Framework · AI-Supported Health Communication

The Orientation-First Editing Framework

A six-stage editorial model for reviewing AI-assisted health content — applied below to one real, unedited AI draft, from first prompt to publication-ready result.

← Back to Research & Portfolio

Illustrative example. The topic (unexplained infertility, first specialist appointment) is used for demonstration purposes only and was fact-checked for illustrative accuracy against public patient-education guidance (ASRM, Mayo Clinic, NHS). This is not medical advice and does not replace guidance from a qualified clinician.

Executive Summary

distinct issues, across five categories, found in one fluent, unedited AI draft — none of which a spell-check, a single read-through, or the AI system itself would have caught.

The draft below was generated from a single prompt on unexplained infertility and reviewed against the same six-stage editorial model applied to every AI-assisted text in this practice — the Orientation-First Editing Framework. What follows is the full diagnostic, stage by stage, and the corrected, publication-ready result.

Editorial Philosophy

Good health communication is not the same discipline as correct health information. A page can be medically accurate and still fail the person reading it — because accuracy answers "is this true," while orientation answers "what does this mean for me, and what do I do now." Both are necessary. Neither substitutes for the other.

AI has changed how fast a first draft can be produced. It has not changed what makes health communication trustworthy: a claim that can be traced to its source, a tone that neither minimizes nor alarms, and a clear boundary between what a piece of content can tell someone and what only a qualified clinician can. AI does not yet reason about any of these three things — it produces text that resembles reasoning about them. That difference is the entire discipline.

This is why AI-assisted editing is treated here as a governance function, not a proofreading step: someone has to decide, deliberately, what a reader is told, what is deliberately held back until they are ready for it, and what is left to a clinician entirely. I call this discipline the Orientation-First Editing Framework — a six-stage editorial model, applied below to one real topic.

— Bettina Müller-Farné

Stage 01 · Input

AI Draft

Diagnosis starting point: single prompt, no follow-up, no editing — the unedited output most content teams are already producing at volume.

AI Draft — UneditedUnexplained Infertility: What to Expect at Your First Appointment If you've been trying to conceive for a while without success, unexplained infertility can be frustrating, but the good news is that most couples in this situation go on to have a baby! At your first appointment, your doctor will run some basic tests to check your ovarian reserve, hormone levels, and your partner's sperm count. Don't worry — these tests are quick and painless. Based on your results, your doctor will usually recommend starting with fertility drugs like Clomid, and most patients respond well within 3 cycles. If that doesn't work, you'll likely move on to IUI, and eventually IVF if needed. Many patients get pregnant within 6 months of starting treatment. Make sure to bring your insurance card and be prepared to discuss your medical history. It's a good idea to write down any questions you have beforehand. Remember, unexplained infertility doesn't mean something is wrong with you — try to stay positive, and trust the process!

This reflects typical single-prompt output, not a cherry-picked worst case — which is precisely why it requires a structural review, not a spot-check.

Stage 02 · Evidence & Accuracy Review

Evidence & Accuracy Review

Diagnosis: claims that sound reasonable but are not sourced — the failure mode AI produces most often and most invisibly.

Unsourced outcome claimRisk: High
"most couples in this situation go on to have a baby" — no timeframe, no source, no acknowledgment that outcomes vary enormously by age and duration of infertility.Intervention: removed; replaced with an appropriately hedged, sourced framing.
Fabricated precisionRisk: High
"most patients respond well within 3 cycles," "many patients get pregnant within 6 months" — no clinical source supports these as general statements.Intervention: invented statistics removed entirely.
False default pipelineRisk: Medium
Clomid → IUI → IVF presented as a fixed, universal sequence.Intervention: reframed as one common pathway among several, dependent on age, cause and clinic protocol.
Undefined jargonClarity: Medium
"ovarian reserve" and "IUI" used with no plain-language definition.Intervention: one-line definitions added inline at first use.

Stage 03 · Patient Comprehension

Patient Comprehension

Diagnosis: language that is technically accurate but still creates avoidable friction for a reader who is not a clinician.

Vague phrasing hides real logisticsClarity: Medium
"some basic tests" glosses over cycle-day-dependent blood draws and a separate semen-analysis appointment for a partner — both of which affect how someone should actually plan the visit.Intervention: replaced with concrete, specific description of what is actually being asked of the reader.
Reading loadClarity: Medium
long, compound sentences and passive constructions throughout.Intervention: shortened, and rewritten to state clearly who does what.

