sooner at the firm’s request, and every deletion is logged.

A BAA is signed before any record is uploaded.

The business associate terms are in the data processing agreement. The portal requires it accepted before any upload.

Workers’-compensation applicant attorneys are billed directly. We never bill through a lien, and we never file one.

The quote, the card authorization and the charge all go to the firm’s own account. Nothing we build has a path that routes an invoice through a WCAB lien.

Cases run through the self-serve portal: upload, quote, approval and delivery.

A firm requests access by the form on this site. Each request is reviewed and approved before the firm’s first upload.

Productions from several custodians are combined into one record set, with every page’s origin tracked.

Also a promise, from the same ruling of 18 September 2026. Send the case in however many productions it reached you. One record set comes back with every page’s origin tracked.

That most AI vendors in this market offer to summarize and analyze the file.

Read from what vendors publish about themselves, last looked at in September 2026. No vendor is named, and this never turns into a number.

What US nurse-reviewed and offshore chronology services charge, and how long they take.

Read from vendors’ own published pricing and given as ranges. It is what they say, not what we measured.

What self-serve AI chronology tools charge, and that the tier ships no human review.

Published pay-as-you-go rates, read from the vendors’ own pricing pages. The tier is characterized and no vendor is named.

How AI chronology platforms price their product.

Characterized from industry reporting surveyed on 27 August 2026. Attributed where it appears, and not sharpened into a precision the sources do not have.

What it costs a firm to do this in-house.

Arithmetic rather than a measurement: hours times a rate, with every assumption printed beside it. No study is claimed behind it.

What our malpractice pages say about certificate-of-merit and affidavit-of-merit requirements.

That many states have one, and that what it is called and who may sign it vary by state. That the document rests on a qualified expert’s signed opinion. It is a description of the law for the attorneys we work with, not legal advice.

What our long-term disability pages say about how ERISA-governed benefit denials are reviewed, and about the Department of Labor’s claims rules.

In most ERISA-governed cases a reviewing court looks mainly at the claim file. Not in every court or under every policy. A description of the law for the attorneys we work with, not legal advice. Paraphrased from 29 CFR 2560.503-1 and checked in September 2026.

A completed security audit or certification: SOC 2, HIPAA, or any other.

What would have to be true first: our own security checklist finished. Then an audit performed by someone who is not us.

Client outcomes, settlement figures, or a count of firms served.

What would have to be true first: delivered work we can measure and cite. No outcome number is published, and any we ever publish will carry its method and its source.