How to Write Facebook Ad Scripts with AI That Actually Convert
Table of Contents
In the past six months, we have written over a thousand Facebook ad scripts using AI. They've generated millions in revenue for e-commerce brands.
Here's what so many people get wrong. The difference between ads that flop and ads that print money isn't the AI tool you're using. It's the system you feed into it.
I know you're already using AI to write your scripts. Everyone is. The problem is most people are just typing stuff like "Write me a Facebook ad for my product" and hoping for the best.
That's why your ads sound generic. That's why they don't convert.
Here's the five-step system we use to turn AI from a mediocre copywriter into a conversion machine.
The Two Components of Every High-Converting Ad
Before we get into the system, you need to understand this framework.
Every high-converting Facebook ad has two components: angle and format.
The angle is your core marketing message. It's the main perspective you use to present your product. A protein bar could be "the perfect post-workout fuel" or "a healthy afternoon snack for busy parents." Same product. Completely different angles.
The format is how you execute that angle. A founder-style video. An us-versus-them comparison. A static ad with bullet points. These are formats.
Most people feed AI a product description and ask it to be creative. But AI doesn't know your customer's pain points. It doesn't know which angle will resonate. It doesn't know what format performs best in your niche.
That's what this system fixes.
Step 1: Find Ideas That Actually Have a Chance
You don't need to reinvent the wheel. The best ideas already exist. You just need to know where to look.
TikTok feed. Scroll through TikTok and look for content getting traction in your industry. Also look at adjacent industries just outside your niche. Open a Google sheet and save videos with the most views and engagement.
Ad Spy tools. Search for your favorite brands and competitors, then filter by shares. This shows you ads that have gotten the most traction. Make sure you're not looking at ads that are too old. Stick to ads that performed well in the past 6 to 12 months.
Meta Ads Library. Be cautious here. The fact that an ad has been running for a long time doesn't mean it's a winner. Some brands deliberately leave older ads active that aren't getting any spend. They do this to confuse people who copy their work. Use your own judgment and look for inspiration only.
Foreplay or Atria. These are like a second layer to Meta Ads Library. You can save ads, easily get transcripts, and browse their historical database.
Your existing winners. If you have a video that's already converting, don't just leave it sitting there. Use it for inspiration. What made it work? Can you apply that to a new angle?
The key here is pattern recognition. You're exploring what angles resonate with your audience and adapting them to your brand.
Step 2: Do Your Research (Don't Skip This)
Even professional creative strategists don't spend enough time on research. Jumping straight into writing is more fun. But without good research, you're wasting the creator's time, the editor's time, and ad spend.
Here's what you should know about your product before writing a single word:
- Features: What does it do?
- Benefits: What does the customer get?
- Pain points: What problem does it solve?
- Objections: Why might someone not buy?
- Failed solutions: What have they tried before that didn't work?
- Trigger events: What moment makes them realize they need this?
- Driving emotions: What emotions motivate the purchase?
How to gather this information:
Start with your product page. Copy it, paste it into Claude, and ask it to extract all the elements above. If your product page doesn't have enough information, move to the next step.
Mine Amazon reviews. Find a similar product on Amazon. Use Apify to scrape the reviews, then run them through a customer review mining prompt. This gives you real customer language. Gold for ad copy. Segment positive and negative reviews separately because both are equally valuable.
Read Reddit threads. Search your product category. Look for threads where people ask for solutions. This is where you find trigger events and failed solutions.
Read your Meta ad comments. Look for frequently asked questions and objections that keep coming up. If people are saying "too expensive," you need to address that in your script.
Set up post-purchase surveys. I use Triple Whale for this. Questions I like: "Which of the following led to your purchase today?" "Are you buying this as a gift?" "What made you almost not buy today?" "What is the main reason you bought?"
Put all of this into a marketing research document. This becomes your home base. Every script you write pulls from this.
Step 3: Write Your Scripts (Here's My Controversial Take)
Write some scripts without AI first.
After doing research, your brain is full of ideas. You can often jump into writing right away. Write until you've put all your ideas on paper. Only then start using AI tools.
Why? Because you can use those fresh ideas that are uniquely yours to kickstart the process. Your uniqueness is your competitive edge.
When you do start writing with AI, use Claude. I've tried ChatGPT, Gemini, Grok, others. Nothing sounds more natural than Claude so far.
Method 1: The Copycat Method
This is a simple two-step prompt.
Step one: Input your inspiration script and write "Analyze the tone, style, and flow of this script."
Step two: Follow up with this prompt: "I need a new script for a new creator. The script I gave you was a winning script. I want you to take inspiration from it, but I don't want it to be too similar. People have already seen the previous script. This would be for a brand called [X]. The product is [X]. I've added the research document for info. Try to incorporate the reviews and the research sheet." Then paste your research.
You can also run it without the research document first to utilize the base knowledge of the model.
Method 2: The Context Document Method
Create a Claude project where you teach it how to make a certain style of ad.
For example, make a guide around how to write a problem-solution Facebook ad script. Upload your brand research and ask it to follow the guide to write new ads.
To create these context documents, you can take transcripts from your favorite YouTube videos and ask an LLM to turn the video into an actionable guide. But honestly, if you can use your own unique knowledge, that's where your advantage comes from.
The key is specificity. Give the LLM enough of the right knowledge without overwhelming it with pages of fluff. Treat it like an assistant who needs clear direction.
Step 4: Revise Until It Sounds Human
Your first draft is never your final draft. Ever.
Read it out loud. Does it sound natural? If you stumble over a sentence, rewrite it. If it sounds like a robot wrote it, rewrite it.
Cut it down. AI tends to write too long. Your job is to trim the fat. Every sentence should serve a purpose. If it doesn't move the viewer closer to the sale, delete it.
Simplify the reading level. Paste your script into the Hemingway Editor. Aim for fifth grade reading level. Short sentences. Simple words. Clear ideas.
Use revision prompts. Some of my favorites:
- "Write it from a third person perspective"
- "Make it shorter"
- "Give me five hooks that utilize open loop psychology"
Revision is where good scripts become great scripts.
Step 5: Bring Your Script to Life
You have a few options here.
Shoot it yourself. If you're the founder or comfortable on camera, this is the most authentic option. People trust familiar faces.
Use AI video tools like Arcads. These generate UGC-style videos from your script. It's fast, it's cheap, and it works for testing. When editing, use a lot of real-world B-roll to cover up the AI creator. Think 80% B-roll, 20% creator face.
Hire a UGC creator or actor. You can use platforms, but our best creators have come from running our own ads with specific criteria. We set our own questions, filter for quality, and build a custom roster. We run recruitment ads on Meta, X, LinkedIn.
Here's something to keep in mind: if you want to maintain your competitive advantage, don't hire creators from the same places as everyone else. When a creator doesn't charge much, lots of companies use them. The faces become saturated. People recognize them from other ads. That makes it harder to convert.
Find your own unique sources.
Editing matters. Keep it tight. Add captions. Use pattern interrupts to hold attention.
The Recap
Writing a high-converting Facebook ad script comes down to five steps:
- Ideas: Find what's already working and adapt it
- Research: Know your customer better than they know themselves
- Script writing: Use structure and AI to speed up the process
- Revising: Cut, simplify, and make it sound human
- Content: Bring it to life with video
Follow this system and you'll never struggle with ad scripts again.
If you're spending $30K or more on ads every month and want help implementing this system, get in touch with us!