Stage 04 · Orientation Architecture

Orientation Architecture

Diagnosis: where the reader actually is, and what they genuinely need next — the layer AI drafts consistently skip.

Missing emotional acknowledgmentOrientation: High
no acknowledgment that "unexplained" is its own distinct emotional experience — often harder to sit with than a named diagnosis, not easier, because there is nothing concrete to target.Intervention: named directly and validated in the opening lines, without dwelling on it.
Missing boundary statementOrientation: High
no statement of what this text can and cannot tell the reader.Intervention: an explicit "what this can / can't tell you" block added.
Missing genuine next stepOrientation: High
the only "next step" offered was administrative ("bring your insurance card").Intervention: added one genuine, low-commitment next step — preparing a short, focused question list.
Flat narrative structureOrientation: Medium
one continuous narrative, no situational scaffolding.Intervention: rebuilt into a before / during / after structure that matches how someone actually plans for an appointment.

Stage 05 · Editorial Tone

Editorial Tone

Diagnosis: forced positivity and unsupervised clinical framing — removed without swinging into clinical coldness or false alarm.

Toxic-positivity patternTone: High
"the good news is," "don't worry," "stay positive, and trust the process" — phrasing that minimizes a real, valid emotional experience instead of acknowledging it.Intervention: replaced with calm, direct acknowledgment — no false reassurance, no despair framing either.
Unsupervised clinical framingGovernance: High
a specific brand-name medication (Clomid) named as a default — a clinical decision, not an editorial one.Intervention: described by drug class instead of brand, leaving the specific choice to the reader's own clinician.

Diagnostic Summary

Twelve interventions, five categories, one paragraph of AI output

Unsourced or fabricated claims removed

Oversimplified generalizations corrected

Comprehension frictions resolved

Orientation elements added

Tone & governance corrections

Stage 06 · Publication

Publication

Outcome: every intervention from Stages 02 through 05 integrated into one publication-ready piece.

Final, Published VersionUnexplained Infertility: What to Expect at Your First Appointment An "unexplained" diagnosis can feel harder to sit with than one with a clear cause — there's no single thing to target, and that uncertainty is a real part of what you're carrying into this appointment, not something to push past. Before your appointment: your care team will typically want your cycle history and any previous test results, and — if relevant — your partner's semen analysis. Some tests are tied to specific days of your cycle, so it's worth asking in advance whether timing matters for you. At your appointment: expect a review of your history and, often, tests to check ovarian reserve (an estimate of remaining egg supply, usually via blood test and ultrasound) and, for your partner, a semen analysis. These are standard starting points, not signs that something further is already suspected. What happens next varies by clinic, age, and how long you've been trying — there is no single default sequence. Common next steps include ovulation-support medication, intrauterine insemination (IUI), or IVF, often in that general order, but your specific path depends on your situation and should be discussed directly with your care team rather than assumed in advance. What this page can tell you: what a first appointment usually covers. What it can't tell you: how long your specific path will take, or what will work for you — those depend on factors only your care team can assess with you. One concrete next step: write down your cycle history and your top three questions before you go. A short, focused list is more useful in a fifteen-minute appointment than a long one.

What this demonstrates. AI does not replace this work. It changes what the work is: less time producing a first draft, more time ensuring the draft is safe to hand to someone who is scared, exhausted, or deciding under pressure. The Orientation-First Editing Framework is the same six-stage review applied to every AI-assisted text in this practice — one paragraph here, an entire content library in a client engagement.

What This Framework Demonstrates

AI Governance — treating AI output as a reviewed input, not a finished asset
Editorial Risk Assessment — distinguishing high-risk fabrication from low-risk phrasing
Health Information Architecture — restructuring content around a reader's situation, not a topic
Human-Centered Communication — tone correction without minimizing or alarming
Patient Decision Support — building a genuine next step into every piece of content

Related system: this same editorial governance runs behind Praxis Liebenswert's German-language editorial content (Redaktion) — applied there at library scale, not just to one paragraph.

See how this applies to your content

This framework is applied as a service — see AI-Supported Health Communication Workflow on the For Organizations page.

Start a project inquiry →