Methodology: sources, dates and our conflict of interest
Short version: every fact on this site comes from a primary source or the vendor's own page, read on the date shown. We did not test any product with patient records. We make Medrecords AI, mark it as ours everywhere, and name the cases where another tool fits better.
Limits: We read every rule, ruling and product page on the date shown and did not test any product with patient records. Court summaries here are not legal advice; how your court treats AI use depends on the jurisdiction and the judge, so ask retaining counsel.
What we read
- Search results: the Google top 10 for "best ai for expert witnesses", "ai for expert witnesses", "expert witness ai tools" and "physician expert witness ai disclosure", pulled 2026-11-12, plus searches on Rule 707 status and on discovery of expert prompts.
- Rules: Federal Rule of Evidence 702 and its committee notes, and Federal Rule of Civil Procedure 26, from the Legal Information Institute at Cornell.
- Rulings: Matter of Weber from the New York courts' official reporter. Kohls, CLF v. Shell and Villanueva through published summaries by law firms and e-discovery writers that quote the orders, linked on each page. The 3M trial through CBS News reporting.
- Professional guidance: AMA Code of Medical Ethics Opinion 9.7.1, the American College of Surgeons statement on physician expert witnesses, and the UK Academy of Experts AI guidance.
- Products: each vendor's pricing, product or security page, linked in its row with the capture date.
How we chose the tools
A product made the tools table if it does a task a physician expert does (reading the record, finding literature, general drafting help under a BAA) or is sold to the firms that retain experts and names experts as an audience. We left out consumer chatbot plans, which offer no BAA, and expert-finding platforms. We list facts, not scores: price, trial, page citations to your file, BAA status and the catch.
What we did not do
- We did not upload patient records to any product or measure any product's error rate.
- We did not read sealed filings or full dockets; for 3 rulings we relied on summaries that quote the orders.
- We could not read OpenAI's BAA help article on 2026-11-12 (timeouts on every route we tried). The OpenAI row uses a capture of the same page made the same day for a sibling site we publish.
- We did not find a published account of the outcome of the 2026-10-15 evidence rules mini-conference.
- Nothing here is legal advice. Discovery and admissibility rulings turn on the court and the facts.
Synthetic examples
The worked example and the sample log rows are synthetic, with no real patient. Their page counts, times and findings are invented to show the method. They are labelled that way where they appear.
How we handle owning Medrecords AI
This site is published by Medrecords AI, which makes 1 of the products it covers. That is a conflict of interest. We don't claim the site is neutral about it; we try to make it useful anyway, under these rules:
- Medrecords AI is marked "(ours)" wherever it appears and gets the same columns as every other product.
- We never call it the best overall. Where we pick it, we name the use case (large multi-provider files needing a page-cited chronology), and we name the cases where another product fits better, including cheaper ones for small files.
- Facts about Medrecords AI come only from its published pricing page and llms.txt, the same standard as every vendor.
- Links to medrecords.ai appear only on the Medrecords AI page and in boxes marked as ours. Each carries a tag so we can see which visits came from this site.
- Links to every other company carry a nofollow attribute.
- The legal pages (FRE 702, deposition, disclosure) don't depend on which tool you buy, and we wrote them to be useful if you never buy ours.
Corrections
Vendors and readers can send corrections to aj@medrecords.ai. We check the source, fix confirmed errors within 2 business days and update the date on the page.