The Problem With Using AI For UGC

Introduction

You've seen it all by now. 

AI avatars that look human.
Synthetic voices that sound almost real.
Platforms promising unlimited UGC at a fraction of the cost.

On the surface, the pitch seems compelling.

Why pay creators $300 per video when you can generate infinite content for pennies?
Why manage 50 relationships when software can do it in minutes?

I get the temptation. Content costs keep climbing, and dealing with creators can be genuinely exhausting.

But I need you to understand what you're actually trading away once you decide to pull the trigger and go all in on AI UGC.

Introduction

You've seen it all by now. 

AI avatars that look human.
Synthetic voices that sound almost real.
Platforms promising unlimited UGC at a fraction of the cost.

On the surface, the pitch seems compelling.

Why pay creators $300 per video when you can generate infinite content for pennies?
Why manage 50 relationships when software can do it in minutes?

I get the temptation. Content costs keep climbing, and dealing with creators can be genuinely exhausting.

But I need you to understand what you're actually trading away once you decide to pull the trigger and go all in on AI UGC.

Let's back up for a sec. Why does UGC convert better than polished brand content in the first place?

It's not because it’s mobile footage. It's not the casual framing. Those are surface features that are easy to replicate.

What's hard to replicate is the actual reason viewers trust UGC.

When a real customer says "this actually worked for me," there’s a quick conversation that goes on in the viewer’s mind.

Who is this person? Are they being paid to say this? Do they seem genuine? Would they use this if they weren't being filmed?

When the answers check out, trust transfers. The viewer then thinks: "If it worked for her, maybe it'll work for me."

That trust transfer is the entire mechanism that makes UGC so powerful.

The problem with AI generated content is that it can't do this because there is no real person and there is no real experience.

You're manufacturing social proof that is 100% based on a lie. And although it’s easy and cheap to create, you have to admit to yourself that it’s downright unethical.

In 1970, roboticist Masahiro Mori discovered something strange. 

As robots became more human-like, people's comfort increased, but only to a point. When the robot looked almost human but not quite, something flipped. 

People felt revulsion. A creeping sense that something was deeply wrong.

Mori called this the "uncanny valley."

Everyone’s brain is running a threat detection software they can't turn off.

Humans spent entire civilizations reading faces for survival. Is this person friendly or hostile? Are they lying? Can I trust them?

We developed extraordinarily sensitive pattern recognition. We pick up on blink timing, micro movements around the eyes, the synchronization between lips and speech. 

Most of this processing happens below conscious awareness. You don't think "that person's smile engaged at the wrong velocity." You just feel something is off.

AI faces fail on the micro level.

Current AI has solved the macro problems. Skin looks like skin, proportions are correct. But the micro expressions are wrong. The timing is off.

A real smile involves dozens of muscles firing in coordination developed over a lifetime. AI generates an approximation based on training data. Good enough to pass a glance. Not good enough to pass sustained attention.

When viewers watch a 30 second testimonial, their unconscious processing runs the whole time, flagging inconsistencies they can't articulate. They don't trust the speaker. They don't know why. They scroll.

This phenomenon might be permanent.

There's a hopeful narrative that the uncanny valley is temporary, that AI faces will eventually become indistinguishable from real ones.

Maybe. But there's a competing theory: the uncanny valley isn't a bug in human perception. It's a feature, specifically evolved to detect imposters, corpses, and disease. If true, the valley won't disappear with better rendering. Our brains might be designed to find near-human -but-not-quite extremely disturbing, regardless of how good the technology gets.

Every viral tweet exposing AI content will train viewers to look harder. Every article about synthetic influencers will raise suspicion. And as AI content becomes more common, so will cultural awareness of AI content.

Your analytics will show poor conversions, and you’ll have no idea why. (At least, now you do)

You're a DTC brand doing $10M a year. You've spent four years building trust with real reviews from real customers. You’ve built a proper community.

Then someone screenshots one of your AI generated testimonials. Posts it with: "lol this brand is using fake people to sell you stuff."

The post goes viral. Quote tweets pile up. People start scrutinizing all your content. Other AI pieces get surfaced. The narrative becomes: "this brand has been lying to you."

Suddenly you'll no longer be dealing with an ad conversion problem. You'll be dealing with a full-blown PR crisis.

As if that’s not enough, there are at least three other main problems you have to deal with when you decide to use AI for UGC

This basically means your AI ads aren't proprietary assets. Competitors can legally copy them. You have no legal recourse. You'll be spending money to create content you won't own.

Running undisclosed AI testimonials will put your ad accounts at risk.

This is currently the system our clients at Refunnel are using to run UGC campaigns at scale:

Right now, customers are posting about you. Unboxings, reviews, before and afters, recommendations in comment threads. Most vanishes because you're not watching.

The tracking setup:

You need coverage across Instagram, TikTok, YouTube, and Twitter/X. Tool options:

For brands under $3M, a trained VA doing manual sweeps twice daily also works. Budget $500 to $800/month for this if you prefer going that route.

The rights workflow:

When you find content, send this:

Hey [name], saw your post about [product]. Loved it, especially [specific detail]. Would you be open to us sharing this? We'd credit you and send [product/gift card] as thanks.

Response rates: 60 to 70% say yes to reposts. For ad usage rights, offer $50 to $150 per piece.

Benchmark: At 1,000 orders/month, aim for 20 to 40 pieces monthly (2 to 4% of customers posting something usable).

Pull affiliate data from the last 90 days. Sort by conversion rate, not clicks.

High conversion affiliates are creating content that persuades. Your top 10% by conversion rate are your content candidates.

Reach out directly:

Hey [name], your conversion rate is in our top 10%. Whatever you're doing is working. I'd love to talk about licensing some of your content for our channels. Would you have 15 minutes?

What to offer: $100 to $300 per piece for ad usage rights, or $500 to $1,500 monthly retainer for 4 to 8 pieces. Build content agreements into affiliate relationships from the start.

Stop spreading thin across hundreds of transactional creator relationships. Find 20 people who genuinely love your product. Build real partnerships.

How to identify them:

The relationship structure:

The math: 20 creators × $1,000/month = $20,000. At 4 pieces each, that's 80 pieces for $20k, or $250 per piece. Same range as one off UGC, but quality is dramatically higher because these people actually use and understand your product.

More importantly, these relationships compound. By month six, they're producing content random creators couldn't match at any price.

Some products generate content naturally. Others don't. The difference is usually intentional design.

Order your own product. Ask at each stage: where would I reach for my phone?

Elements that trigger shares:

AI tools are getting cheaper every month. Within a year, every brand will have access to unlimited synthetic content. Feeds will flood with AI faces saying AI things about products they've never touched.

When everything looks synthetic, real content will stand out.

Your competitors using AI content will start accumulating a trust debt they'll eventually have to pay.

Meanwhile, you could be building real relationships and real systems that compound over time.

The window is open. Build now.

We built Refunnel to solve the capture and deployment problem. With our solution, you’ll be able to monitor, track, and manage user-generated content and simplify securing permissions for content usage.

Click on the link below and I'll show you exactly how it works.