How content gets made, and reviewed before it ships
Every tool, page, and post follows one path and clears one review. This site funnels toward a regulated telehealth clinic, so every claim is treated as advertising under FTC rules: true, sourced, and never a promise.
The pipeline
From idea to live, in seven stages
- 1 Spot a content gap A blog, tool, page, funnel leaks
- 2 Research AI/Perplexity, PubMed, & other research
- 3 Brief + pick format tool · blog · page · social · partnership
- 4 Draft sourced claims, brand voice, copy best practices
- 5 Compliance review the compliance gate, explained below (WIP)
- 6 Publish final human approval, merge dev → main
- 7 ↻ Improve workflow distribute and repurpose via social, audit the news and space for new content gaps (which feed step 1)
Approved content is re-reviewed on a schedule (default 12 months, sooner if the science changes).
Stage 5, a real one
What a compliance check actually does
A lightweight version of pharma's MLR (Medical-Legal-Regulatory) review.
The clearest way to show it is to run one. Below is the real check on our hardest piece: the "What is enclomiphene?" blog post. It names an off-label drug and ends with a link to Veedma's paid assessment, so the FTC treats the whole page as advertising. That is the deepest review there is.
You point Claude at the draft. It returns a table of findings, claim by claim. A human works each one, then records the call. A sample of what came back:
| The draft said | What the check found | Outcome |
|---|---|---|
| "…raised testosterone into the normal range while preserving the fertility signals." | A drug-efficacy claim. The post links to a paid clinic, so the FTC treats the whole page as advertising. Kept, but attributed to the specific trials and escalated as needing a lawyer, not just a reword. | Tier 3 → counsel |
| A "47% vs 13.8%" adverse-event figure. | Claude could not find this number in the cited study. Cut from the post entirely. | Removed |
| Trial numbers: 604 ng/dL, OR 0.18, +274 ng/dL. | Specific efficacy statistics that each need a real source. Checked against PubMed; the PMIDs were recorded in the log. | Verified |
| "Not FDA-approved, off-label." | A required fair-balance disclosure for any drug. Present and prominent in the lead, the table, and a dedicated section. | Passed |
| Em dashes and a couple of hype words. | Brand-voice and style rules. Em dashes removed; hedge words kept where they keep a claim honest. | Fixed |
What you end up with
The decision gets logged
Every check ends as one dated, named entry in the review log. That entry is the audit trail: what was found, what was cut, and exactly what risk the approver knowingly accepted. The real entry for this post:
"Approved for launch by the owner, 2026-06-09. Maria signed off as the human approver, accepting the FTC/FDA risk. Clinician and healthcare-marketing-attorney review were NOT obtained; the owner chose to proceed without them."
No entry, no launch. A clean AI pass is never the approval. A human signing the log is.