A medical chronology is evidence. It goes to the expert who signs the affidavit, into the demand, into the deposition outline, and eventually in front of a witness who will be asked where a row came from. Software that builds one is therefore not a productivity tool in the ordinary sense; it is a tool that produces something a court may examine. That changes what to look for. The questions below are the ones that decide whether a chronology built in software survives its first contact with the other side, followed by an honest comparison of the vendors that publish their prices, and a script you can run on any product, including ours, before you pay for it.
The chronology guide sets out the by-hand method: inventory the production, Bates it, sort it, extract every dated clinical event into an eight-column table, cite each row to the page, flag gaps, and check the whole thing. Software exists to do the extraction and citation steps at a speed a person cannot match, on productions of thousands of pages, without dropping the entries in the middle. Everything else in a chronology, the significance of an event, the flag that says this is where the standard of care was breached, is judgment, and no vendor should be selling it as automated. Evaluate the software on the mechanical work: does it find the events, does it date them correctly, does it tell you where each one is, and can you prove it.
The first test is the only one that matters if it fails. Take a record you know, run it, pick any row, and open the cited page. The quoted text has to be there. Then pick a row that looks wrong and do the same. A chronology whose rows cannot be traced to a page is a summary, and a summary is exactly what opposing counsel will call it.
Look for the physical PDF page and the Bates number both, because they diverge: a production of forty PDFs has forty "page 1"s, and the Bates stamp is the identifier the other side and the court use. Look for whether the citation is a link you can click to the page image, or a number you have to hunt for. And look for what the tool does when it cannot cite: a row marked as unverified or undated is honest; a row with a page number that turns out to be blank is the failure mode to fear, because it looks sourced and is not. Our own citation auditor exists to run that check on any chronology, from any vendor, in the browser.
Run the same record twice. If the two chronologies differ, ask why. A tool that generates prose with a language model will produce different text on each pass, and sometimes a different set of events; a tool that scans the record for stated events and dates will produce the same table every time. Neither approach is wrong for every job, but they are different products with different evidentiary weight, and vendors do not always say which one they sell.
The distinction has a practical consequence that we measured. In our own testing in September 2026, a synthetic record of roughly 3,500 pages containing 3,533 dated entries was run through a general-purpose model in a single pass and returned 74 events; the same record processed in sized segments returned 1,946. The single-pass table looked complete. The timing write-up has the full numbers. When a vendor says the tool "reads the whole record," ask how they know, and ask for the event count against a record whose entries you have counted yourself. Our free timeline maker is deliberately built the deterministic way: a fixed set of patterns, no generation, and if a date is not on a page the event is listed as undated rather than guessed.
A meaningful share of any production is images: faxed records, scanned paper charts, handwritten nursing notes, monitor strips. A tool that reads only the text layer of a PDF will silently skip those pages, and a chronology with no events from the paper chart is a chronology with a hole where the nursing notes were. Ask three questions. Does the tool OCR image pages at all, and is that included or an add-on? Does it tell you which pages were OCR'd and with what confidence, so a reviewer knows where to read the image rather than trust the text? And does it keep the page number stable through OCR, so a row extracted from a scanned page still cites the physical page in the production?
Test it with a page you know is an image: a scanned consent form or a handwritten progress note. If the tool's output has nothing from that page, you have your answer. The Timeline Builder OCRs scanned pages inside the case and keeps the physical page and Bates number on every entry; the free timeline maker does not read images and says so on the page, which is the honest version of the same limitation.
A chronology can be accurate on every row and still be wrong, because it left out the row that mattered. Completeness is hard to verify by reading the output, so look for tools that expose it. Does the tool report how many pages it processed against how many the production contains? Does it report how many dated entries it found, and can you compare that to a sample count you did by hand on fifty pages? Does it let you ask the record a direct question, every troponin value, every visit with a named provider, every page with the word "fall," and get every instance with its page, so you can test whether the chronology caught them? That last capability is the fastest completeness check there is, and it is why Ask the Record is a separate tool from the chronology builder: a lookup that returns every occurrence is a check on a table that might have missed some.
The chronology leaves the software. It goes to an expert who works in Word, a paralegal who works in Excel, and a case management system that wants a file. Look for CSV export with the columns intact, a Word or PDF export that an expert can annotate, and a format that preserves the citation as text rather than as a link that dies outside the product. Look, too, at whether the export carries the columns you will add by hand, significance and flag, so the finished chronology lives in one document rather than two. A vendor whose only export is a PDF of its own screen is keeping your work product inside its product.
Medical records are protected health information, and a chronology vendor that receives them from a covered entity's law firm is, in HIPAA's terms, handling PHI on someone's behalf. The Privacy Rule's standard is at 45 CFR 164.502(e): a covered entity "may disclose protected health information to a business associate" only "if the covered entity obtains satisfactory assurance that the business associate will appropriately safeguard the information," and those assurances "must be documented through a written contract or other written agreement" meeting 164.504(e). The practical questions for a vendor are five. Will you sign a business associate agreement, and before or after I upload? Where are the records stored, and for how long? Are the records used to train any model? Can I delete a case myself, and does deletion propagate? Is there a written security page, and does it say what is not certified as plainly as what is? A vendor that answers those in a paragraph is a vendor you can put in front of a client's general counsel. Our answers are on the security page and in the BAA itself.
