🧠 What I Learned from Testing AI-Generated Video Content So Far
This week I ran an experiment: I started posting almost 100% AI-generated videos to brand-new accounts. Each video takes about five minutes to make using Sora 2 and ChatGPT for captions.
The results: Pretty average — around an 8% like-to-view ratio, which is standard for marketing-style content, but the difference is time. At five minutes per video, even ā€œaverageā€ engagement becomes really good ROI.
Here’s what I’ve noticed so far šŸ‘‡
  • Relatable content still wins. Videos about the headaches of running a business do better than ones about success. People connect with what they’ve lived through — not what they envy.
  • Your audience mirrors what they see. When I use a 30-something female character with a modern office vibe, the audience skews toward people who look like her, dress like her, or strive to be her. It’s a reminder that visuals are a form of targeting — people are drawn to reflections of themselves. Representation shapes your audience. Be intentional about who appears on screen.
  • Music choice and vibe matter a lot. The more upbeat the song and faster the pacing, the higher the engagement. The tone and rhythm of a video often matter more than its message to keep your audience engaged. People feel your content before they think about it.
  • CTA Matters for conversion. Even the best visuals and storytelling fall flat if your call to action isn’t clear. Viewers shouldn’t have to guess what to do next, you have to tell them. ā€œLearn how belowā€, ā€œVisit >website< to start,ā€ or ā€œMessage for more informationā€ — it all works when it’s direct.
Key Takeaway:
AI hasn’t changed what works marketing, it just saves us time and allows for more testing. There's no filming and almost no editing, but the core principles haven’t moved an inch:
  • Relatability
  • Representation
  • Energy
  • Clarity
That’s what connects with people — not the tool you used to make it.
So if you’ve been holding off on creating because you think it has to be perfect, it doesn’t.
Start experimenting, gather data, and adjust as you go. Some eyes on your product are better than none, and by putting yourself out there, you learn what works and what doesn't.
I’ll be posting a full case study soon with all the videos, prompts, and metrics once the experiment wraps up.
Have you tried Sora 2 yet? What have you noticed between your best and worst-performing videos? Drop your insights below. I’d love to compare notes :)
2
1 comment
Katerina DiFatta
4
🧠 What I Learned from Testing AI-Generated Video Content So Far
powered by
Lucidium Executive AI Strategy
skool.com/unlock-the-machine-mind-8635
Learn to use AI strategically like an executive—build clarity, systems, and confidence in every decision.
Build your own community
Bring people together around your passion and get paid.
Powered by