Monday morning, I almost opened a fake “client refund” video
Last Monday at 9 a.m., I was sorting through client messages in a Hangzhou café when a video under 20 seconds popped up. A familiar voice said, “Help me cover a refund first; I’ll pay you back later.” I stared at it for a few seconds, my finger hovering over “Transfer.” It wasn’t the person I knew—the voice and image had been cloned with AI.
Markus Cuban publicly questioned whether Facebook knew about AI deepfake scam ads while continuing to give them reach. Related reporting suggested the scams could involve up to $16 billion in fraudulent revenue. For anyone building a personal brand, freelancing, or running a small team, the frightening part isn’t expensive software. It’s customer trust being harvested.
This isn’t a chat-tool problem; it affects every step of the client pipeline
Today’s ad platforms serve video ads whose voices, lip movements, and on-screen people may be generated or replaced. The useful move is to verify the person in the video, the payment account, and the communication account separately—not just check whether the face looks familiar.
My friend Lin Yue, who provides cross-border design services in Guangzhou, once received a video inquiry from a “German client.” She contacted the person through the email listed on the client’s official company website and confirmed the company bank account used in the contract, avoiding a request for an upfront “materials fee.”
I’ve also gotten stuck here. At first, I confirmed the identity only by phone. When the person changed their profile picture, I realized the same voice might not belong to the same person. We don’t need to master every AI term first. A simple habit is enough: verify through a second channel before sending money.
Try it today: $0, about 30 minutes, almost no technical barrier
Cost: $0. Time: about 30 minutes. Technical barrier: no coding required; email and a browser are enough.
First step: Open Meta Ads Library, Meta’s public ad library. Search for your brand name, industry terms, and high-risk terms such as “refund,” “investment,” and “part-time job.” Open each result and check where its link actually leads.
I’d also give myself a 10-minute verification checklist: Can I confirm the person in the video through another channel? Does the receiving account match the legal entity on the contract? Is the person pressuring me to pay outside the platform?
I got this wrong too. I treated “looks like a real person” as “identity confirmed.” Now I insist on contacting people through a second channel. This method isn’t necessary for everyone, and it’s fine not to try it today, especially if you aren’t running ads yet.
Use the step that fits your stage, not one universal checklist
If you’re just starting out, I’d keep two official communication channels and put client agreements, quotes, and payment details in one table before buying detection tools.
If you have one or two steady clients, I’d require a second confirmation through a known number or email before every payment or account change, then save contracts and communication records.
If you’re scaling, I’d separate ad review, employee training, and payment approval: anyone can respond to clients, but only one designated person can change payment information. I’d also use the public ad library to check regularly for ads impersonating the brand.
This tool and method won’t guarantee that we recover losses. It’s more like a cheap reminder: trust can bring in orders, but verification protects our cash flow.