Chronology software is priced three ways, and the model tells you who the product was built for. Per page suits high-volume users with predictable productions and punishes the solo practitioner with one 8,000-page case. Per matter suits firms with a steady docket. Per seat suits firms that want a tool on every desk. Ask what happens at the boundary: the overage rate per page, the page cap per matter, the cost of a second user. And ask for the price in writing, because a vendor that will not publish a price is telling you the price depends on what you look like.
This comparison lists only vendors with a public pricing page, with the figures as those pages showed them when we checked in September 2026. Prices change; check the linked page before relying on it. Vendors whose pricing is by quote only, including Supio, EvenUp, Filevine's MedChron, DigitalOwl, Wisedocs and LegalMation, are not compared, because we will not print estimates as if they were prices.
| Product | Published pricing | Included allowance | Free tier | Exports stated | BAA / data handling stated |
|---|---|---|---|---|---|
| Chartely | Free $0; Pro $49/mo; Scale $199/mo | Free 50 record pages/mo; Pro 1,000 pages/mo then $0.06/page; Scale 7,500 pages/mo then $0.04/page | Yes, no card | CSV and JSON on all tiers; Word export on Pro and above; API and MCP server on Pro | "Records processed in memory, never stored"; the pricing page links a security page and does not itself mention a BAA |
| ProvaLens | Solo $249/mo ($2,490/yr); Practice $499/mo; Firm $999/mo; Enterprise custom | Solo 75 active matters, 5,000 pages per matter; Practice 250 matters, 7,500 pages; Firm 600 matters, 10,000 pages | 30-day trial of the Practice plan, no card | Not specified on the pricing page | "BAA included"; HIPAA and SOC 2 Type II stated |
| MedLegal AI | Pay-per-case $19; Pro $49/mo; Professional $249/mo (5 users); Firm $499/mo | Pro is one user with fair use of 5,000 pages/mo and up to 20 cases; Professional and Firm are unlimited cases | Free trial with 3 case analyses, no card; free timeline maker and chronology demo, no account | CSV and Word (.docx) from the timeline maker; Timeline Builder exports the chronology with Bates cites for an expert or a demand | BAA signed with every customer before PHI upload; records not used to train models; user-initiated case deletion; not SOC 2 certified, stated as such |
Two readings of that table. Chartely and MedLegal AI are priced for the solo and small firm, with a free tier or free tools that produce a page-cited output before any money changes hands; ProvaLens starts at the price of a small firm's monthly subscription and is built around matters rather than pages. And the BAA column is where the vendors differ most in what they say, not necessarily in what they do: a pricing page that is silent on a BAA is a question to ask, not an answer.
Run this on any product, in this order, with a record you already know.
If a vendor will not let you run that script on a trial, that is a finding too. The free timeline maker runs steps two, three and six on a pasted record with no account, and the Timeline Builder runs the whole script on a full production, with free cases to start. Bring your own record and keep score.
It should extract every dated clinical event from a medical record, date it correctly, and cite each row to the physical page and Bates number so a reviewer can open the page and find the quote. Significance and flags are the reviewer's judgment; the software's job is the mechanical extraction and citation, done completely and the same way every time.
Count the dated entries in a sample of pages by hand, run the record, and compare. In our September 2026 testing, a general-purpose model run in a single pass on a roughly 3,500-page synthetic record with 3,533 dated entries returned 74 events, while processing the same record in sized segments returned 1,946. A tidy table is not evidence that the whole chart was read.
Only if it OCRs image pages. A tool that reads the PDF text layer alone silently skips faxed and scanned pages, including handwritten nursing notes. Test with a page you know is an image and check whether the tool reports which pages were OCR'd and keeps the physical page number on rows drawn from them.
Under 45 CFR 164.502(e), a covered entity may disclose protected health information to a business associate only with satisfactory assurances that the information will be safeguarded, documented in a written contract. Ask the vendor whether it signs a BAA before upload, where records are stored, whether records train any model, and whether you can delete a case yourself.
As of September 2026, Chartely (Free, Pro $49/mo for 1,000 pages, Scale $199/mo for 7,500 pages), ProvaLens (Solo $249/mo for 75 matters, Practice $499, Firm $999) and MedLegal AI ($19 per case, Pro $49/mo, Professional $249/mo, Firm $499/mo) publish pricing pages. Supio, EvenUp, Filevine MedChron, DigitalOwl, Wisedocs and LegalMation price by quote.
The free timeline maker returns a dated, page-cited table from pasted or uploaded record text with no account. The Timeline Builder indexes a full production, OCRs scans, and keeps the page and Bates number on every entry.
Open Timeline Builder →Sources. Chartely pricing page and ProvaLens pricing page, as displayed September 26, 2026; MedLegal AI pricing. 45 CFR 164.502(e) (quoted). MedLegal AI, timing write-up, September 2026 (event counts on a synthetic record). Vendors listed as pricing by quote are those whose public sites showed no price when checked; no third-party price estimates are used. Product descriptions of other vendors are limited to what their own pricing pages state.
MedLegal AI is software, not a law firm. This article is general information for attorneys, paralegals, legal nurse consultants and patients; it is not legal or medical advice and does not create an attorney-client relationship. Verify every statute, rule and figure against the current source before relying on it. Questions: [email protected]