Ugc Ad Production

UGC Ad Production

Use this skill to plan, script, generate, edit, and review short-form ads that should feel native to creator-led feeds while still being truthful, substantiated, disclosed, licensed, and ready for platform review.

This is production guidance, not legal advice. For regulated products, material performance claims, health/finance/children/political content, synthetic human likeness, or unclear disclosure duties, stop and ask the client for legal/compliance approval before producing or publishing.

Operating principle

UGC ads work because they borrow the cues of spontaneous creator content: handheld framing, first-person speech, product proximity, lived details, casual edits, and plain-language proof. Do not let the style imply a false source, false experience, fake review, undisclosed sponsorship, or unrealistic result.

Treat every UGC ad as four linked artifacts:

1. The creative: hook, story, demo, proof, CTA.
2. The evidence file: substantiation for every objective express or implied claim.
3. The rights file: who appears, what assets are used, where they may run, for how long, and whether paid amplification/whitelisting is permitted.
4. The disclosure file: commercial relationship disclosure, AI/synthetic media disclosure, and platform upload settings.

If any artifact is missing, ship only a clearly labeled draft for client review, not a publish-ready ad.

Fast intake

Capture these items before scripting:

• Product, offer, price, discount terms, guarantee, shipping/availability, promo expiry, landing page URL, and required disclaimers.
• Target buyer: awareness level, pain point, objections, alternatives, trigger event, language/region, age sensitivity.
• Placement: TikTok organic/Spark/In-Feed, Instagram/Facebook Reels or Stories, YouTube Shorts, other; paid vs organic; creator handle vs brand handle.
• Desired format: testimonial, product demo, app walkthrough, founder-style, street-style, problem-solution, before/after, comparison, myth-bust, unboxing, avatar/synthetic creator, voiceover-only, split-screen, or remix of supplied footage.
• Evidence constraints: allowed claims, prohibited claims, study data, customer review source, screenshots, app analytics, product measurements, demonstrations the product can actually perform.
• Creator/persona: real creator, actor, employee, customer, avatar, synthetic human, cloned voice, or faceless POV.
• Rights: talent release, usage term, paid usage scope, exclusivity, music/SFX license, location/property permissions, product/logo permissions, third-party app/site footage permissions.
• Compliance reviewers and escalation path.

Ask for missing evidence before making objective claims. If the client only provides "make it convert," default to subjective, experience-framed copy and product-visible proof rather than measurable outcomes.

Compliance guardrails

Claim substantiation

Documented fact: FTC advertising substantiation policy expects advertisers to have a reasonable basis for express and implied objective claims before dissemination. Objective claims include spoken lines, captions, demonstrations, before/after visuals, product screenshots, superimposed text, and implications created by editing.

Production rules:

• Convert each claim into a row: claim, where it appears, type, evidence, risk, approved wording.
• Separate subjective opinion from objective claim. "I love how light it feels" is subjective; "cuts cleanup time in half" needs evidence.
• Substantiate implied claims created by sequence. A cut from "my skin looked dull" to a bright after shot implies product effect even if the voiceover is vague.
• Avoid absolutes unless the client has strong evidence and platform/legal approval: "guaranteed," "instant," "cure," "risk-free," "best," "#1," "everyone," "no side effects."
• Align ad, landing page, price, terms, demo, and disclaimer. A platform can reject mismatches even when each isolated piece is technically true.
• For before/after, show typicality or limitations. Do not use lighting, posture, retouching, filters, compression, or edit timing to exaggerate outcomes.

Escalate to counsel/client approval for health, medical, financial, earnings, safety, children, political/social issue, comparative superiority, environmental, AI-capability, or "results" claims.

Testimonials, endorsements, and reviews

Documented fact: FTC endorsement guidance treats endorsements as promotional messages consumers are likely to believe reflect someone other than the advertiser, including depictions of an individual's likeness or experience. Endorsements must be honest, not misleading, and material connections should be clearly disclosed.

Production rules:

• Never fabricate a customer, review, star rating, testimonial, expert opinion, or "real user" experience.
• If a creator is paid, gifted, employed, affiliated, or receives a commission, script and placement must make the material connection obvious.
• A testimonial may not claim results the advertiser could not claim directly.
• If the endorser's experience is not typical, disclose the generally expected result or clearly limit the claim.
• Do not suppress or imply review consensus by cherry-picking only favorable reviews in a way that misrepresents consumer sentiment.
• Do not ask an AI avatar or actor to say "I bought this," "this changed my life," "my honest review," or "as a customer" unless that statement is true of the person whose experience is represented.
• If an employee/founder appears in a creator style, disclose or frame the relationship: "I'm on the team," "founder demo," or brand-handle content, not "random customer."

