How to Use AI for a Dermatology Practice in 2026: Medical + Cosmetic Playbook
Published May 5, 2026 · 14 min read · Happycapy Guide
TL;DR — for the practice owner / MD
- The two highest-ROI AI wins in dermatology are ambient scribing for medical-derm visits and AI-drafted biologics prior authorizations. Together they save 45-75 min/provider/day and cut PA cycles from ~10 days to ~3.
- AI lesion-analysis tools (DermAssist, SciBase Nevisense, DermEngine/MoleScope, VisualDx) are decision-support aids, not diagnoses. A licensed dermatologist must close every loop.
- Cosmetic marketing: never generate synthetic before/after photos. The FTC Endorsement Guides and state truth-in-advertising laws treat this as deceptive. Use AI for captions, scheduling, and de-identification only.
- PHI goes only into HIPAA-BAA tools (ModMed EMA Dermatology AI, Nextech AI, EZDerm AI, DeepScribe, Abridge, VisualDx enterprise). Never into consumer ChatGPT/Gemini/Claude.
- Owner rule: every AI-drafted note, letter, or marketing claim is reviewed and signed by a human clinician before it leaves the building.
Why dermatology is different from general medicine
Dermatology sits in a peculiar position in 2026. You are simultaneously running a medical specialty (biopsies, Mohs, biologics, pediatric derm, inpatient consults) and a retail cosmetic business (Botox, fillers, lasers, peels, body contouring, skincare lines). Each side has its own regulatory floor — and both are being rapidly rewritten by AI vendors, payers, and enforcers.
That means you have to hold two mental models at once: a clinical compliance floor (HIPAA, state medical board, FDA device rules, Medicare PFS), and a retail compliance floor (FTC Endorsement Guides, state truth-in-advertising, FDA OTC cosmeceutical claims, state medspa rules). AI tools cross both.
The compliance floor (read this first)
- HIPAA Privacy & Security Rules: no PHI into any tool without a signed BAA. Covers EMR, scribe, PA drafting, photo storage.
- State medical board rules: CA MBC, TX TMB, FL BOM, NY OPMC, and others have issued 2024-2025 AI guidance — supervision, informed consent, and scope-of-practice limits.
- FDA device framework: DermAssist, Nevisense, DermEngine-class tools are cleared as CADe/CADx decision-support, not replacements. Mirror vendor IFU exactly.
- Medicare PFS + derm CPT: biopsy 11102-11107, shave 11300s, destruction 17000-17004/17110-17111, Mohs 17311-17315, path 88305. AI cannot code unreviewed.
- FTC Endorsement Guides (2023 revision + 2024-2025 updates): material connections disclosed; typical results depicted; no fabricated before/after imagery or AI-written fake testimonials.
- FTC Act §5 UDAP: covers "clinically proven," "FDA approved," "permanent," and similar superlatives — AI copy must be verified.
- State truth-in-advertising: CA B&P §17500, FL §817.06, NY GBL §349/§350, TX DTPA §17.46 — applied to cosmetic claims.
- State medspa / supervision rules: who can inject, who can fire a laser, who can delegate — AI intake must route correctly.
- FDA OTC monograph + cosmeceutical claims: drug vs. cosmetic line — AI-written product descriptions cannot claim structure/function without substantiation.
- Anti-Kickback / Stark: watch AI-generated "recommended products" or referral language in patient communications.
The dermatology AI stack in 2026
- Derm-specialty EMRs with native AI: ModMed EMA Dermatology AI, Nextech AI, EZDerm AI, EMA Derm, PatientNOW (cosmetic-heavy), Nextech IntelleChart.
- Ambient scribing: DeepScribe, Abridge, Suki, Heidi, Nuance DAX Copilot — all HIPAA BAA, chart-aware, derm-tuned where possible.
- Clinical decision support: VisualDx (differential + image library), DermEngine + MoleScope/DermLite, DermAssist (Google Health), SciBase Nevisense (EIS), 3Derm.
- Pathology & Mohs: PathAI, Paige.AI (research-grade), Inspirata, plus your EMR's Mohs module with slide tracker.
- Cosmetic imaging & consult: Canfield IntelliStudio + VISIA + VECTRA (3D), Quantificare LifeViz, with patient consent and HIPAA-compliant storage.
- Front desk & patient comms: Klara, Weave AI, Phreesia AI, Doctible, RevenueWell — intake, recall, consent forms, post-op check-ins.
- Marketing: Jasper / Copy.ai for draft-only copy (never fabricated testimonials), Canva Magic, Later/Hootsuite AI — all human-reviewed pre-publish.
- Biologics & PA automation: CoverMyMeds, Glidian, Myndshft — AI-assisted drafting; clinician still signs.
10 copy-paste prompts for a 2026 derm practice
Use these only inside a HIPAA-BAA tool. Replace bracketed placeholders with real values. Every output must be reviewed and signed by a licensed clinician before it enters the chart, goes to a payer, or reaches a patient.
1. New-patient intake synthesis (medical derm)
You are a dermatology MA. Summarize this new-patient intake for [Dr. ____] before a 15-min
medical-derm visit.
