Make AI UGC Believable: Precision Prompts and Deliberate Imperfection

What you’ll make

A reusable “realism checklist” and prompt template. The source’s point: generic prompts produce the same polished, invisible output everyone else gets. Specific detail and small imperfections are what make UGC feel real.

Steps

1. Ban generic subjects. Replace “woman with product” with a specific person in a specific moment: hairstyle, clothing condition, where the product is, what expression she’s caught in.
2. Add imperfections on purpose. Messy hair, uneven window light, a wrinkled shirt, clutter in the background, product held at arm’s length.
3. Describe the camera like a phone. Front camera, slight tilt, not perfectly centered, mid-sentence.
4. Build a checklist you run every prompt through (below). Ask Claude to score a prompt against it and fill the gaps.
5. Spend the time on the prompt, not the re-rolls. The source’s argument is that the prompt is the bottleneck, not the model.

Prompt templates

Realism checklist:

Person: age range, hair (state, not just style), clothing (fit + condition)
Moment: what they're doing mid-action, expression mid-word
Product: exact position, hand grip, label facing camera?
Light: source + flaw (window glare, lamp warm cast, overexposed sky)
Camera: front/back phone cam, angle, distance, slight tilt
Background: 2–3 specific, ordinary objects
Imperfection: at least two

Example:

Phone front-camera shot of a woman in her late 20s with a messy bun and a
slightly stretched gray hoodie, holding the serum bottle at arm's length,
mid-sentence half-smile, uneven daylight from a window on the left,
unmade bed and a mug in the background, frame slightly tilted.

Mistakes to avoid

• Prompting for “high quality, cinematic, 8K” on UGC. That’s the opposite of the look.
• Making every frame imperfect in the same way. Vary the flaws.