How to Brief AI UGC for Conversion, With a Fill-In Brief Template
To brief AI UGC, write down the decisions a creator would otherwise make on set: what the image must get the viewer to do, where it runs, who is in it, how they handle the product, where they are, what changes across the batch, and what must not appear. The fill-in template is right below, followed by a worked example and the reasoning behind each field. One limit no brief fixes: an AI persona can show your product, but it cannot stand in for a real customer's review or result.
6 min read

The fill-in AI UGC brief template
Copy this into a doc or your project tracker. Fill every line; leave nothing to “whatever looks good.”
Brief ID: [date]-[product]-[number]
Action: [click the ad / add to cart / save the post / open the email]
Channel and format: [Meta feed 4:5 / Reels and TikTok 9:16 / product page 1:1 / email header wide]
Product: [name, color, size; label must / need not be readable]
Persona: [AI Expert name; same reference face as previous batches]
Interaction: [holding at chest height / applying / wearing / unboxing / product on counter, not held]
Scene family: [e.g. small apartment bathroom, morning]
Varies across batch: [axis 1, e.g. window light vs warm bulb] × [axis 2, e.g. mirror selfie vs eye level]
Composition: [close-up / medium / full body; eye level / slightly above; empty space top third for text]
Details that make it real: [3–5 objects: towel on the rail, phone on the sink, half-used tube]
Hook being tested: [one sentence a media buyer can judge, e.g. “direct eye contact beats looking at the product on cold traffic”]
Angle (caption/voiceover): [problem / use case / objection]
Do not: [other brands' logos, clutter, forced smile, claims the product cannot support]
Disclosure: [platform AI label on/off; caption wording; who approves]
QA: [same face? product accurate? hook clear in 1–2s? hands and eyes natural? right ratio?]
Next brief changes: [filled in after results]
A filled-in example
Action: click the ad (cold traffic).
Channel and format: Reels and TikTok, 9:16; also export 4:5 for Meta feed.
Product: 750 ml insulated bottle, sage green; logo readable.
Persona: the brand's commuter Expert, late twenties, office-casual.
Interaction: pulling the bottle out of a tote bag, lid closed.
Scene family: train platform and carriage, weekday morning.
Varies: overcast daylight vs carriage fluorescent × over-the-shoulder vs eye level.
Details: laptop sleeve in the bag, earbuds case, folded umbrella.
Hook: “bottle coming out of a bag with no wet patch” beats “bottle on a desk” for the leak-proof angle.
Angle: problem: bottles that leak into your bag.
Do not: transit-operator logos, other drink brands, visible water inside the bag.
Disclosure: AI label per platform policy; no customer-review wording in the caption.
The decisions behind each field
1. The action comes first
An ad image and a product-page image for the same product should look different, because they do different jobs:
- Click an ad: a thumb-stopping frame: person and product clear at a glance, one thing happening.
- Add to cart: product clarity, scale against a hand or body, the product in use.
- Save or comment on a post: feed-native, looks like something a person posted.
- Open or click an email: readable at small sizes on a phone, with the offer obvious.
2. Channel and format before the scene
- Product pages: clean light, product as the hero, several angles in context.
- Meta and Instagram feed: a person in the shot, slightly raw, square or 4:5.
- TikTok and Reels: candid, vertical, nothing important where the platform's buttons and caption sit.
- Pinterest: styled, tall vertical, rich surroundings.
- Email headers: wide, uncluttered, product visible when small.
- Marketplaces: follow the marketplace's image rules first; see the multi-marketplace product photo guide.
3. Persona: cast for the buyer, then keep them
Describe the persona the way you would in a casting brief: age range, style, wardrobe, energy. The AI Expert should look like someone in the buyer's world. Then reuse the same Expert across batches: ppl.studio keeps an Expert's face consistent from its reference, and that consistency is what makes a persona recognizable. If you change an Expert's look, note it in the brief so you can tell drift from a deliberate change. Testing two or three Experts against each other on the same scene is one of the more useful early tests.
4. Scene: a family, with two things that vary
“Kitchen” is not a scene. “Small rental kitchen, laminate counter, morning light from a window on the left, a kettle and a fruit bowl” is. Specific, slightly imperfect details are what make a scene read as real rather than as stock.
Then decide what varies. Specify one light, one angle and one time of day and the whole batch comes out near-identical. Vary everything and you learn nothing. Fix the scene family and vary two axes inside it, such as light and camera angle. Looks (visual presets) are a quick way to keep the rest of the treatment constant.
5. Product interaction: say exactly what the hands do
- Holding: the most versatile. Say where: at chest height, beside the face, in a bag.