Inputs: chief complaint, HPI, PMH, medications, allergies, family hx melanoma / psoriasis /
atopy, sun-exposure hx, tanning-bed hx, Fitzpatrick, prior biopsies, prior biologics.
Output:
1) 3-line clinical snapshot
2) Red flags (ABCDE, new/changing lesion, immunosuppression, family hx melanoma <50)
3) Candidate differentials to consider
4) Relevant prior therapies + response
5) 3 open questions the MD should ask
No diagnosis. Flag any missing intake fields. Do not paste into external tools.
2. Dermoscopy / lesion triage second-opinion
You are a dermatology decision-support assistant. I am a board-certified dermatologist. I
will paste a structured lesion description (location, size, color, pattern, dermoscopy
features) from a DermLite / MoleScope capture stored inside [EMR].
Return:
1) Top 3 differentials with supporting features
2) Features that argue AGAINST each
3) Recommended next step (observe / photograph / biopsy type / refer)
4) Any guideline reference (AAD 2025 pigmented-lesion, ACS melanoma)
You are NOT making a diagnosis. I am. Do not output a probability of malignancy as a
number.
3. Biopsy / procedure note draft
You are a derm scribe. Draft a procedure note for a [shave / punch / excisional] biopsy.
Inputs: site (with anatomic laterality and measurement from landmark), size (mm), depth,
anesthetic (agent + volume), technique, hemostasis, closure, specimen to path, patient
tolerance, post-op instructions given.
Output:
- CPT suggestion (11102/11104/11106 for shave/punch/excision as applicable)
- ICD-10 candidates
- Note in SOAP-lite format for EMR
- Post-op handout bullets for the patient
- Path requisition text (clinical context + rule-outs)
Flag any missing element for the MD to complete before signing.
4. Mohs case coordination brief
You are Mohs coordinator support. Build a pre-op brief for tomorrow's Mohs schedule.
Inputs per case: patient, age, tumor type (BCC/SCC/melanoma in situ per MART-1 if
applicable), anatomic site, prior biopsy path, clearances meds anticoag, expected complexity,
repair contingency (primary / flap / graft / referral to plastics).
Output per case:
1) One-line summary
2) Red flags (anticoag not held as ordered, prior radiation, peri-ocular/peri-auricular,
aggressive histology)
3) Repair plan A/B
4) Items the team must stage before patient walk-in
5) Estimated stages
No Mohs appropriate use criteria (AUC) override — MD confirms indication.
5. Pathology report plain-language summary
You are a derm scribe. Given this dermatopathology report inside [EMR], produce a
plain-English summary for the patient chart message at an 8th-grade reading level.
Include:
- What the biopsy showed in 2-3 sentences
- Whether margins are clear / involved
- What happens next (re-excision, Mohs, observation, topical, referral)
- Next visit timeline
- One reassurance sentence if benign
Do NOT diagnose beyond the path report. Do NOT promise outcomes. Do NOT include pricing.
MD reviews and signs before message is sent. FTC/state disclosure: this is a clinical
communication, not marketing.
6. Biologics prior-auth letter (psoriasis / AD / HS)
You are a PA drafting assistant. Draft a medical-necessity letter for [drug] for a patient
with [dx: plaque psoriasis / atopic dermatitis / hidradenitis suppurativa / etc.].
Inputs: PASI/EASI/BSA/DLQI/HiSCR scores, affected anatomy, QoL impact, prior therapies
and failure/intolerance/contraindication with dates, labs (QFT-TB, HBV/HCV, CBC, LFT),
vaccine status, comorbidities (PsA, IBD, uveitis).
Output letter sections:
1) Patient + dx + ICD-10
2) Clinical severity with objective scores
3) Step-therapy documentation (drug, duration, outcome, why failed)
4) Why requested agent is medically necessary per payer criteria + guideline (AAD/NPF
2020+ psoriasis; AAD/AAAAI atopic derm; NAMS/USHHS HS 2019; etc.)
5) Safety screen complete
6) Requested dose + loading + maintenance
Flag anything missing. MD signs. Do not submit AI output unreviewed.
7. Cosmetic consult write-up (FTC-safe)
You are a cosmetic-derm scribe. Draft a consult summary for [patient] for a [neurotoxin /
filler / laser / body contouring / chemical peel] consult.
Include:
- Goals in patient's own words
- Clinical findings (Glogau, Fitzpatrick, Baker-Gordon if peel, FACE-Q domain if relevant)
- Plan with anatomic areas and approximate units/ml/pulses
- Realistic outcome expectations (ranges, not promises)
- Material risks discussed
- Alternatives declined
- Photography consent status
- Pricing / package presented
FTC-safe language rules:
- No "guaranteed," "permanent," "clinically proven" unless substantiated
- No comparisons to physicians outside this practice
- No patient testimonials inserted
- If results will be shared on social, confirm separate written media release is on file
8. Before/after photo + caption (real patient, consented only)
You are a cosmetic-derm marketing writer. I will paste a real patient's consented before
and after photo metadata (never the images themselves into an unapproved tool) plus their
treatment summary.
