Where synthetic creator video genuinely works in beauty, where it gets your ad rejected, and the exact shot list we use. Written so you can run it yourself — and so you know what you're buying if you'd rather not.
Written by Aman Rai · ₹50Cr+ / US$6M+ ad spend managed · 280+ brands since 2020, most of them in aesthetics, beauty and D2C · Last reviewed 3 September 2026
Skincare is the category most people ask us about, and the honest answer is that it's the best and the worst fit at the same time.
It's the best fit because beauty ads are mostly texture and routine. A cream that whips under a fingertip, a gel-to-oil cleanser turning milky, a mist catching light, a hand tapping serum into a cheek at 7am. Those are close, short, product-in-hand shots, and image-to-video models are now very good at them. On Meta and TikTok that kind of footage outperforms polished studio film because it looks like something a customer posted, which is the whole point of UGC.
It's the worst fit because the thing beauty buyers most want to see is a result, and a result is exactly what a synthetic person cannot honestly show. An AI face that has never had a breakout cannot demonstrate one clearing. The moment you script that, you've made a fake testimonial, and both the platforms and the regulators treat it as one.
Generate the routine. Never generate the result.
Three rulebooks matter here and none of them care how the footage was made. The FTC endorsement guides treat an implied personal experience as an endorsement, so an AI presenter saying “I've used this for a month” is a claim about an experience that never happened. Meta's advertising standards restrict ads that assert or imply personal attributes, including skin conditions, and reject before-and-after images that show unlikely results. TikTok's creative best practices push native, front-loaded video and its policies reject exaggerated results the same way.
| What you want to say | Runs? | Say it like this instead |
|---|---|---|
| “Cleared my acne in 3 weeks” | No | “Formulated for breakout-prone skin” and let a real review carry the outcome |
| Before-and-after skin split screen | Rarely | Texture shot of the product, then a real customer photo with consent on the landing page |
| “You have wrinkles” hook to camera | No | “If fine lines are on your list this year” — the viewer, not the presenter, decides it applies |
| “Dermatologist-recommended” | Only if true | Name the dermatologist and the context, or leave it out |
| “Clinically proven” | Only with the study | “In a 4-week study of 40 people…” with the study linked from the page |
| Ingredient facts (niacinamide, SPF 50, fragrance-free) | Yes | State them plainly — ingredient and format claims are the safest material you have |
| “Best serum in India / the US” | No | Specific, checkable comparisons or nothing |
Notice what survives: ingredients, formats, textures, routines and feel. That's most of the ad anyway. The result belongs on the landing page, attached to a real person.
Our general six-shot AI UGC storyboard drops the “proof” shot for beauty, because proof is the one shot we refuse to generate. Five shots, one prompt each.
| # | Shot | What it has to do | Model |
|---|---|---|---|
| 1 | The mirror moment | Someone mid-routine at a real bathroom mirror, already talking. Bad overhead light, wet hair, phone propped on the tap. Earns the second second. | Text-to-video |
| 2 | The swatch or texture | Macro of the actual product: a dot on a wrist, a drag with two fingers, the sheen. Two to three seconds. This is the shot buyers screenshot. | Image-to-video from a real frame |
| 3 | The routine step | Where it sits in the order — after cleanser, before SPF. Hands in frame, cap coming off, product going on. Ordinary, not staged. | Image-to-video |
| 4 | The honest expectation | What it does on day one that you can feel: absorbs fast, no pilling under sunscreen, no scent. Feel is a claim you can make. | Text-to-video |
| 5 | The ask | One line, spoken. Where to buy it — your site, Nykaa, Sephora, Amazon, wherever the click goes. | Text-to-video |
Everything in [brackets] is yours to swap. Both assume you've taken one clean phone photo of the real product on a plain surface first.
Tretinoin, hydroquinone, anything dispensed through a clinic. The buyer needs a clinician and real evidence, and the ad rules are stricter. Use dermatologist-led content instead.
If the entire pitch is a visible change — acne, melasma, hair regrowth — you need real before-and-after material with consent, and a real creator to carry it. AI can do the supporting shots, not the hero.
Foundation and concealer shade matching. Generated skin tones drift, and a wrong shade on screen creates returns. Real swatches on real skin, ideally several.
Cleansers, moisturisers, serums, sunscreen, body care, haircare, lip and eye products sold on texture, format and routine. This is where AI UGC earns its place.
If your product is in the first three boxes, we'll tell you on the call. Running AI UGC on a results-led medical product wastes your budget and risks your ad account, and we'd rather lose the engagement than do that.
We run this inside your own Meta and TikTok ad accounts, so the learnings stay yours.
Every engagement is scoped and quoted on a call, with no fixed packages. Our Meta work for beauty and aesthetics brands is what most of the ₹50Cr+ we've managed since 2020 was spent on, and the AI UGC service page covers the broader format for any category.
Yes, for the parts of skincare that are about texture, routine and feel. Swatches, the way a serum spreads, a morning routine at a real sink, a hand pressing a jar shut. Where AI UGC does badly is results: anything that promises a visible change to skin needs a real person, real time and, on most platforms, a claim you can back up. Use AI for the routine and the product, use real evidence for the outcome.
No, and you should not write scripts that imply it. A synthetic presenter has not used anything. The FTC's endorsement guides treat an implied personal experience as a claim about the endorser, so a line like I have used this for three weeks and my acne is gone is a fabricated testimonial. Write the AI script as a demonstration and an explanation instead, and let real customers supply the experience through reviews and real creator content.
Not for being synthetic. Rejections come from the same places as human-made ads: before-and-after skin images, promises about acne, wrinkles or pigmentation, negative self-perception hooks, and text overlays with claims the landing page does not support. Meta's ad standards and TikTok's creative rules both apply to the words and images, not to how the footage was made. Follow them and AI UGC clears review at the same rate as filmed creative.
We start every product shot from a real frame. A clean phone photo of your actual bottle, tube or compact goes into an image-to-video model such as Kling or Higgsfield, and the model animates that frame instead of inventing packaging. Text-to-video alone gets the bottle roughly right and the label wrong, and in a category where the label is the brand, that is worse than no product shot.
A written shot list per concept, a compliance pass on every script against Meta and TikTok rules, image-to-video product shots from your real packaging, twenty to forty finished variants in 9:16 and 4:5, captions, a naming convention that keeps your reporting readable, and a monthly rewrite of hooks based on what actually won in the account. Every engagement is scoped and quoted on a call, and if AI UGC is the wrong format for your product we say so on that call.
Send me the product and where you sell it. I'll tell you which shots to generate, which to film, and whether this format fits — including when it doesn't.
Message Me on WhatsApp Prefer not to message? Send your details instead →These come straight to me. I reply myself, usually the same working day — WhatsApp is faster if you are in a hurry.