Home / Blog / AI Workflows

Vibelets · AI How-To

How to remake a competitor's winning ad with Claude and AI tools, legally

A competitor ad that has run for months is a proven structure sitting in the open. Here is how to use Claude to tear it into its reusable parts, then rebuild those parts around your own product with AI image and video tools, so nothing of theirs ever ships.

By Siddhant, Product Marketing Manager at Vibelets

Built for people who run and create their own Meta ads

When a competitor's ad has been running for months, that is a signal. Meta keeps showing what keeps converting, so a long-running ad has already passed a test you have not paid for. The instinct is to screenshot it and make your version look the same. That is the move that gets you into trouble, and it skips the part that actually matters.

The value in that ad is not the footage or the logo. It is the structure underneath: how the hook opens, how the body carries the promise, how fast it moves, how the offer lands, and the shape of the call to action. Structure is not ownable. A specific script, a specific image, a brand name, and a logo are. So the job is to read the structure with Claude, keep only that, and rebuild every visible part around your own product with AI tools.

This is the exact path: where Claude reads and plans, where an image or video tool renders your own creative, and the one line you never cross. There is also a faster way to run the whole thing in one place, and the manual version below works today.

Key takeaways

A competitor ad that has run for months is a validated structure. That structure is what you study, and structure is not something anyone owns.
Claude is strong at teardown. Paste the ad and it names the hook type, the body beats, the pacing, the offer, and the CTA shape as a reusable skeleton.
The legal line is clean: keep the structure, and rebuild every asset. No copied footage, logo, script, or brand identity ever ships in your ad.
An AI image tool renders your statics and an AI video tool renders your motion, both from your own product shots, so the skeleton wears your brand.
Meta still wants broad targeting and enough budget to clear the learning phase. The teardown changes your creative, and the media math stays the same.

01 Why a long-running ad is worth studying

Ad platforms are ruthless about spend. If a competitor's ad has been live for weeks or months, Meta is choosing to keep serving it, which means it is earning its place against everything else that account is running. You are looking at a winner that someone else paid to validate.

What you want from it is the reason it works, and that reason is almost never the specific pixels. It is the underlying decision set: which pain the hook grabs in the first two seconds, how the middle keeps that promise, how quickly it cuts, where the offer appears, and how it asks for the click. Those are patterns, and patterns are fair to learn from the same way every copywriter studies great ads. The mistake is copying what your eyes see instead of what the ad is doing.

02 Use Claude to tear the ad into its parts

Watch the ad a few times, then write down what happens second by second, or transcribe the script and describe each scene. Paste that into Claude with a clear brief: "Break this ad into its reusable structure. Name the hook type, the body beats in order, the pacing, where the offer lands, and the shape of the call to action. Ignore the specific product and brand. I want the skeleton, not the copy."

Claude is good at this because a strong ad follows a logic you can name. It will tell you the hook is a problem-agitation open, that the body runs three quick proof beats, that the pace cuts every two seconds, that the offer arrives at the two-thirds mark, and that the CTA is a soft "see how it works" rather than a hard sell. That is the reusable asset. None of it belongs to anyone.

Hook

the open + lever

→

Body

beats + pacing

→

Offer + CTA

placement + shape

Claude turns a competitor ad into this skeleton. You keep the skeleton and rebuild every part.

Ask Claude to write the skeleton as a fill-in-the-blank template with your product left as a variable. "Hook: [pain] stated as a question. Beat 1: [proof point]. Beat 2: [transformation]. Offer at 0:12. CTA: soft invite." Now you have a proven shape waiting for your own words and your own visuals.

03 The one line you never cross

This is the part that keeps the whole method clean, so be strict about it. You are allowed to learn the structure. You are never allowed to ship anything that is theirs. Keep the skeleton and replace every visible and audible element with your own.

Keep (structure)

The hook type and emotional lever
The order and count of body beats
The pacing and cut rhythm
Where the offer lands in the timeline
The shape of the call to action

Rebuild (never copy)

Their footage, images, and B-roll
Their script, voiceover, and captions
Their logo, brand name, and colors
Their music and on-screen text
Their product, model, and set

A simple test: if a viewer could tell your ad apart from the original with the sound off, you have done it right. Your product is on screen, your words are in the captions, your brand look is throughout, and the only thing shared with the competitor is a sequence of decisions that no one can own. Have Claude review your rebuilt draft against the original and flag anything that reads as a copied phrase or a lifted visual idea.

04 Rebuild the structure around your product with Claude

Now fill the skeleton with your own material. Feed Claude your product page or a short description and ask it to write into the template: the hook rewritten around your customer's pain, the body beats rewritten around your proof, the offer in your terms, and the CTA in your voice. The shape stays, and every word is yours.

Push Claude to write several versions of each beat so you have range to test. Ask it to keep the same pacing discipline the original used, the same tight opener and quick cuts, because that rhythm is part of why the structure worked. Tell it to avoid hype and write the way a real person talks, so the ad sounds like your brand and reads clean.

