The AI UGC Brief Template for DTC Marketers

The brief is the single most operationally undervalued artefact in the AI UGC stack. Brands running AI tooling without a structured brief produce variant noise; brands running AI tooling with a structured brief produce variant signal. The difference is not in the model — Veo 3.1, Sora 2 Pro, Kling 3.0 Pro, and Seedance 2.0 will all produce dramatically different output from the same product depending on the brief that fronts the prompt. What follows is a working brief template for AI UGC in DTC wellness, with the structural reasoning for each field and a complete example for a magnesium-glycinate sleep supplement.

The framework is opinionated and built from the practical observation that the briefs producing the best variant cohorts share a specific set of constraints. The brief template that works in 2026 is not the brief that worked for human-creator agencies in 2020.

Quick answer

The brief is the leverage point in AI UGC. Brands typing prompts directly into the tool produce variant noise; brands writing an eight-field structured brief upfront produce variant signal.

• The eight fields: subject and cast, setting and time-of-day, the product moment, voiceover content, hook archetype, pacing and shot count, CTA and on-screen text, brand voice constraints.
• Each omitted field produces a specific failure mode — generic creator, perfectly-lit influencer kitchen, improvised product interaction, plausible-but-non-compliant voiceover, generic montage opener.
• One structured brief generates 20-40 parametric variants by varying single fields (subject, setting, voiceover register, hook archetype).
• Brief authoring takes 30-45 minutes for the first canonical brief, 5-10 minutes per cohort to vary parametrically.
• The AI-tool brief differs from the human-creator brief in that it expects to be obeyed rather than interpreted.

Why the AI UGC brief differs structurally from the human-creator brief

A human-creator brief is a creative-direction document — it sets the tone, gives the creator room to interpret, and trusts the creator's judgement on framing, pacing, voice, and texture. The AI-tool brief is a constraints document — it removes interpretation by specifying the parameters that the model will otherwise hallucinate.

The structural shift is: where a human-creator brief expects to be interpreted, the AI-tool brief expects to be obeyed. The creative-direction layer that lived in the creator's head moves to the brief author's responsibility. Brands running AI UGC tooling without this shift in brief practice produce variant cohorts that are technically completed but creatively homogeneous.

The eight fields below are the load-bearing brief structure. Omit any one and the variant programme degrades in a specific failure mode.

The eight-field brief template

1. Subject and primary cast

Who is on camera, in what demographic, with what archetype. Specify age range (35-45 not "middle aged"), ethnicity if relevant to the audience segment, professional archetype (creator, knowledge worker, athlete, parent), wardrobe register (athleisure, business-casual, founder-uniform).

Failure mode if omitted: the model defaults to a generic 25-35 white female creator in athleisure. Every variant looks the same. The category-specific creator archetypes that drive conversion in collagen, electrolyte, and skincare are not surfaced.

2. Setting and time-of-day

Where the scene takes place, in what light, at what time of day. Specify the room type (kitchen island, home office, bathroom counter), the light source (morning window light, afternoon golden-hour, mid-day overhead), the season cues (autumn jumpers, summer linen) if relevant.

Failure mode if omitted: the model defaults to a perfectly-lit influencer kitchen at indeterminate time. The category's authenticity primitive collapses.

3. The product moment

What happens with the product on camera, in what sequence. For a sleep supplement: pill bottle close-up, two capsules into hand, water glass at bedside, sip, lights-off. For a collagen powder: scoop close-up, pour into coffee, stir, drink. For an electrolyte sachet: sachet held up, pour into water, dissolve, drink. Specify the order and the camera move.

Failure mode if omitted: the model improvises the product interaction and frequently does it wrong — pours from a closed bottle, holds the supplement awkwardly, skips the dissolve. Product credibility collapses.

4. Voiceover content (or none)

The exact voiceover script, or "no voiceover" if the asset is visual-only. Specify the voiceover register (educational, conversational, comparative, founder-POV), the approximate length (15s, 30s, 60s), and the claim boundaries (what can be said, what cannot — particularly important in regulated categories per The AI UGC trust crisis: what the data actually says).

