10 guides breaking down the systems behind our results.
Every top-performing ad has a structure underneath it. A hook type, an emotional trigger, a proof mechanism, a close. Once you understand the structure, you can replicate the logic without copying the creative. Here's the exact workflow.
Open Meta Ad Library (facebook.com/ads/library). Search your competitor's brand page. Filter to "Active" ads, anything still running after 3+ weeks is a signal it's working. Screenshot the copy, or if it's a video, note the first 3 seconds and the CTA.
Open Claude and paste this prompt:
Claude will give you the structural DNA of the ad. Now ask: "What is the underlying framework, stripped of brand-specific details? Give it to me as a reusable template."
Now prompt:
Ask for 3 variations: one with a curiosity hook, one with a transformation hook, one with a social proof hook. You now have a testing slate in under 10 minutes.
Brief your creative team or a UGC creator with the scripts. Send static versions to a designer. Test all 3 in one creative testing batch. The one with the best hook-to-watch-time ratio gets scaled.
Running this process monthly against your top 3 competitors gives you a constant stream of battle-tested creative angles, without starting from scratch.
Most brands treat script writing as a bottleneck. A copywriter spends a day on 3 scripts. A creator goes off-brief. The ad underperforms. Repeat. We broke this cycle by building a script machine, a repeatable AI workflow that produces volume without sacrificing quality.
Three inputs feed into Claude, which outputs scripts. The three inputs are: (1) Brand voice doc, (2) Winning angle bank, (3) Format template. Set these up once. Then scripts take minutes, not hours.
One-time setup, 30 minutes. Document: tone of voice (e.g., "direct but warm, no corporate speak"), words to use, words to avoid, brand story in 100 words, core differentiators, main objections. Paste this into a Claude Project or save it as a reusable prompt prefix.
An angle is a specific reason someone would buy. Not "it tastes great", that's a feature. An angle is "You've tried every supplement and nothing works because they're all the same formula, ours isn't." Keep a running list of 20-30 angles, pulled from: customer reviews, competitor ad comments, Reddit threads, your own DMs.
The 5 formats we use most:
Run every output through one filter: would a real person say this? If it sounds scripted, ask Claude to "make it sound more off-the-cuff and conversational, like a voice note."
Once your inputs are set, generating 40 scripts takes under 2 hours. Batch them by angle, 8 scripts per angle, 5 angles per week. Each angle gets one format variant per week, rotating the 5 formats above. The result is a full creative testing matrix, every single week.
The manual process of checking Meta Ad Library weekly is fine when you're small. But at scale, you want this happening automatically. Here's a no-code setup using Make.com and Claude to monitor, analyze, and brief creative based on what your competitors are running.
A workflow that: (1) scrapes new active ads from a competitor's Meta Ad Library URL, (2) sends them to Claude for analysis, (3) outputs a creative brief, (4) drops it into a Notion database or Google Sheet for your team to review every Monday.
Use Apify's Meta Ad Library scraper actor. Set it to run weekly, targeting your competitor's ad account. Output fields: ad text, image URL, status, first seen date. Connect the Apify run output to a Make.com webhook trigger.
In Make.com, add a filter module: only continue if "First Seen Date" is within the last 7 days. This keeps the workflow focused on new creative, not old campaigns you've already analyzed.
Add a Make.com HTTP module pointing to the Claude API. System prompt: "You are a creative strategist. Analyze ad creatives and extract their persuasion framework." User prompt:
Add a Notion module to create a new database entry with: competitor name, ad copy, Claude's analysis, generated brief, date scraped. Tag it "To Review" so your team knows what needs attention each Monday.
After 8 weeks, you have a living database of competitor creative strategy, analyzed and briefed with zero manual work. You'll see which hooks they return to, which offers they're testing, and when they ramp spend. That's a competitive moat built on automation.
The hook is the ad. Everything else, the copy, the offer, the CTA, only matters if the hook holds attention. We never scale an ad without testing at least 3 hook types first. Here's the full framework.
They test the creative, different visuals, colours, spokespeople, but run them all with the same hook. You're not testing the ad. You're testing the thumbnail. Hook-first testing means: same visual, different opening 3 seconds. You'll find the winner 5x faster.
"Most people don't know this about [PRODUCT CATEGORY]..."
Works best for supplements, health tech, or anything that
counters a mainstream narrative.
"In 90 days, here's what happened to my [METRIC]..."
Works best for fitness, weight loss, any before/after product
category.