Draft 3 caption options (Instagram, TikTok, practice website) that:
- Disclose the treatment type and provider name
- State "individual results vary; typical results [describe]" when showing any result
- Do NOT promise outcomes
- Do NOT imply the provider is the "best" without objective basis
- Include a material-connection disclosure if the patient received any compensation,
discount, or complimentary service
- Comply with state truth-in-advertising (CA B&P §17500, FL §817.06, etc.)
- Never use AI image enhancement, skin smoothing, or synthetic generation on clinical
before/after photos
Flag any wording that might trigger FTC Endorsement Guides scrutiny.
9. Recall + follow-up outreach (HIPAA-safe)
You are the front-desk coordinator. Draft recall messages for:
Segment A: total-body skin exam due in 30-60 days, prior hx of skin cancer
Segment B: biologic infusion / injection refill due, last labs <6 months
Segment C: cosmetic post-op day 7 / day 30 / day 90 check-in
Segment D: annual cosmetic touch-up recall at 10-12 months
Constraints:
- No diagnoses or treatment details in SMS (TCPA + HIPAA minimum necessary)
- Two-way conversation through portal or secure messaging only
- Opt-out language in every SMS
- Never disclose "biologic," "biopsy," "skin cancer" in plain SMS
- Call-back script for patients who reply "unsubscribe" vs. "reschedule"
Output each segment as: subject, SMS (<=140 char), portal message, call-back script.
10. Owner weekly scorecard prompt
You are my practice analyst. Read this week's export from [EMR + PM] and produce a
one-page owner scorecard covering:
Medical derm: new patients, established, TBSE volume, biopsies, biopsy-positive rate,
Mohs stages, biologic starts, PA approval rate, PA turnaround days, no-show %, wRVU /
provider / day.
Cosmetic: consults booked, consult-to-treat conversion, avg ticket, injector utilization,
laser utilization, membership adds / churn, post-op complication rate, 90-day retention.
Financial: collections, AR >60, denial rate top 5 codes, referral mix.
Compliance watch:
- Any AI-drafted note signed without edit (flag %)
- Any marketing asset used without human sign-off
- Any BAA lapses / new vendors
Surface 3 wins, 3 risks, and 3 decisions I must make by Monday. No fluff.
Common mistakes that cost derm practices money (and licenses)
- Pasting patient photos or PHI into consumer ChatGPT / Gemini / Claude. HIPAA breach. Use only BAA-covered enterprise tiers.
- Letting the scribe sign notes. Every note must be reviewed and edited by the billing clinician. Unreviewed AI text has produced state-board complaints.
- Synthetic or AI-enhanced before/after photos. FTC enforcement priority. State AGs in CA, NY, TX, FL are active. Assume every piece of marketing may be audited.
- AI-written patient testimonials. Per-post FTC exposure plus state bar-style medical board exposure. Never do it.
- Treating an AI lesion-analysis score as a diagnosis. Document the clinician's independent judgment separately. Mirror the FDA IFU language.
- Using AI to auto-code biopsies or Mohs without review. Medicare audit risk. Keep coder + physician sign-off in the loop.
- Forgetting the cosmeceutical line. AI-written product descriptions can accidentally state structure/function claims ("restores collagen"). That converts a cosmetic into a drug in FDA's view.
- Skipping state medspa/delegation rules. AI intake bots that route cosmetic bookings to non-physicians without required supervision can put the medical director's license on the line.
- No AI governance doc. State boards and payers increasingly ask: "Who approved this tool? Where is the BAA? Who monitors output?" Have a 2-page policy.
- Ignoring FRE 901 / authenticity for clinical photos used in court. Derm photos turn up in PI, workers' comp, and malpractice cases. Maintain chain-of-custody for any photo an AI tool has touched.
A 60-day rollout that does not blow up the practice
Do not boil the ocean. Layer AI in four two-week sprints, verifying compliance and ROI at each step.
- Days 1-14 — Governance + ambient scribe pilot. Sign BAAs with EMR AI, scribe, and any decision-support vendor. Write a 2-page AI governance memo. Pilot ambient scribe on 2 providers, medical-derm visits only.
- Days 15-28 — Biologics PA automation. Route PAs through an AI-drafting workflow (CoverMyMeds + your EMR's AI or a specialty pharmacy portal). Measure cycle time + approval rate weekly.
- Days 29-42 — Cosmetic ops + FTC guardrails. Deploy consult-summary prompt + caption-draft prompt. Train marketing on synthetic-photo ban. Have attorney or compliance officer sign off on social-media policy.
- Days 43-60 — Owner scorecard + lesion decision-support. Stand up the weekly scorecard. Introduce VisualDx / DermEngine / DermAssist as second-opinion in the exam room (not in marketing). Review outcomes; drop any tool that is not earning its keep.
Want a full operator-level AI playbook tuned to your practice?
Happycapy publishes weekly playbooks like this across medical and cosmetic specialties — compliance-first, vendor-agnostic, and written for the clinician- owner who actually has to sign the notes.
Browse more playbooks →