Beauty copy lives in a narrow lane: claim too little and the page feels flat, claim too much and you trip the US FTC, EU/UK cosmetic-claims rules, or China’s NMPA. These 15 prompts cover the angles a skincare, makeup, or fragrance brand needs — a single hero ingredient, texture and sensory copy, results framed without medical claims, honest clinical context, and the cross-market traps when you localize for the US, UK, EU, and China.
TL;DR
- Give the model the rules and it will draft inside them; the rules are the prompt. Paste your INCI list, a per-market banned-word list, and a real study (N + duration) into every prompt.
- The expensive mistakes are claims, not grammar: “anti-aging,” “cures acne,” and “erases dark spots” read as drug claims in the US, EU, and China.
- As of June 2026, the EU fragrance-allergen list jumps from 26 to 82 substances — new declarations are mandatory on products placed on the EU market from 31 July 2026 (Regulation (EU) 2023/1545). Your copy and ingredient blocks must match the new label.
- A regulatory reviewer signs off the final draft. No model — GPT-5.5, Claude Opus 4.7, or Gemini 3.1 Pro — can certify compliance for you.
Who this is for
Skincare and beauty DTC founders, indie cosmetic brands, marketplace beauty sellers (Amazon Premium Beauty, Sephora, Mecca, Tmall Global), and copywriters at agencies serving cosmetic clients.
When not to use these prompts
Do not use these to write therapeutic or drug claims (treating acne or eczema, “reversing” aging, “erasing” dark spots). Those are regulated and need legal review and clinical substantiation, not AI prose. The US FTC’s Health Products Compliance Guidance requires “competent and reliable scientific evidence” — for a health claim that means randomized, controlled human testing, not marketing copy.
Which model to use
Any current model can run these once you supply the rules. As of June 2026 the practical picks:
| Model | Strength for beauty copy | Plan |
|---|---|---|
| Claude Opus 4.7 / Sonnet 4.6 | Best at holding a long banned-word list and a full INCI in context (1M tokens) without slipping back into “anti-aging” | Claude Pro $20/mo |
| GPT-5.5 (Thinking) | Strong at structured PDP outlines and table output; reliable JSON | ChatGPT Plus $20/mo |
| Gemini 3.1 Pro | Cheapest API for bulk listing runs ($2/$12 per 1M in/out); 1M context | Google AI Pro $19.99/mo |
The differentiator is not the model — it is the system context you paste. Feed any of them the INCI, the claims policy, and a per-market word list and the output converges.
Prompt anatomy / structure formula
A beauty copy prompt should always carry six elements:
- Role: who the AI plays (luxury copywriter / Amazon listing strategist / DTC brand voice / paid-ads hook writer).
- Context: product, brand voice, target buyer, platform, price tier, season — anything that shifts copy.
- Goal: one concrete deliverable — 5 bullets, a 150-word hero, 13 tags, 10 hook lines, a refund reply.
- Constraints: must / must-not (FTC claims, banned words, character limits, tone, no emoji, no superlatives).
- Output format: numbered list, table, JSON, or labeled blocks so you can paste straight into the seller backend.
- Examples / signal: 1-2 reference lines you like, or anti-examples (“not like this competitor”).
Best for
- Skincare PDP hero and bullets
- Makeup launch detail pages
- Fragrance notes and storytelling
- Ingredient-led education content
- Amazon / Sephora / Tmall listing copy
15 copy-ready prompt templates
1. Hero ingredient narrative
Lead with one star ingredient at a named %; everything else is supporting cast.
You are a beauty copywriter for a {category — skincare / makeup / haircare} brand. Write a 90-word hero paragraph for {product}. Anchor on one hero ingredient at {%}, sourced from {origin / supplier}. Explain what it does in plain language. No medical claims (no "cures", "reverses", "treats"). End with one sensory hook (smell, texture, finish).Variables to swap: category, product, hero ingredient, %, origin/supplier
Optimization: If output drifts into clinical territory, add: “Reframe every benefit as a visible / sensory experience, not a medical outcome.”