High-risk lines to rewrite:

• "I found this brand and had to tell you" when the creator was briefed or paid.
• "Real customer results" without verified customer permission and evidence.
• "Doctors recommend" without substantiated expert endorsement support.
• "Everyone's switching" without evidence of market behavior.

Synthetic persona, AI, avatar, and voiceover

Documented fact, volatile as of 2026-07-10: major platforms are actively changing AI/synthetic media labeling. YouTube requires creators to disclose realistic AI-generated or meaningfully altered content during upload; TikTok requires labeling for realistic AI-generated images/audio/video and prohibits some impersonation or unauthorized private-person likeness uses; Google Ads documentation says AI label settings for AI-created/edited ad assets are rolling out gradually, with required disclosures in some geographies; Meta labels AI-generated organic content and has separate ad standards and political-ad rules. Re-check platform rules at production time.

Production rules:

• Disclose synthetic humans, synthetic voices, face swaps, voice clones, realistic AI scenes, realistic product demos created by AI, or AI-altered real people when platform rules or consumer understanding require it.
• Do not depict a real private person without permission. Do not clone voices or likenesses without explicit written consent for the specific use.
• Do not use a synthetic persona to create a fake testimonial. An avatar can present a scripted brand message, demonstrate a product, or narrate a hypothetical scenario, but cannot have a personal product experience it did not have.
• Avoid "I" claims for synthetic personas unless the line clearly refers to the presenter role, not real use: prefer "Here's how it works" over "I used this for 30 days."
• Keep provenance records: generation tool, prompt summary, source images/audio, consent, AI label decisions, and where the disclosure appears.
• If a synthetic ad will target or run in a jurisdiction with specific AI-ad rules, require legal review.

Platform policy checks

Always re-check policies in the ad account or platform help center immediately before final export. Treat specs, labels, eligible products, and review rules as volatile.

Baseline checks:

• Product category allowed for the target placement and geography.
• Claims supported and not exaggerated.
• No misleading UI, fake buttons, fake notifications, clickbait overlays, or simulated platform interactions that imply functionality.
• Landing page matches product, offer, price, discount, terms, language, currency, and availability.
• Disclosures are visible/audible enough for the placement and do not disappear too quickly.
• Sensitive attributes are not implied in a prohibited way: avoid "Are you broke?", "Need to lose weight?", "Struggling with addiction?", or other direct personal-attribute callouts unless platform policy explicitly allows the framing.
• Music, creator, likeness, screenshots, trademarks, and product footage are licensed for paid use.

Creative system

UGC ad structures

Pick one primary structure per ad. Native-feeling ads are usually clear, not overstuffed.

1. Problem-solution: "I was doing X the hard way → found Y → proof → CTA."
2. Product proof demo: "Watch this specific thing happen → why it matters → offer."
3. Objection handling: "I thought X would be annoying/expensive/fake → here's what changed my mind."
4. Before/after concept: "Old workflow/state → intervention → realistic new workflow/state." Use only truthful, typical, non-manipulative comparisons.
5. App walkthrough: "Open app → complete task → result → who it's for."
6. Listicle: "Three reasons this is in my bag/workflow/cart."
7. Founder/employee transparency: "We built this because X → here is the demo → launch offer."
8. Street-style or social proof: "People keep asking about X → product reveal → proof." Avoid faking public reactions.
9. Myth-bust: "You don't need X; do Y instead." Avoid unsupported comparative claims.
10. Creator challenge: "I tested it for a day/week." Use only when the test actually happened or is framed as a planned demo, not a testimonial.

Hook frameworks

Write 5-10 hooks before choosing one. Test variants by hook type, not just wording.

• Pattern interrupt: "Stop doing [common behavior] until you see this."
• Specific pain: "If [annoying moment] happens every week, try this."
• Proof-first: "This took [task] from [old state] to [new state] in [observed demo]."
• Curiosity gap: "I didn't understand why people liked this until I tried [specific feature]."
• Contrarian: "The viral version of [category] misses the point."
• Confession: "I almost returned this because [objection], then [turning point]."
• Demonstration cold-open: no intro; start with product doing the thing.
• Price/value: "I stopped buying [alternative] because this covers [specific use]."
• App shortcut: "Here's the fastest way to [task] on your phone."
• Persona mirror: "For anyone who [identity-neutral behavior], this is the fix I wish I had sooner."

Avoid hooks that make the viewer feel accused or targeted for a protected/sensitive trait. Prefer behavior or situation framing over personal-attribute framing.

Script beats

For 9-15 seconds:

• 0.0-1.5s: hook or visual proof.
• 1.5-5s: product/context setup.
• 5-10s: one proof moment or one objection answer.
• 10-15s: CTA, offer, or next step.