Failure mode if omitted: the model generates plausible-sounding voiceover that may or may not survive a compliance review. Particularly dangerous in regulated categories (fertility, hair-loss, nootropic).

5. Hook archetype (first 3 seconds)

The category-specific hook primitive: the application shot (skincare), the dissolve (electrolyte), the morning-ritual scoop (collagen), the focus-context (nootropic), the pet-reaction (pet supps). Specify the hook archetype and reference one or two ad-library exemplars.

Failure mode if omitted: the model generates a generic montage opener. The first-3-second drop-off rate on Meta is brutal for non-category-specific hooks.

6. Pacing and shot count

How many distinct shots in the asset, at what pace. Most performing UGC formats land at 4-6 shots in a 10-second clip; long-form educational creative lands at 8-12 shots in a 30-second clip. Specify the shot count and any specific cut requirements.

Failure mode if omitted: the model produces a single long-take that does not match the platform-native pacing for Meta or TikTok.

7. CTA and on-screen text

The closing call-to-action (subscribe, shop, learn more, try free), the on-screen text overlay if any, the brand logo placement at end. Specify the text exactly and the placement.

Failure mode if omitted: no CTA, or a generic "shop now" that does not match the brand's voice. On-screen text the model invents may overclaim.

8. Brand voice constraints

The brand-voice document attributes — tone, vocabulary, cinematography preferences, music register, comparable-brand reference set. This is what makes Glossier's variant not look like Drunk Elephant's. Tonic Studio's brand-kit feature implements this structurally; competitors implementing the same primitive include the prompt-engineering layer inside ad-creative tools.

Failure mode if omitted: the model produces creative that is technically completed but not visually distinguishable from the next brand's creative in the same category. The brand-equity carrier collapses.

A complete example: magnesium-glycinate sleep supplement

A worked example brief for a 10-second sleep supplement ad targeting Meta women 35-54.

Subject and primary cast: 38-year-old woman, mixed-ethnicity, hair tied back, no makeup, wearing a linen pyjama set. Calm, slightly tired energy. Speaking to herself, not to camera.

Setting and time-of-day: bedroom side-table, 10:30pm, single bedside lamp casting warm yellow light, the rest of the room in shadow. Bedding visible, neatly made.

The product moment: pill bottle in foreground on side-table, hand picks up bottle, twists cap, taps two capsules into palm, places bottle down. Glass of water lifted, capsules taken with sip. Lamp clicked off. 4 seconds total for the product sequence.

Voiceover content: educational register, 8-second voiceover starting at second 1. Script: "The third night I tried this, I slept through. Magnesium glycinate. The form that actually works." No specific outcome claims beyond the personal anecdote; no quantified sleep-duration claims.

Hook archetype: bedside-routine intimate close-up. Reference: top-performing Meta ads from Olly Sleep, Calm Soothing, and Mag-Easy in Q4 2025 ad libraries.

Pacing and shot count: 5 shots in 10 seconds. (1) Bottle close-up on bedside, (2) Hand reaching for bottle, (3) Capsules in palm with bottle in soft focus, (4) Sip from water glass, (5) Lamp click, room goes dark with afterglow on her face.

CTA and on-screen text: closing 1.5s shows brand logo bottom-third with text overlay "Real magnesium. Real sleep." Closing voiceover: "[brand name] dot com slash sleep."

Brand voice constraints: warm, intimate, no music, no fast cuts, naturalistic lighting. Reference voice: Loftie, Apothékary, Magic Mind. Avoid: high-energy fitness creator pacing, club-music register, hyper-saturated colour grade.

How the brief feeds the variant programme

A single brief in this format generates 20-40 variants by parametrically varying single fields:

• Subject variants: same brief, change age range to 28-38, 45-55. Three variants.
• Setting variants: change to bathroom counter brushing teeth, home office before-bed, hotel-room business-trip context. Four variants.
• Voiceover register variants: educational, conversational, comparative ("I tried four magnesium brands"), founder-POV. Four variants.
• Hook variants: bedside routine, bathroom counter, evening reading. Three variants.

