Before-and-after ads are easy to understand because the viewer can see a contrast immediately. They are also easy to get wrong with AI. A generated “after” can improve lighting, pose, skin, body shape, packaging and background at the same time, making the product look responsible for changes it did not cause.
A useful AI UGC before-and-after ad starts with evidence, not generation. Decide what the product can truthfully demonstrate, capture or verify that outcome, then use AI for presentation, creator continuity and supporting shots. If the result cannot be supported, change the concept from a result claim to a clearly labelled demonstration.
Define the change in one sentence
Write the exact comparison before making any images. “This organiser reduces the space taken by these six cables” is measurable. “This changed my life” is vague. “Hair looks smoother immediately after applying the product” is different from “repairs damaged hair,” which implies a stronger and potentially longer-term effect.
Separate three things:
- What viewers can directly see
- What the creator can truthfully say
- What evidence the brand holds for objective claims
The UK CAP Code on misleading advertising says marketers should hold evidence for objective claims and must not materially mislead. It also prohibits fake consumer reviews and requires documentary evidence that testimonials are genuine. An AI actor should not claim a personal experience they never had.
Use the creator as a presenter when necessary: “Here is how the product is used” is safer and clearer than “I used this for two weeks and got this result” when no such experience occurred.
Choose the right kind of before-and-after
Not every product needs a transformation claim. Pick a format that fits what the product actually does.
A physical demonstration works for organisation, cleaning tools, storage products, kitchen accessories and similar items. Show the same objects before and after one observable action.
A styling comparison can show two legitimate choices, such as an outfit with and without an accessory. Avoid suggesting the accessory changed the person's body.
A process comparison can show an inefficient setup followed by a more organised workflow. The product's role should be visible.
A verified customer result can be used only when the customer supplied the real material, gave permission and the claim can be substantiated. Do not generate a fictional customer's face or outcome and present it as authentic UGC.
For cosmetics, wellness, finance and other sensitive categories, review the rules and evidence requirements that apply to the specific claim before production. A disclosure that something is AI-generated does not make an unsupported product claim acceptable.
Lock the comparison before generating creative
The two sides should differ only where the product is meant to make a difference. Keep the camera position, focal length, crop, subject distance and main light as close as possible. Use the same objects and quantities. Keep clothing and grooming stable unless they are part of the stated change.
Create a simple continuity card:
- Camera: fixed, eye-level or top-down
- Framing: exact crop and subject position
- Light: direction, softness and colour
- Product: approved SKU, colour and label
- Before state: what is visibly wrong
- Action: what the product does
- After state: the supported outcome
If the “before” uses dim, flat lighting and the “after” uses warm studio light, the viewer may attribute the visual improvement to the product. Match the conditions even when a more dramatic contrast would be easier to generate.
For a product-led sequence, start with clean references and use the same approved product image throughout. The guide to preventing product drift in AI UGC explains how to protect packaging, colour and proportions across shots.
Build the ad as separate shots
A reliable short sequence can use five clips:
Hook: Show the comparison quickly, but do not hide the conditions.
Before: Establish the original state with enough time to inspect it.
Action: Show the product performing one understandable step.
After: Return to the matched camera setup and show the supported change.
Close: State the product, offer or next step without adding a new claim.
Generate or edit each shot separately. This gives you more control than asking one video model to create the entire transformation. Approve the still frames first, then animate simple actions. If a precise interaction matters, such as a measured quantity or a real product mechanism, filmed footage or a controlled composite may be more dependable than fully generated motion.
The AI UGC product demo workflow covers shot planning and when hybrid footage is the safer choice. For the opening, use the hook variation workflow to test problem-led and result-led versions while keeping the evidence unchanged.
Prompt for continuity, not extra improvement
Tell the model what must remain fixed and name the one permitted change. For example:
Before frame: “Vertical creator-style shot of the approved adult presenter at the same bathroom counter, fixed camera, soft daylight from camera left. Six labelled skincare items are scattered across the counter. Use the exact approved product references. Natural phone-camera realism.”
After frame: “Return to the identical camera, crop, presenter, counter, objects and lighting. The same six skincare items are now stored inside the referenced organiser. Do not change the presenter's face, skin, body, outfit, room, product labels or amount of light.”
Avoid prompts such as “make the after look beautiful” or “dramatic transformation.” They invite unrelated improvements. If only a small region needs changing, use a targeted edit around that region and inspect everything outside it for unintended changes.
Make simulations and AI use clear
If the visual is an illustrative reconstruction rather than a documented result, label it as a demonstration or simulation where viewers can see it. Do not bury the qualification in a caption after a strong visual claim.
TikTok's current AI-generated content guidance requires labels for AI-generated or significantly edited content showing realistic people or scenes. Check the advertising policies for every platform and market where the creative will run, since a platform label and an advertising disclosure serve different purposes.
Keep disclosure readable at normal phone size and on screen long enough to register. “AI-generated presenter” explains the person. “Illustrative demonstration” explains the result. Use both when both facts matter.
Run a claim and continuity check
Watch the final edit once without sound and once audio-only. Without sound, does the visual imply a stronger result than the script states? Audio-only, does the presenter sound like a real customer testimonial? Then inspect the before and after side by side.
Check that the product, quantity, camera, lighting and time period are represented accurately. Confirm that captions match the spoken claim. Remove beauty filters, colour shifts or speed changes that exaggerate the difference. Save the source material, permissions, claim evidence and final approved export together.
The best before-and-after ad is not the most extreme one. It is the one where the viewer can understand exactly what changed, see the product's real role and trust that the comparison was not manufactured for the camera.
Explore AI video generator for the workflow, then create a video in Rasgo.
Create your next Rasgo visual
Turn the ideas from this guide into generated images, creator content, product shots, or video-ready concepts.