How we pick, and how we handle our conflict
Answer in brief: every pick passes 5 tests scored from the vendor's own public pages, read on 2026-11-12. We sell Medrecords AI, which is our pick in 1 category of 6, so we follow 7 written rules to keep that from tilting the rest.
Limits: Vendor facts come from each vendor's own website (1 from trade press), read 2026-11-12. We haven't run every product, and prices change, so check the vendor before you buy.
The 5 tests behind each pick
| Test | What passes | Why it matters to a physician |
|---|---|---|
| 1. Setting fit | Sold to, or usable by, a physician or clinic team | A tool sold only to law firms or insurers won't fit your workflow or contract |
| 2. PHI position | States a BAA or HIPAA position, for any tool that handles PHI for you | No BAA means no patient data, whatever the features |
| 3. Traceable output | For reading tools: page or sentence citations you can open | You sign the opinion; you need to find the source fast |
| 4. Price visibility | A published price or free tier, or a clear statement of who it sells to | You can budget before a sales call |
| 5. Try first | A trial or free tier on your own work | A demo on the vendor's sample proves little |
When 2 tools pass all 5, the pick goes to the one that fits the most common physician setting for that job. Where no product in a category passes test 4 (coding), we pick on the other 4 and say so. Exceptions are named on the page: ChatGPT Health fails test 2 for physicians and is picked for patients only, with that catch stated first.
Where the facts come from
| Source | Used for | Captured |
|---|---|---|
| Each vendor's own website, help pages or blog | Prices, trials, security statements, who it sells to | 2026-11-12, linked per table row |
| Fierce Healthcare (trade press) | The athenaPatient connector for Claude | 2026-11-12 |
| Google results for 4 searches | Which products to check, and what the ranking pages leave out | 2026-11-12 |
| medrecords.ai/pricing.md and llms.txt | Facts about our own product | 2026-11-12 |
Pages were read with a mix of plain downloads and a web data service. Where a page loaded its prices by script and we couldn't read them, the table says so instead of guessing. We didn't run every product on a test file; the picks rest on published terms, and each page says that.
Our conflict of interest, and the rules we follow
AI Health Studio LLC, doing business as Medrecords AI, publishes this site and sells Medrecords AI, which appears in 2 categories: as the pick for outside file summaries and as a runner-up for chronology. That is a conflict. Our rules:
- A disclosure line sits at the top and bottom of every page.
- Medrecords AI is labeled "ours" every time it's named.
- It is listed only in categories where it does the job. It's absent from scribes, retrieval, coding and patient apps.
- It gets the same columns as every other product, including the catch.
- Where a competitor is cheaper on our own worked example, we print that.
- Links to medrecords.ai appear only on the Medrecords AI page and in marked boxes on 4 pages, each tagged so we can see the traffic came from here.
- No vendor pays to be listed, and any vendor can correct its own row by emailing aj@medrecords.ai; confirmed fixes go live within 2 business days.
We don't claim to be neutral about our own product. We claim to apply the same rubric to it, and to tell you where to buy something else.
How often picks are rechecked
Every table is rechecked at least every 45 days. A pick changes when a product fails a test it used to pass (a BAA withdrawn, a free tier ended) or when a runner-up passes a test the pick fails. Each page's "Updated" date is the last fact check.
Questions about how picks are made
Did you test every product?
No. Picks rest on each vendor's published terms, scored against the 5 tests. Each category page says where we didn't run the product, and suggests what to check in your own trial.
Why is your own product picked at all?
Because outside file review and summary for physicians is the job it was built for, and it passes all 5 tests there. It is not picked, or even listed, in the 4 categories it doesn't serve.
Can a vendor ask to be added?
Yes. Email aj@medrecords.ai. We add products that sell into a physician setting and publish enough to score against the rubric.