Twenty unique variants from one structured brief, generated parametrically in Tonic Studio's brand-kit-driven workflow. The variant-volume framework that drives the unit economics is mapped in Creative volume economics: AI video and the 25-variant month; the iteration-speed comparison against human creators is in AI video iteration speed vs human creator turnaround.

The discipline

Three operational disciplines separate brands producing variant signal from brands producing variant noise.

Write the brief before opening the tool: brief authoring is a creative-direction job, not a prompt-engineering job. The brief should be written as if briefing a senior creative director, then translated into the tool's prompt fields. Brands typing directly into the tool's prompt input produce the worst output because they skip the brief-authoring discipline.

Re-use the brief across the variant cohort: the brief is the constant; the variant fields are the variables. Changing the brief mid-cohort produces inconsistent output and breaks the creative testing framework documented in The AI video creative testing framework for DTC brands.

Run a brief review on every new product launch: the brief should be reviewed against the brand-voice document and the regulatory constraints at every new product launch. Brands re-using a brief from a previous product line frequently produce off-brand or off-claim creative.

The brief is the leverage point. Brands that get the brief right run AI UGC tooling at the unit economics the technology promises. Brands that get the brief wrong run AI UGC tooling at higher cost than human-creator procurement, because they generate 200 variants and use 4. The discipline pays back at every variant beyond the tenth, and it is the operational separator in 2026.

Frequently asked questions

Why is the AI UGC brief different from a human-creator brief?

A human-creator brief is a creative-direction document — it sets the tone, gives the creator room to interpret, and trusts the creator's judgement on framing, pacing, voice, and texture. The AI-tool brief is a constraints document — it removes interpretation by specifying the parameters that the model will otherwise hallucinate. The structural shift: where a human-creator brief expects to be interpreted, the AI-tool brief expects to be obeyed. The creative-direction layer that lived in the creator's head moves to the brief author's responsibility.

What goes wrong if I skip the brand voice constraints field?

The model produces creative that is technically completed but not visually distinguishable from the next brand's creative in the same category. Glossier's voice ends up looking like Drunk Elephant's; Drunk Elephant's looks like The Ordinary's. The brand-equity carrier collapses across the variant set. Brands skipping the brand-voice field produce homogenised slop at AI-tooling unit economics — which is worse than producing fewer variants at higher unit cost through human-creator agencies. Brand-voice encoding is the load-bearing field.

How long does brief authoring take vs typing prompts directly?

A first canonical brief takes 30-45 minutes to author properly. Each subsequent variant cohort (varying single fields like setting, creator archetype, voiceover register) takes 5-10 minutes. Typing prompts directly takes seconds per attempt but produces lower-quality variants and frequently requires regeneration — the per-usable-variant time often exceeds the structured-brief approach within 3-4 generations. The break-even on brief authoring is at roughly the fifth variant, and the brief approach scales asymptotically better past 10 variants.

Can I reuse the same brief across product launches?

Partially. The brand-voice constraints, the audience archetype, and the hook archetype usually carry across launches. The product moment, the voiceover content (especially the claims), and the on-screen text need re-authoring per product. Brands launching multiple SKUs in the same product line (e.g. a flavour expansion for an electrolyte brand) reuse 5-6 of the 8 fields and re-author 2-3. The reuse efficiency is what makes the brief-template approach scalable past the first canonical product.

How does the brief change for compliance-sensitive categories?

The voiceover-content field carries materially more constraint specification — explicit claim boundaries (what can be said, what cannot), citation requirements, regulator-specific language. The on-screen-text field needs the same treatment. For fertility, hair-loss, and nootropic, brands run a compliance counsel review on the voiceover-content field of every canonical brief, then re-use the approved language across the variant cohort. The brief is the document that survives compliance review; the variants are the parametric applications.