Turn one idea into a week of patient-education content

TL;DR: The reason most practice marketing dies isn’t a shortage of ideas — it’s that every piece of content restarts the whole job: research it, write it, check it, format it, five separate times. The fix is structural: verify one idea’s facts once, then have AI fan that single verified core out into a week of formats — an education page, social posts, a newsletter piece, FAQs — under one hard rule: the fan-out never adds a claim. Verification happens once, upstream, by you. Everything downstream is formatting. This guide walks the workflow end to end and gives you the prompt and checklist to install it.

Who this is for: any clinician doing their own marketing who posts in bursts and then goes quiet — the pattern this replaces. You’ll need one drafting engine (Claude or ChatGPT — see the tool map) with its settings done per the Safe Setup Checklist. About an hour a week once installed.

This workflow is designed for a solo clinic — consistent output without a marketing team. The worked example below uses a deliberately generic topic; your version runs on whatever your patients are actually asking.

---

The insight: verification is the expensive step — so do it once

Think about what actually makes clinical content slow. Writing is fast. What’s slow is the part your licence sits behind: checking that every factual statement is true, sourced, and worded without overreach. That’s as it should be — health content is held to a higher standard, and your college holds you to it, not your AI tool.

Most clinicians pay that verification cost per post. Five posts, five rounds of checking, and by Thursday the checking loses to the day job — which is why feeds go quiet.

The workflow inverts it: all the facts live in one small, verified document — the seed — and every piece of the week’s content is generated from the seed and nothing else. One verification pass covers the entire week, because nothing downstream is allowed to introduce anything new to verify. AI does what it’s actually reliable at (restructuring, reformatting, matching a voice — see guide one on why), and never what it’s unreliable at (being a source of facts).

This is also what makes the workflow defensible: the answer to “how do you know your AI content is accurate?” is “the AI never generates a factual claim — it reformats the ones I verified.”

The workflow, end to end

Four steps. The first is the only one that needs your clinical brain at full attention.

Step 1 — Pick one idea and verify it (the clinician step)

The best source of ideas is your own week: the question you answered three times in the treatment room is the question your next patient is typing into a search bar. One idea. Not a content calendar — an idea.

Then do the verification you’d do anyway, once, deliberately: pull the 5–8 factual points you want the content to make, and source each one to something real — a guideline, a review, a reputable clinical reference. Points you can’t source don’t go in the seed. This step is yours; nothing here is delegated.

Step 2 — Build the seed

The seed is a half-page document with four parts:

1. The topic and audience — one sentence each (“plantar fasciitis; adults searching about heel pain; grade-8 reading level”).
2. The verified points — your 5–8 sourced facts, each with its source noted.
3. The boundaries — what the content must not do: no doses, no product recommendations, no cure/guarantee language, no diagnosing the reader. (These go into the prompt too, but stating them in the seed keeps them visible to you.)
4. A voice sample — a paragraph of your own writing, so the outputs sound like you and not like everyone else’s AI content. (There’s a much more powerful version of this — see the end of this guide.)

The seed takes fifteen minutes and is the whole week’s factual universe.

Step 3 — Fan out

Paste the seed into the master prompt (the installable asset below) and generate the week in one pass:

• One education page or blog draft — the cornerstone; thorough, structured, the thing the rest points back to.
• Three short social posts — different angles on the same verified points: a hook-led explainer, a myth-vs-fact, a checklist-style post.
• One newsletter paragraph — for whatever email you send patients or subscribers.
• Five FAQ pairs — for the education page, your site’s FAQ, or replies you’ll reuse.

That’s a week of consistent presence from one sitting — every piece traceable to the same verified core.

Step 4 — Review and ship (the second clinician touch)

Review each output against the seed with three questions: Did anything get added? (Any fact, number, or study reference not in the seed gets deleted, not checked — the rule is no new claims, not new claims verified later.) Did anything important get lost or distorted? Does it sound like me? Then the standard finish: educational disclaimer on anything public, and schedule the pieces across the week.

This review is fast because of the rule. You’re not fact-checking five pieces; you’re diffing five pieces against half a page.

