Governance
AI Use and Content Production Policy
Artificial intelligence tools may assist with drafting, structuring, and checking content on this site, but no factual claim is published on the authority of a model. Every clinical claim traces to FDA labelling or a named trial, every commercial claim to a dated capture, and medical content is reviewed by a named clinician before publication.
Key takeaways
- AI tools may assist with drafting and structure; they are never the authority for a factual claim.
- Every clinical claim traces to FDA labelling or a named trial with verified identifiers.
- Every commercial claim traces to a dated first-party capture from a named source.
- Model-generated citations are never published without independent verification of the identifier.
- Medical content is reviewed by a named, NPI-verified clinician regardless of how it was drafted.
| AI-assisted drafting | Permitted |
|---|---|
| AI as source of fact | Prohibited |
| Citation verification | Every identifier checked against ClinicalTrials.gov or PubMed independently |
| Medical review | Required regardless of drafting method |
| Mass generation | Prohibited — every indexable page must add original information |
Why is model-generated citation the specific risk?
Because language models produce plausible identifiers. A fabricated PMID or DOI looks exactly like a real one, and a reader who does not click through cannot tell the difference. On this site every identifier is checked against the registry or PubMed before publication, and where that check has not been completed the numeric claim is withheld and the page says Verification Pending.
What about mass-generated pages?
They are prohibited. Producing many near-identical pages by varying a drug name, provider name, or query phrasing is the dominant failure mode in this category, and it is tested for: pages above a meaningful-similarity threshold fail the build rather than being published.
What controls are applied?
- Independent verification of every trial identifier before publication.
- Named clinical review of medical content.
- Automated duplicate-similarity testing across all indexable pages.
- Fail-closed handling — unverified facts are withheld, not smoothed over.
- No hidden text, invisible keywords, or instructions directed at AI systems.
How much of this site's data is actually verified?
| Group | Records |
|---|---|
| Trials with full identifiers | 8 |
| Trials partially verified | 1 |
| Trials withheld pending audit | 1 |
| Providers with verified all-in price | 5 |
| Providers excluded with reason | 19 |
| Evidence status | Definition | What may be published |
|---|---|---|
| Verified | Checked against a primary source with a recorded date | The claim, with its identifier and date |
| Partially verified | Some elements confirmed, others outstanding | Confirmed elements only; the rest labelled |
| Verification Pending | Not yet checked against a primary source | No factual assertion — the gap is shown |
What does this page cover, and what does it deliberately leave out?
This page addresses Human verification, medical review and anti-fabrication and provenance, organised around the primary question of AI medical content policy. Each of those elements is treated separately below rather than blended, because they carry different evidence weights and a reader is entitled to know which parts rest on randomised data and which rest on a captured commercial claim or a regulatory document.
| Element | Treatment here | Evidence basis |
|---|---|---|
| Human verification | Covered on this page | Regulatory or policy source |
| Medical review | Covered on this page | Regulatory or policy source |
| Anti-fabrication and provenance | Covered on this page | Regulatory or policy source |
| Individualized clinical instruction | Deliberately not covered | Belongs with the corrections process |
What are the limits of what this page can tell you?
Every page on this site rests on a specific published policy, and that record has boundaries worth stating plainly rather than leaving a reader to discover them. The limitations below are specific to the material presented above.
- The evidence here describes groups, populations, or captured records — it does not describe you, and no page can substitute for the corrections process.
- Figures carry the date on which they were verified. In a market where terms change frequently, an undated figure functions as a claim about the present that nobody has checked.
- Elements marked Verification Pending are genuinely unknown to this publication rather than merely omitted for brevity, and should not be inferred from surrounding content.
- Where a source conflicts with another, this site shows the conflict rather than resolving it, which means some questions are left open on purpose.
What would change the conclusion on this page?
This conclusion is held open to the following evidence:
- New primary evidence bearing directly on AI medical content policy.
- A change to FDA labelling affecting any statement made above.
- A verified correction submitted through the corrections process and accepted on the evidence.
- A material change to a captured record, including a price, term, or regulatory status.
- Completion of a verification currently marked pending, which would replace a gap with a stated fact.
Frequently asked questions
Do you use AI to write content?
AI tools may assist with drafting and structure, but no factual claim is published on a model's authority.
How do you prevent fabricated citations?
Every identifier is checked against ClinicalTrials.gov or PubMed independently; unverified ones are withheld.
Do you generate pages at scale?
No. Similarity testing fails the build for near-duplicate pages.
Is there hidden content aimed at AI systems?
No. No hidden text, invisible keywords, or ranking instructions are used.
Change history
| Date | Change |
|---|---|
| 2026-07-22 | Page published with current dataset snapshot. |
Dates change only for substantive edits, never for cosmetic changes. Corrections: corrections policy.