For 20-35 seconds:

• 0-2s: hook.
• 2-6s: problem or status quo.
• 6-16s: demo sequence with 2-3 concrete beats.
• 16-24s: proof, differentiator, or objection.
• 24-35s: offer, disclosure reinforcement if needed, CTA.

For 45-60 seconds:

• 0-3s: hook.
• 3-10s: stakes and persona context.
• 10-35s: product walkthrough, proof, or story arc.
• 35-50s: objection handling, social proof, comparison, or terms.
• 50-60s: CTA and final product shot.

Script in columns:

time | spoken audio | on-screen text | shot | proof/evidence | disclosure/disclaimer | notes

Every on-screen text line should either sharpen comprehension or provide required disclosure. Do not duplicate the full voiceover in large text unless captions are intentionally burned in.

Creator shot language

Use organic visual cues intentionally:

• Selfie monologue: handheld, eye-level, imperfect but audible.
• Tabletop demo: hands, product, packaging, close texture/screen interactions.
• POV use: camera becomes the buyer's eyes.
• Split-screen: face reaction plus product/app proof.
• Green-screen review: creator points to product page, receipt, screenshot, or app screen; verify screenshots are licensed/accurate.
• Unboxing: packaging, included parts, first-use sequence.
• Day-in-life insert: product in realistic context, not stock b-roll.
• Screen recording: finger taps or cursor path, no private data, no deceptive UI.
• Receipt/proof insert: only if real and permission-cleared; blur personal info.
• Texture/sound proof: close-up audio/visual sensory cues.

Prefer "messy-real" over "low-quality." Bad audio, unreadable captions, shaky product demos, and dark footage reduce trust and review quality.

B-roll and proof

The proof shot is the ad. Do not hide the product until the end unless mystery is the concept.

Good proof:

• Product visibly solving one task.
• App screen completing one meaningful flow.
• Side-by-side workflow with identical conditions.
• Customer quote with source and permission.
• Measured attribute: weight, size, duration, output quality, ingredient/feature, certification.
• Packaging/label/guarantee shown clearly.

Weak or risky proof:

• Generic smiling lifestyle footage with no product action.
• AI-generated "results" that look like real customer outcomes.
• Dramatic before/after with changed lighting or retouching.
• Unlicensed competitor screenshots.
• Fast-cut claims no viewer can verify.

Captions and on-screen text

Use captions by default for social ads. WCAG guidance treats captions as synchronized text for speech and relevant non-speech audio; W3C contrast guidance uses 4.5:1 for normal text and 3:1 for large text.

Production rules:

• Caption all speech and meaningful sounds if they affect understanding.
• Keep caption chunks short; align to speech; correct auto-caption errors.
• Burn open captions when the placement may autoplay muted or when caption files are not reliable.
• Keep captions and CTAs inside platform safe zones; test in previews.
• Use high contrast, text backing, shadow, or stroke over busy footage.
• Avoid covering product proof, face, price, discount terms, or disclosures.
• Keep claim disclaimers on screen long enough to be read.

Variant matrix

Generate variants as controlled experiments. Change one major variable at a time unless producing a broad exploratory batch.

Minimum matrix for a serious UGC ad set:

| Axis | Variant examples | What it tests | |---|---|---| | Hook | pain, proof-first, objection, curiosity | Thumb-stop angle | | Persona | customer, founder, expert, synthetic presenter, faceless POV | Trust source | | Proof | demo, app walkthrough, testimonial quote, comparison | Believability | | Length | 9-15s, 20-35s, 45-60s | Attention depth | | CTA | shop now, try quiz, download, learn more, claim offer | Intent level | | Caption style | kinetic keywords, full captions, minimal text | Comprehension | | Offer framing | discount, bundle, guarantee, trial, waitlist | Conversion lever |

Name variants with intent: hook-proof-demo-15s-tt-v1, not final_final_3.

For each variant, log:

• Hypothesis.
• Primary audience.
• Claims and evidence.
• Disclosure/AI-label decisions.
• Platform/placement.
• Edit notes.
• Export filename and specs.

Platform delivery notes

Re-check official specs before export; the following were verified on 2026-07-10 and can change.

• TikTok: for commercial content, use the platform's commercial content disclosure setting when promoting a brand/product/service. TikTok's in-feed ad safe zone depends on dimensions, captions, and ad format; use its downloadable safe-zone templates or in-account preview.
• YouTube Shorts ads: Google recommends vertical 9:16 assets for Shorts, says only the first 60 seconds play in the Shorts feed, and recommends social-first creative with sound and text overlays.
• YouTube/Google vide

…

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: calesthio
• Source: calesthio/generative-media-skills
• License: MIT
• Homepage: https://github.com/calesthio/OpenMontage

Install and usage instructions live in the source repository linked above.