A worked example — the shape of a week

Say the seed topic is plantar fasciitis (the same generic example the tool map’s starter prompt used). The seed holds your sourced points — what it is, who commonly gets it, the general categories of what helps, when imaging or referral is actually warranted. The fan-out returns, from that single seed:

• Monday, the cornerstone: a structured education page — “Heel pain that’s worst with the first steps of the morning: understanding plantar fasciitis” — answering the main question in the first paragraph, then working through the seed points in plain English.
• Tuesday, hook-led: a short post opening on the recognizable moment (“Those first steps out of bed…”) and giving one seed point with a pointer to the full page.
• Wednesday, myth-vs-fact: one common misconception from your treatment-room conversations, corrected using only what the seed supports.
• Friday, checklist-style: “questions worth asking your provider about heel pain” — the decision-oriented format that positions you as the guide, not the gatekeeper.
• Sunday, the newsletter paragraph: the cornerstone’s core idea in three sentences with a link.

Notice what happened to the claims: every piece above makes the same verified points in different shapes. Five pieces of content; one verification pass; nothing for the Wednesday post to get confidently wrong, because it wasn’t allowed to generate anything checkable that you didn’t already check.

Why this stays on the right side of the line

Worth making explicit, because “AI content” and “regulated health profession” collide badly when done casually:

• It’s education, not treatment. Every output describes conditions and options in general terms and routes individual decisions to a consult — the same posture as everything else on this site. The seed’s boundaries section enforces it.
• Every public claim has a source — because it came from the seed, and the seed is sourced. Your college’s advertising standards (no outcome guarantees, no misleading efficacy claims) are baked into the prompt’s constraints rather than remembered at midnight.
• No patient data touches the tool, ever. The idea came from your week, but the seed describes the topic, never a person — the generic-prompt habit from the Safe Setup Checklist. If a real situation inspired the idea, the seed’s version of it is unrecognizable by design.
• You are the reviewer of record. The workflow has a clinician at both ends — sourcing upstream, sign-off downstream. AI occupies only the middle.

---

Install this: the master prompt + weekly checklist

Copy both into your practice docs. The checklist is the routine; the prompt is the engine. Fill the bracketed lines once and reuse weekly.

# [PRACTICE NAME] — One idea → one week (weekly content workflow)
Companion to our AI Use Policy and tool map. Time: ~1 hour, one sitting.

## The weekly checklist
- [ ] ONE idea picked (best source: the question patients asked most this week)
- [ ] Seed built: topic + audience · 5–8 verified points WITH sources ·
      boundaries · voice sample  (facts you can't source don't go in)
- [ ] Fan-out prompt run (below)
- [ ] Every output diffed against the seed:
      nothing added · nothing lost · sounds like me
      (anything not in the seed is DELETED, not checked later)
- [ ] Educational disclaimer on everything public
- [ ] Scheduled across the week — done until next week

## The fan-out prompt
I'm a [PROFESSION] in [REGION]. Below is my SEED document: a topic, my
verified factual points with sources, boundaries, and a sample of my
writing voice.

Produce the following, using ONLY the facts in the seed:
1. A ~600-word patient-education page draft — answer the main question
   in the first paragraph, then clear sections with descriptive headings.
2. Three short social posts on the same points, different angles:
   (a) hook-led explainer, (b) myth-vs-fact, (c) a "questions to ask
   your provider" checklist post.
3. One 3-sentence newsletter paragraph linking to the education page.
4. Five FAQ question-and-answer pairs.

Hard rules:
- Use ONLY facts from the seed. Add NO new claims, statistics, study
  references, doses, or product names. If a section needs a fact the
  seed doesn't contain, write [NEEDS SOURCE] instead of inventing one.
- Plain English at a grade-8 level; technical term in parentheses on
  first use.
- No cure, guarantee, or "better than your doctor" language. General
  education only — individual decisions belong in a consultation.
- Match the voice sample. Keep my phrasing where the seed provides it.
- End each public piece with: [DISCLAIMER]

--- SEED DOCUMENT ---
[PASTE SEED HERE]

The [NEEDS SOURCE] instruction is the quiet workhorse: it converts the tool’s worst habit — inventing a plausible fact to fill a gap — into a visible flag you handle in review.

---

What’s next

The seed’s voice sample is the training-wheels version of something better: a standing practice-brain file — one document that teaches AI your voice, your services, your phrasing, and your boundaries so thoroughly that every draft starts eighty percent done. That’s the next guide, and it’s the mechanic everything else in this library builds on.

---

Sources

This guide teaches a workflow rather than making factual claims about tools or health topics, so it travels light on sources. The framing it relies on is established elsewhere in the library:

• What AI actually is for a clinician — and the one privacy rule that comes first (why drafting engines are reliable for reformatting and unreliable as fact sources)
• Which AI tool for which job — a clinician’s map (choosing the drafting engine; the no-new-claims repurposing rule this guide expands into a full workflow)