- Test one offer in three formats, a short video, a service image and a before-and-after, and keep what brings inquiries.
- Label realistic AI video and images where a platform or law asks for it, and never present an AI person as a real customer.
- Good editing makes versions faster. Your offer, your real photos and your reply speed still decide the result.
A cleaning company has one advert. It has run for four months. Nobody knows if a different picture or a shorter video would do better, because making a new version means a new shoot.
It doesn’t have to be that way. With a simple editing routine, a small business can make three versions of one offer in a day, test them on a small budget and keep the winner.
What matters just as much is honesty. The rules on labels and testimonials are stricter than ever. This is how we run our ad video service.
What counts as a video ad for a local business?
Short video ads come in a few forms. For a local business, that usually means:
- your real clips, edited into a short vertical video
- a presenter explaining your offer, on camera or as a voice-over
- your real photos turned into a short animated video
- several edits of one video for different platforms
“UGC” means user-generated content. UGC-style ads look like a normal person filming on their phone. They often feel more natural than polished adverts.
What should a local business test?
Don’t test ten ideas at once. Test one offer, in three formats:
| Version | What it is | Why test it |
|---|---|---|
| A. Short video | A 15 to 30 second vertical clip that asks the question your customer is already thinking | Stops the scroll |
| B. Service image | One clear image that shows what is included, with one button | Fast to read |
| C. Before and after | The same room or result, before and after | Shows the change instantly |
Run all three with the same budget and the same audience for about a week. Then look at inquiries, not likes.
What must you label?
Platforms and laws are moving in one direction: people should know when realistic content is made by AI.
- Meta (Facebook and Instagram) can add an “AI info” label, and has extra rules for some ad categories.
- TikTok asks creators to label realistic AI-generated content.
- In the EU, the AI Act requires that realistic AI-generated or manipulated images and video are disclosed.
A simple habit covers most cases: if a viewer could mistake it for a real person or a real scene, label it as AI-generated.
The testimonial rule you must not break
This is the line that matters most.
An AI presenter can explain your service. An AI presenter must not pretend to be a customer.
“I used Bright Clean last week and my kitchen has never looked better”, said by a made-up person, is a fake testimonial. In the US, the FTC’s rule on fake reviews and testimonials covers this, including content made with AI. Other countries have similar consumer laws.
Real reviews from real customers are fine to show, with their permission.
Step by step: your first ad test
- Pick one offer with a clear price or a clear next step.
- Write down approved facts: what is included, the area you cover, how to book.
- Make three versions following the table above.
- Label them where the platform or the law asks.
- Run a small test with equal budgets for about a week.
- Reply fast. An ad that brings messages is wasted if replies take a day.
- Keep the winner, pause the rest, then test one new idea against the winner.
How it works: three ad versions from one offer
- When this happensYou approve a new offer, for example a spring boiler service.
- The system doesWe write three short scripts and captions from the offer and your approved facts.
- ThenWe edit a short vertical video, a service image and a before-and-after layout, each labeled where a platform asks for it.
- ThenYou approve the versions, then all three run on a small, equal budget.
- ThenAfter a week, the system reports inquiries per version and pauses the weakest.
- You onlyApprove every script and visual before it runs, and decide the budget.
- CapCut
- Canva
- Meta Ads Manager
- Google Sheets
Common mistakes
Fake customers. The fastest way to lose trust, and in some places a legal problem.
Testing too many things at once. You won’t know what made the difference.
Judging by likes. Count inquiries and bookings.
Slow replies to ad messages. Every message from an ad is a paid lead. See how fast you should reply to a new lead.
A simple next step
Choose the one service you most want more of this month. Write one sentence that explains the offer. That sentence is the start of all three versions.
One good video can also feed your social pages for a week, as one video, ten posts shows.
Want help planning your first test? Book a free video call.
Sources
- Federal Register: FTC rule on the use of consumer reviews and testimonials (2024)
- TikTok: new labels for disclosing AI-generated content
- EU AI Act, Article 50: transparency obligations
Written by MD Al Amin from our day-to-day client work. Research and drafting tools were used, and a person checked every fact before publishing.
Quick questions
Do I have to say an ad was made with AI?
It depends on the platform and the country. Meta and TikTok have labels for AI content, and some laws require disclosure for realistic AI people or scenes. When in doubt, label it.
Can an AI presenter say they used my service?
No. A made-up person presented as a real customer is a fake testimonial. In the US, the FTC treats fake reviews and testimonials, including AI-generated ones, as deceptive.
Do I need a professional shoot?
Usually not. Clear phone clips and real photos of your work are enough for most local ads. Good editing turns them into several versions quickly.