Map your product onto each slot

Take the skeleton and, slot by slot, write in what your product does. The competitor's hook grabbed a specific pain, so name the equivalent pain your customer feels and open on that. The point is a one-to-one swap of meaning, with none of their wording carried over.

Gets you: a script that follows a proven shape in your own words

Have Claude pressure-test the draft

Paste your rebuilt script back next to the original and ask Claude to check two things: does the structure still hold, and is there any phrase, claim, or visual cue borrowed from the competitor. Rewrite whatever it flags. This is the review pass that keeps you on the safe side of the line.

Gets you: the structure intact and nothing of theirs left in it

05 Render your own statics and motion with AI tools

With the script done, produce the actual creative from your own assets. Take your product photos into an AI image tool for the statics and an AI video tool for the motion. Most of these read a reference image and a short prompt, so you feed them your product shot plus the scene description from your skeleton.

Two rules keep the output usable and clean. Match the first frame of the video or the top third of the static to the hook, so the promise and the visual agree the way the original's did. And build only from your own footage and product images, because the whole point of the method is that the pixels are yours.

Cut rhythm: match the pacing you tore down, open on the hook frame, keep every asset your own

Produce both a video and an image version of your strongest cuts, because the same structure can win in one format and stall in the other. Now you have a set of ads that follow a validated shape and are one hundred percent your brand.

The shortcut

From a reference ad to your brand in one step

Vibelets reads your store from a URL, then remixes a reference ad's structure into your own brand: your product, your look, your words, rendered as video and image ads and launched to Meta. You keep the proven shape and none of their assets, without juggling a stack of tools. It flags what to scale and pause with the signal behind each call, and you approve the moves.

Start with 50 free credits ↗ app.vibelets.ai

06 Launch on Meta and read the first signals

Launching is the same media math as any other campaign. Go broad on the audience and let the algorithm find the buyers, because in 2026 the creative carries most of the targeting signal. Run 3 to 4 of your rebuilt cuts per ad set, give each enough budget to clear roughly 50 conversions per ad set per week, and resist editing mid-learning, since a budget change over 20 percent quietly resets the phase.

Once it is live, export the early numbers and paste them into Claude. Ask it to read click-through and hook rate first, since those tell you whether your version of the structure earned attention the way the original did, then cost per result and conversion rate to confirm it. Claude surfaces which cut is carrying delivery. You keep the scale and pause calls, so a single good day never moves real budget on its own.

07 The takeaway

A competitor's long-running ad is a free lesson in what converts, and the lesson is the structure, never the assets. Use Claude to tear the ad into its hook, beats, pacing, offer, and CTA shape, rebuild that skeleton around your own product in your own words, and render the statics and motion from your own photos with AI image and video tools. Keep the shape, replace everything visible, and launch broad on Meta. The two decisions that need you, the product you build around the shape and the winners you scale, stay yours. The rest hands off.

Frequently asked

How do I recreate competitor Facebook ads with AI?

Tear the competitor ad into its structure with Claude, then rebuild that structure around your own product. Paste a scene-by-scene description into Claude and ask for the hook type, body beats, pacing, offer placement, and CTA shape as a reusable skeleton. Then write your own script into that skeleton and render the statics and motion from your own product photos with AI image and video tools.

Is it legal to copy a competitor's ad structure?

Learning the structure of an ad is fair, and it is what copywriters have always done. Copying the assets is where you cross the line. Keep the hook type, beat order, pacing, and CTA shape, and rebuild the footage, script, logo, brand name, and product entirely as your own, so nothing that belongs to the competitor ever ships in your ad.

Can Claude analyze a competitor's ad for me?

Yes. Paste a transcript or a second-by-second description of the ad into Claude and ask it to name the hook type, the body beats in order, the pacing, where the offer lands, and the CTA shape. Claude turns the ad into a labeled skeleton you can reuse, and it can also review your rebuilt draft for anything borrowed.

What AI tools do I need to remake an ad?

A language model like Claude to tear down the structure and write your new script, an AI image tool for statics, and an AI video tool for motion, all fed from your own product photos. Ads Manager is where you launch. That is three or four tools stitched together, which is what an all-in-one platform collapses into one step.

How is this different from just cloning an ad?

Cloning copies the visible assets, which is both risky and ineffective because it wears someone else's brand. Remaking keeps only the underlying structure and rebuilds every asset as your own, so you get the benefit of a proven shape while your product, words, and look are the only things a viewer sees.

How do I know which competitor ad is worth remaking?

Look for ads that have been running for a long stretch, since Meta keeps serving what keeps converting. A long run is the platform validating the structure with real spend, which makes that skeleton worth studying and rebuilding around your own product.

Can AI rebuild a reference ad into my own brand automatically?

Yes. Vibelets reads your store from a URL and remixes a reference ad's structure into your brand in one step, generating video and image ads with your product and look, then launches them to Meta. You keep the proven shape and none of the competitor's assets.

Sources

Meta learning phase and ~50 conversions per ad set: facebook.com/business/help
Broad targeting and creative-led delivery in 2026: practitioner benchmarks
Ad structure teardown and creative testing: 2026 industry reporting

Vibelets · vibelets.ai