"47,000 people switched to [PRODUCT] last month. Here's why..."
Works best in crowded categories where trust is the primary
conversion blocker.
"If you're still doing [COMMON PRACTICE], stop."
Works
best for supplements with misconceptions, fitness programs,
diet-related products.
"Get [SPECIFIC RESULT] in [TIMEFRAME]. Here's how."
No
warmup. Straight to value. Works best for retargeting and high-intent
audiences.
Take your best-performing ad. Rewrite just the first 3 seconds in all 5 hook types. Test as 5 separate creatives with identical bodies and CTAs. Budget $50 per variation. After 3 days: kill the bottom 3, scale the top 2.
This is the most underutilized skill in paid media. Not media buying optimisation. Not creative strategy. Simply: looking at what's already working and understanding why. Here's the exact process. It takes 5 minutes once you've done it a few times.
Watch or read the first 3 seconds only. What specific emotion does this trigger? Curiosity, fear, desire, validation? Write it down in one word.
What happens immediately after the hook? Does it agitate the pain, introduce the product, show social proof, or build authority? This is what converts attention into interest.
What makes you believe the claim? A stat, before/after, testimonial, live demo, or authority figure? Note the specific type, this is what makes the claim land.
How does it close? Urgency, scarcity, soft ask, or straight offer? Is it friction-low ("learn more") or friction-high ("buy now")?
Give the structure a name so you can reference it later. "Problem/Agitate/CTA with social proof bridge." Now it's a reusable template.
Build 3 variations for your brand using the same skeleton. This isn't copying, it's understanding what persuasion structure the market is currently responding to and applying it to your offer. Do this with 10 ads in one sitting and you'll have a pattern library most agencies charge thousands to produce.
UGC fatigue is real. But it's not UGC that's failing, it's lazy UGC. Same talking head. Same "I tried this and it works" script. Same hook you've seen 500 times. These 3 formats are consistently outperforming everything else in our accounts right now.
Structure: Show a normal day, embed the product naturally. No hard sell. The product is incidental.
Hook: "Come with me on a typical Tuesday..." or "What a normal morning looks like for me now..."
Why it works: Pattern interrupts the "I'm about to watch an ad" filter. Feels like content. The product reveal lands harder because there's been no setup for it.
Script skeleton: Open on B-roll of daily routine → natural product use moment → brief explanation ("I started using [X] about 3 months ago, just because...") → close on the result. Length: 30–45 seconds.
Hook: "Everything you've been told about [CATEGORY] is wrong. Here's what actually works..."
Why it works: Positions your brand as the informed outsider. Triggers curiosity and slight discomfort in people who hold the myth, which is most of your audience.
Script skeleton: State the myth → explain why it's wrong or incomplete → introduce your product as the correction → proof point → CTA. Length: 45–60 seconds. Best for: supplements with novel mechanisms, fitness tools with counter-intuitive approaches.
Hook: "We asked people who'd been using [PRODUCT] for 90 days..."
Why it works: Three voices are more convincing than one. Social proof in numbers. The rapid editing maintains attention better than a single long testimonial.
Script skeleton: Interviewer question (off-camera) → 3 × 8-second responses from different people → product shot → CTA. Length: 25–35 seconds. Best for: any product with happy customers to film.
When Perfect Body came to us, they were running 4 ads, all variations of the same creative. Same hook, same spokesperson, same format. ROAS had plateaued at 2.1x for 11 weeks. The problem wasn't the media buying. It was creative monoculture. Here's the system we installed.
Most brands test 5 things at once and don't know what moved the needle. We test one variable per batch. Always. This is non-negotiable, without this discipline, you're just guessing.
Take the winning ad body (everything after the first 3 seconds). Write 5 different hooks. Run all 5 with identical bodies and CTAs. Budget: $30–50 per variation. After 3 days: kill everything with below-average CTR. Identify the top 2 hooks.
Take the top 2 hooks from Phase 1. Test each in 3 formats: talking head, B-roll with voiceover, text-on-screen. 6 total variations. Budget: $30–40 per variation. Identify which hook/format combination produces the best cost-per-click-to-landing-page.
Take the winning hook + format. Test 3 different ad bodies: Problem/Agitate/Solution, Transformation story, Straight demo. Budget: $40–60 per variation. Identify the top performer by CPL or ROAS depending on your funnel.
Take your winning hook + format + body. Test: urgency CTA, offer CTA, soft CTA. This is usually the smallest lever but often improves CVR by 8–15%.