2. Claims-safe benefit bullets
Write 5 bullet points for {product} that read benefit-led but stay claims-safe. Format: emoji + sensory or appearance benefit + 1 supporting ingredient. Banned words: cure, treat, reverse, eliminate, anti-aging, medical. Use instead: visibly, feels, helps support, appears.3. Texture & sensorial paragraph
Write a 70-word texture paragraph for {product}: how it feels on application, how it absorbs, what it leaves behind. Use sensory nouns ({silk, gel-cream, balm-to-oil, powder-soft}). No "luxurious" or "indulgent".4. Ingredient glossary block
For {product}, generate a 5-row ingredient table: Ingredient — % (if disclosed) — What it does (one verb) — Source / form. Voice: educator, not marketer. Include any "free from" callouts ({silicones, sulfates, parabens, fragrance}).5. Fragrance notes (top / heart / base)
Write fragrance copy for {scent name}: 60-word evocative opener + structured pyramid (Top, Heart, Base) with 2-3 notes each. End with a one-line wear context ({day, evening, summer, winter}).6. Skin-type fit guide
Write a "Who this is for / Who should skip" block for {product}. Best for skin types: {list}. Skip if: {sensitivities, conditions, ingredient allergies}. Voice: honest, no upsell.7. Routine placement copy
Explain where {product} fits in a routine: AM or PM, before / after which steps, pairs well with, do not layer with. 80 words, plain language, no jargon dump.8. Clinical / dermatologist-tested context
For {product} with {N-week clinical / consumer panel}: write a 100-word section that cites the study honestly. Include: N participants, duration, what was measured, % who saw the result. Use "in a consumer study of N" phrasing. No "clinically proven" without citation.Variables to swap: N, duration, measured outcome, %, study type
9. Shade / finish description (makeup)
For {shade name} in {product line}, write 50-word copy covering: undertone, finish ({matte / satin / dewy / luminous}), best skin tones, day vs. night wearability. End with one outfit / occasion image.10. Cruelty-free / clean / sustainability section
Write an 80-word values block for {brand}: cruelty-free status (named cert if any), clean-beauty framework, packaging recyclability (specific PCR %, refill program). Avoid "natural" without definition.11. Before/after-style results (compliant)
Write a results paragraph for {product} based on {self-reported consumer study}: lead with the most-improved metric. Phrasing template: "{X}% of {N} users reported {visible / sensory benefit} after {time}." Add a one-line disclosure ("Results may vary").12. Cross-cultural claim adapter
Each market polices claims differently, so adapt rather than translate. As of June 2026: the EU’s six common criteria (Regulation (EU) 655/2013) ban “free from animal testing” style claims and require every benefit be evidenced; China’s NMPA treats whitening, sun protection, anti-hair-loss, and anti-acne as efficacy claims that need human trials, with “spot whitening” registered as a special cosmetic.
Adapt this beauty description for three markets: US (FTC tone, no therapeutic claims), EU/UK (no implied medical benefit; every claim must be evidenced under the six common criteria), China (NMPA — drop "anti-aging" and "whitening" unless the SKU is registered for that efficacy). Mark each adaptation with the change made and why.
[paste source copy]13. Sensitive-skin reassurance
Write 100-word copy for {sensitive-skin product}. Open with the reassurance, not the hero ingredient. Name what is excluded ({fragrance, essential oils, drying alcohols}). End with patch-test guidance.14. Hero quote / testimonial integration
Take this raw customer testimonial and integrate it into product copy as a 1-2 sentence quote with attribution ({first name, age, skin type}). Edit only for grammar and length, never invent details.
{paste testimonial}15. PDP scroll structure (full page outline)
Use this to draft a whole page, then run earlier templates per section.