- Using: applying, pouring, typing. Helps the viewer imagine the experience.
- Unboxing: good for gifts and kits; only show what is really in the box.
- In the environment: the product sits in the scene, not held. For home goods and decor.
“Holding the vitamin bottle at eye level, reading the back label” gives a far more deliberate image than selecting a scene and a product and hoping. Upload the product first; see the product props guide.
6. Hook and angle: two separate variables
The hook is what the visual and the first seconds are built to test. The angle is the story the caption or voiceover carries: the problem, the unexpected use, the objection answered. Write the hook as a sentence someone can judge after the test. If you change hook and angle at the same time, a winner tells you nothing about which one worked.
7. The do-not list
What must not appear matters as much as what should: other brands' logos on laptops and phones, clutter, awkward hands, the forced stock-photo smile, surroundings in colors that clash with your brand, and anything that implies a claim the product cannot support.
8. Disclosure, decided before you generate
Decide how AI use is disclosed before the batch runs, not after legal sees it live. Meta and TikTok have their own AI-content labels; check each platform's current rules for the placement you are buying. Health, finance, alcohol and anything involving minors usually need a second reviewer. And never write the caption as a customer review: an AI persona is not a customer, and the FTC's 2024 rule on fake reviews and testimonials covers AI-generated ones. The AI content disclosure guide goes through it in detail.
9. QA and naming
Write the QA questions before generating, as five to seven yes/no checks: same face as the reference, product accurate and readable where it should be, inside the scene family, hook clear in the first second or two, hands, eyes and posture natural, right aspect ratio. Then name each approved variant by its brief slot, for example brief-id / persona / product / scene / hook / angle / ratio, and use that name in the ad platform. When one wins, you can see which slot it came from.
10. What the next brief changes
The last field closes the loop. If direct eye contact won, the next brief uses more of it and tests something new. If a brief's month-three version looks like its month-one version, it was a print job, not a test. Use a creative testing structure that changes one variable at a time, and refresh winners with small changes (expression, light, angle) as they show ad fatigue.
Common briefing mistakes
| Mistake | What you see | Fix |
|---|---|---|
| Vague brief | Generic, stock-looking images | Name the interaction, light and three real details |
| No channel | Images that do not fit where they run | Set channel and ratio before the scene |
| Product nearby, not used | Person and product in the same frame, no connection | Say exactly what the hands do |
| One brief for every channel | One placement works, the rest do not | One brief per channel |
| Over-styled persona | Reads as an ad, not as a person | Everyday wardrobe, real rooms, natural expression |
| Disclosure after the fact | Assets pulled after launch | Decide disclosure before generating |
| No “next brief” line | Month three looks like month one | Write what changes once results are in |
Build a brief library
Once a few briefs have worked, stop writing from scratch. Keep reusable parts: two or three briefs per channel, one block per core scene (home office, kitchen, gym, commute), a short description per Expert, and seasonal versions for your key dates. A new brief then becomes mostly a matter of picking parts and filling in the hook. For running briefs every week, see the creative production workflow examples; for video, pair the brief with the UGC video script templates.
What to do next
- Create your first Expert: follow the expert creation guide.
- Upload your products: set up your props library.
- Plan a month of briefs: use the content calendar guide.
- Turn a brief into a sequence: see the storyboard workflow guide for carousels and multi-frame ads.
Frequently asked questions
What should an AI UGC creative brief include?
At minimum: the action the asset must drive, the channel and aspect ratio, the product and how the person interacts with it, the persona, the scene with what varies inside it, the hook being tested, a do-not list, the disclosure decision, and how variants will be named. The template on this page has all of them on one screen.
What is the difference between a hook and an angle in a brief?
The hook is the visual and first-seconds idea the asset is built to test, such as direct-to-camera versus mid-application. The angle is the reason to care that the caption or voiceover carries, such as the problem solved or an unexpected use. Vary them separately, or you cannot tell which one won.
How many variations should one brief produce?
Enough to compare the variable you are testing, and no more than you can review properly and fund in a test. A few variants per value of one variable is a sensible start; add more once you know which variable matters.
Is an AI UGC brief different from a creator brief?
Yes. A creator brief describes an outcome and trusts a person to fill the gaps with taste. An AI brief has to state what a creator would decide on set: exact interaction, light, framing, what is not in the frame. It should also say what varies across the batch, so the variation teaches you something.
Put the brief to work
10 free photos. No credit card. Fill in the template and run your first batch.
Start free with ppl.studioFounder of ppl.studio. Building AI tools for product marketing teams who need visual content at scale without the production overhead.