AI-generated UGC is 60-70% awesome, 30-40% meh

Can you use AI-generated UGC for all your advertising needs? Not quite yet, according to Poolday CEO Alexei Chemenda. But it’s close, and it will work well for short-form mobile ads.

“If you compare it to where it was two years ago, it’s absolutely insane,” Chemenda told me in a recent Growth Masterminds podcast. “If you compare it to where the world wants it to be, it’s probably 60% there, 70% there. So it’s a great starting point, and with the proper setup, it’s certainly delivering results.”

So how can you kill it with AI-generated UGC?

Click play and keep scrolling:

AI-generated UGC: what are we talking about here?

AI user generated content (UGC) is already massive and growing. You’ve certainly seen it in mobile ads, even if you didn’t completely notice it or weren’t entirely sure if you were seeing a human or an AI-generated avatar. It often includes one or more of:

• AI avatars, or virtual humans
• Synthetic voices, often in multiple languages
• Script generation, which can be of varying quality
• Scene compositing, which is AI placing avatars in scenes like homes, cars, offices, etc.

Marketers use AI-generated UGC because it’s quick and cheap: you can make hundreds or even thousands of creative variations quickly: think hours, not days or weeks. It can help marketers localize content with AI-driven translation, and it’s great for rapid creative iteration to find the best-performing creative.

Currently Poolday has seen success in verticals such as:

• Hypercasual games
• RPGs and Match-3 games
• Interactive story apps
• Social casino games
• E-commerce brands

The downside can be when you hit the 30-40% gap Chemenda mentioned: lower-quality platforms or even ambitious projects on top UGC generation tools risks introducing visual glitches. Translations might not be spot on, which could mean confusing, annoying, or even offending people in markets you’re trying to enter. And it’s not always clear whether AI-generated ads need to be disclosed, and if so, how.

But for small screens, much of that is less noticeable:

“ What’s important to remember is where that ad is gonna be displayed,” Chemenda says. “If it’s short form content displayed on mobile for users that scroll through, then maybe the details don’t matter as much.”

I’ve personally experienced that. One example: a little AI tool like a setting in my podcast production software that refocuses your eyes directly on the camera vantage point so it looks like you are looking directly at your audience, even if you were reading a script. Some of these things that work on a small screen in your hand don’t look amazing, necessarily, on a giant screen on the wall.

But on a small screen, no-one notices.

5 key factors that impact performance 

Performance advertisers understand: they’re not trying to make something that will hang in the Louvre. They want something that moves the needle in CTR and conversion rates.

Under that definition, it’s less about the artistic value or even how realistic the depiction of a human is than whether your creative generates ROAS.

For that, there are 5 key factors:

1. Hook (the first few seconds)
2. Actor
3. Editing style (jump cuts, continuous, filter usage)
4. Gameplay or product footage
5. Music

Interestingly, the call to action isn’t that critical of a factor, Chemenda says, because most people don’t watch video right to the end. (An interesting item to consider testing: do high-quality users/players who monetize well correlate with people who watch the video to the end?)

You can play with more variables, he adds, but these are the critical ones. Starting with these 5 gives you the greatest chance of focusing on what matters most and ultimately landing on a keeper: a great ad that massively outperforms the others.

(Hint: this is the kind of creative you need to find before you can scale.) 

Finding your unicorn creative with AI-generated UGC

You have to kiss a lot of frogs to find a princess. Or a prince. Or even better a real live genuine unicorn creative: a 10X-better creative that you must find BEFORE scaling your ad spend.

The math is no different when the creative just happens to be AI-generated UGC.

Kiss a lot of frogs.

Make a lot of creative.

3 or 4 is not enough. Many marketers fail with AI UGC, Chemenda says, because they only test a handful of creatives and expect instant success. Instead, you need at least 30 to 40 creatives just to get out of the learning phase.

The upshot: success = volume + iteration.

Just like any other form of creative, winning with AI-generated UGC isn’t about nailing it in one shot … it’s about constant production, rapid iteration, and patient learning. (The patience is the hard part for me too!) This lets you find the messaging, actors, and hooks actually work.

One bright ray in a never-ending ocean of non-royal frogs: one creative can produce multiple winners:

“We built a creative in August ‘22, so it’s been two years and a half now, and that creative and its direct variations, meaning when we swap out the hook or we swap out the AI actor, that creative still remains the top spender for us,” Chemenda, who in addition to being the CEO of Poolday is also the founder of a mobile app publishing studio, says. “And we spend about $400,000 a month.”

That’s … impressive.

And annoying. 

Marketers who find a unicorn creative that just doesn’t ever seem to stop performing generally get annoyed because they keep failing to beat it. (But yeah: you gotta just keep trying!)

Killing creative fatigue

The absolute beauty of AI-generated UGC shines through when you’re starting to experience creative fatigue.

The scenario is that you’ve found an ad that really, really works, and you’ve milked it for all it’s gonna give you … and creative fatigue just starts slowly eroding your results. In the data, you see a long slow fade in performance from that creative.

What to do?

You don’t necessarily have to reinvent the world here and toss that entire once-great-performing creative into the trash can. Instead, swap out the actor. Clearly, something about the FORMAT of the ad and its MESSAGE worked … maybe a different visual delivery will reinvigorate this ad and return it to the glory days of ridiculously high performance.

Chemenda’s seen it work. Give it a shot.

But hey: AI-generated UGC is not a magic wand

Umm … yeah.

I like AI too, but it won’t do my entire job. And the same is true for you. Just like any other form of creative AI-generated UGC is hard work.

AI or not, good UGC takes planning, iteration, and storytelling. A human face alone isn’t enough. A great actor alone isn’t enough. You can’t just slap an actor onto a generic ad and call it a day. Messaging still matters. Story still matters. You still need an emotional hook. You still need a compelling narrative.

So invest the time you need in each of the 5 key factors, the hook, the actor, the editing style, the gameplay or product footage, and the music.

But make sure the core idea, the core story, is awesome. Or iterate until it is.

More in the full podcast!

Watch, listen, tell AI to summarize it for you, but get all the goodies in the full podcast. Here’s what you can expect:

• 00:00 Introduction and Background
• 00:42 Meet Alexei Chemenda: CEO of Poolday
• 02:28 The Power of Curiosity in Marketing
• 03:00 Generative AI for UGC: An Overview
• 05:29 Testing and Iteration in AI UGC
• 07:23 Success Stories and Best Practices
• 11:46 Who Benefits Most from AI UGC?
• 13:54 Creating Effective AI UGC Ads
• 21:58 The Future of Personalized Advertising
• 24:32 Ethics and Tagging in AI UGC
• 27:14 Final Thoughts and Key Takeaways