How this site is researched, reviewed and corrected
The pages here are read by people who have to make a decision that a regulator may later ask about. That sets the standard: primary sources, a named author, a visible date, and a correction route that reaches a human. This page states the rules so you can hold us to them.
Last updated 2026-08-05
How content is researched
Regulatory and compliance content is written from the primary source: the agency, the legislature or the standards body that issued the thing. In practice that means the Centers for Medicare and Medicaid Services, the Department of Health and Human Services and its Office for Civil Rights, ASTP/ONC, the Food and Drug Administration, the National Institute of Standards and Technology, and the state legislatures whose AI statutes now reach clinical work. What that produces is indexed in the regulation tracker, where each page names the issuing document it was written from.
A law firm summary, a vendor blog or a trade publication is not a source. Those are useful for noticing that something moved, and they are where a good deal of research starts, but the claim that reaches a page is checked against the issuing document and cited to it. Where a regulation is genuinely ambiguous, the page says it is ambiguous rather than picking the reading that makes for a cleaner sentence.
Vendor facts are taken from the vendor’s own published material. Where a vendor does not publish something, the page records that it is not published rather than estimating it. “Pricing on request” is a fact; an invented number is not.
Every long-form page carries the sources it was written from, linked and attributed to the publisher, at the foot of the page.
Sources we cite
A sample of the primary sources behind the regulation and compliance pages on this site.
- HIPAA (Health Insurance Portability and Accountability Act)U.S. Department of Health and Human Services
- HIPAA Security Rule guidanceU.S. Department of Health and Human Services
- CMS Interoperability and Prior Authorization Final Rule (CMS-0057-F)Centers for Medicare and Medicaid Services
- Health Data, Technology and Interoperability certification program (HTI)ASTP/ONC
- Artificial intelligence and machine learning in software as a medical deviceU.S. Food and Drug Administration
- AI Risk Management Framework (AI RMF 1.0)National Institute of Standards and Technology
- Colorado SB24-205, Consumer Protections for Artificial IntelligenceColorado General Assembly
- California AB 3030, generative AI in patient communicationsCalifornia Legislature
How often pages are re-checked
Regulation pages are reviewed every 30 days. Rules move, effective dates arrive, and a compliance page that is six months stale is worse than no page at all because it is confidently wrong.
Evergreen pages, which is everything else, are reviewed every 90 days. A review means a person re-reads the page against the current source, not that a build stamped a new date on it.
The last review date is visible on every page that carries one, and it is stored per page rather than generated at deploy time. A site-wide freshness stamp that moves whenever anything ships is not a freshness signal, it is a habit crawlers learn to ignore.
Where a page is reviewed and nothing has changed, the date moves and the content does not. That is a normal outcome and we record it as one.
Where AI is used, and where it is not
Drafting on this site is AI assisted. We advise healthcare organisations on deploying AI agents, and it would be incoherent to pretend we do not use the tools ourselves. Structure, first drafts and rewrites are produced with AI assistance, and that is disclosed here rather than left for someone to detect.
Every page is reviewed by a human before it is published. The review checks the claims against the sources, removes anything that cannot be traced to one, and takes responsibility for what is left. The byline on a page names whoever did that, not the tool that produced the draft. The long-form pages here are written and reviewed by the Clunic research team, whose page lists what it is accountable for.
No statistic is published on the strength of a model’s recall. A figure either carries a citation to the body that published it, or it is labelled as an assumption inside a model whose inputs you can change yourself, or it does not go on the page. This is the rule that AI drafting most often threatens, so it is the rule the review is most careful about.
Nothing on this site is generated at request time. Pages are built, reviewed and published; there is no model writing to a reader in the moment.
Corrections
If something here is wrong, tell us and we will fix it. Email [email protected] with the page and what is wrong with it. A correction that changes the meaning of a page is made in the page itself and the review date moves with it.
This applies to vendors as much as to readers. If we have described a product, a contract term or a certification incorrectly, send the published document that shows it and the page will be corrected. We will not remove an accurate statement because a vendor would prefer it were not there.
What we do not do
The commercial model behind a comparison page determines what the page is allowed to say. Here is ours, and the rule it produces is visible on any of the buyer guides, such as the shortlist of HIPAA compliant AI tools.
No sponsored rankings
No vendor can buy a position in a comparison, a shortlist or a recommendation. There is no rate card, because there is nothing on this site for sale to a vendor.
No affiliate fees
We take no commissions, referral fees or revenue share from any vendor named on this site. Outbound links to vendors are plain links and carry no tracking parameters of ours.
No pay for placement in comparisons
Inclusion in a comparison is decided by whether the product is relevant to the question the page asks. Vendors are not told in advance, and are not given a right of reply before publication.
No statistics without a source
A figure either carries a citation to the body that published it, or is labelled as an assumption in a model whose inputs you can change, or it does not appear.