Generate a complete PDP outline for {product}: H1 + 1-line subhead, hero paragraph (90 words), 5 benefit bullets, ingredient highlight (3 boxes), texture paragraph, routine placement, results section, FAQ (5 Qs), reviews placement, cross-sell. Label each block with a target word count.Common mistakes
- Using “anti-aging,” “cures acne,” or “erases dark spots” — these read as drug claims in the US, EU, and China.
- Listing 15 actives in the hero. Readers anchor on one; everything else is noise.
- Inventing clinical results (“clinically proven”) without a real study to cite. The FTC standard for a health claim is competent and reliable scientific evidence.
- Skipping the “skip if” section — buyers who react blame the brand publicly.
- Translating “whitening” or “anti-aging” literally for China. Under NMPA, whitening (and spot whitening especially), sun protection, anti-hair-loss, and anti-acne are efficacy claims that require registered human-trial substantiation.
- Mixing fragrance romance with skincare science in the same paragraph.
- Using “natural” or “clean” without defining what the brand means — neither term has a legal definition in the US or EU.
- Shipping fragrance copy that no longer matches the label. From 31 July 2026 the EU requires up to 82 named fragrance allergens on-pack (up from 26) when they exceed 0.001% in leave-on or 0.01% in rinse-off products.
How to push results further
- Feed the model the full INCI and the brand’s claims policy as system context. A 1M-token model (Opus 4.7, Sonnet 4.6, Gemini 3.1 Pro) holds all of it without dropping rules mid-draft.
- Build a banned-word list per market (US, EU, UK, CN) and paste it into every prompt. This single step kills most over-claiming.
- Anchor every result on a real study (consumer panel or clinical) with N and duration; never invent.
- Use “visibly,” “feels,” “appears,” and “helps support” instead of “treats,” “cures,” or “eliminates.”
- For fragrance, give the model a moodboard of 3-5 reference scents before asking for notes — and reconcile any allergen mentions against the 31 July 2026 EU label.
- Include a patch-test line for any leave-on product with actives.
- Have a regulatory reviewer read the final copy before publishing. No model replaces this step.
FAQ
- Can AI write FTC / EU / NMPA-compliant beauty copy?: It can draft within the rules if you give it the rules. It cannot guarantee compliance — a human regulatory reviewer must sign off. The FTC’s Health Products Compliance Guidance sets the substantiation bar for any health-related claim.
- Are “clinically proven” claims safe?: Only with a real, citable study. Otherwise switch to “in a consumer study of N users” phrasing. For a health benefit the FTC expects randomized, controlled human testing, not a marketing survey.
- How do I localize beauty copy for China without losing meaning?: Use template 12. Replace “anti-aging” with sensory or visible benefits, and drop “whitening” unless the SKU is registered for that efficacy. Under NMPA, whitening, sun protection, anti-hair-loss, and anti-acne claims need human-trial substantiation.
- What changed for the EU in 2026?: From 31 July 2026, products placed on the EU market must declare up to 82 named fragrance allergens (up from 26) under Regulation (EU) 2023/1545. If your fragrance copy names notes, make sure it matches the new on-pack allergen list.
- Which model is best for beauty PDPs?: As of June 2026, Claude Opus 4.7 or Sonnet 4.6 hold long banned-word lists best; GPT-5.5 (Thinking) is strongest for structured outlines and JSON; Gemini 3.1 Pro is cheapest for bulk listing runs. The system context you paste matters more than the model.
- Why does AI keep over-claiming?: It mirrors the over-claiming patterns in its training data. Add an explicit banned-word list and the problem mostly disappears.
Related
- Product description prompts
- Amazon bullet prompts
- Landing page section prompts
- Product FAQ prompts
- Luxury Product Description Prompts for Heritage and Craft
- E-commerce & Marketing Prompts hub
Tags: #Prompt #E-commerce #Beauty #Skincare #Product description