By week 5 we had a clear winner: curiosity hook + B-roll format + transformation body + urgency CTA. ROAS on that specific ad: 6.1x. We scaled budget 3x. ROAS held at 6.8x for 6 weeks before the first sign of fatigue.
Creative fatigue is the silent killer of paid media performance. It doesn't announce itself. It just slowly erodes your numbers until you're asking your media buyer what went wrong. Nothing went wrong. The audience got bored.
Frequency. When your target audience has seen your ad 4–5 times, engagement drops, not because the ad is bad, but because the novelty is gone. The hook no longer surprises them. Thresholds vary: cold audiences fatigue slower (7–10 impressions), retargeting lists fatigue fastest (2–3 impressions).
These show up 7–10 days before ROAS drops. Watch for two or more moving in the wrong direction simultaneously:
Keep the winning ad body. Write 3 new hooks. This fixes 80% of fatigue situations. New hook = new pattern interrupt = ad feels new to a fatigued audience.
Same message, different format. If you've been running talking head, switch to B-roll + voiceover. Same script, completely different look and feel.
New concept, new angle, new everything. Reserve this for when the whole funnel is fatigued. This is where your angle bank becomes invaluable, you should always have 10+ fresh angles ready to deploy.
At $10k–50k/month: Layer 1 every 3–4 weeks, format change monthly, full resets every 8–12 weeks. Above $50k/month: Layer 1 every 2 weeks, format change monthly, full resets every 6–8 weeks.
Treating creative like a perishable asset, not a one-time investment, is what separates brands that maintain efficiency at scale from those that yo-yo.
Bad briefs are expensive. A creator goes off-brand, the footage is unusable, and you've wasted money and 10 days of turnaround. Good briefs eliminate this almost entirely. This template takes 30 minutes to fill out. Creators consistently deliver usable footage on the first attempt.
Three sentences: what the product is, who it's for, what makes it different. No fluff. Think: "This is a collagen supplement for women 30–50 who've tried everything and are tired of products that don't work. What makes it different is [X]. The brand voice is warm and direct."
One specific objective. Not "get sales." But: "Drive purchases from cold audience on Meta at a target CPA of $40."
Age range, gender if applicable, key psychographics (what they believe, fear, want), where they consume content. One specific person described in 4–5 sentences.
This is non-negotiable. The creator can improvise later in the video, but the first 3 seconds must match your tested hook exactly. Write it out verbatim.
3 bullet points only. These are the only 3 things the viewer needs to understand. The creator can use their own words, but these concepts must land.
The one piece of evidence that makes the claim believable. A stat, a personal result, a review, a certification. One only, more dilutes.
Legal restrictions, competitor comparisons, claims you can't substantiate, off-brand phrases. This section prevents costly reshoots.
"Go to [URL] and use code [CODE]." Nothing left to interpretation.
Number of raw cuts, aspect ratios required, max file size, delivery deadline, file naming convention.
We're at an inflection point. The tools available today can produce video ad creative at a fraction of the traditional cost and in a fraction of the time. This is our current AI video stack, and how the three pieces connect into one workflow.
Claude, Creative strategy and scripting. All scripts, hooks, and briefs run through Claude before anything goes near a camera.
MakeUGC, AI-generated UGC-style video. Upload a product image, select a creator avatar, provide a script, get back a talking-head style video with realistic lip sync.
Google VEO (via MakeUGC's Extend feature), Scene extension and B-roll generation. Takes short clips and extends them, or generates scene-based footage from text prompts.
Browse available AI creator avatars. Filter for demographic match to your target audience. Upload your product images and script. MakeUGC renders a talking-head style video with natural lip sync, realistic enough for testing.
Use MakeUGC's VEO Extend feature to add scene-based B-roll between talking head segments. Prompt example: "Person working out in a gym, morning light, natural movement." VEO generates 3–5 seconds of matching footage per prompt.
Import into CapCut or DaVinci Resolve. Add captions (essential, 85% of viewers watch without sound). Add product shots. Export at 9:16 for Reels/TikTok, 1:1 for Feed.
Use this stack to find your winning concept at low spend. Once validated, brief a real creator to produce a high-quality version for scaling. AI output consistently outperforms in testing but sometimes hits a ceiling above $10k/day.
Traditional UGC (1 creator, 1 concept): $300–800 + 2–3 weeks. AI stack (same concept): $0–40 + 2–3 hours. 10x faster. 90% cheaper. The compound effect on your testing velocity is enormous.