They want you to believe so badly in the AI-avatar slope…
it might be a slippery slope…
The better AI influencers get, the more people seem to crave human tiktok “creators”.
Why is that so:
Organic AI‑influencer playbooks hinge on mass volume and burner accounts.
That’s because they still mostly suck. No emotion, no acting chops, no hot takes, no realness.
So you spam volume to compensate. Spray-and-pray clipping style. Infinite shots at the algo.
Viewers come for the story:
The AI avatar in those POV Camera videos exists only to carry the story.
Hence the plain HeyGen avatars and not more fancy more evolved models.
(a 2d cartoon character with a chat bubble would hit just as hard.)
In fact, carousels were designed to convey stories like that.
Second, full‑screen AI vids give you maybe three formats, tops.
You can try a talking head on a green screen, but AI alone can’t carry your TikTok alone.
You can run a long‑hook video with an AI shocked face, but you still must write those hooks, since text‑AI writes are so mushy.
o3 could improve things, but you still need many examples and treat it as an “assistant”, it cannot take the driver’s seat.
If you post enough times, you can probably cover weak AI hooks, but it may cost more than partly writing them yourself.
Pay a person for one hour to film hundreds of 6‑7‑second shocked clips; AI is overkill here.
So, so,
The talking head looks robotic and stiff. It’s still weak unless you corner it in the background as a small part of your video.
Even then, only big volume can win over the algorithm.
Non‑HeyGen avatars often look too perfect; shiny skin, overly polished. Most create those made-for-ads “perfect” characters.
These clips last only a few seconds. You can stitch them or use the face‑cover trick to show an app but you still write hooks and record the demo product parts yourself.
Tech will fix some of this, and I believe in it. But two things won’t change:
Account community and comments can matter a lot for conversion, but it’s almost impossible for many AI accounts to manage them.
And the human slope which makes it feel real. Same old AI vs creativity debate.
When a format becomes doable through AI, everyone jumps in, it becomes the new average, and diminishing returns hit fast.
Social growth engineering is (consistently) riding the hidden new micro‑strategies.
You play both against the algorithm and against your content competitors.
As a standalone, they suck.
10% AI-avatar content can rock though
As a minority asset towards building up a story, they can be phenomenal.
POV Camera (and a bunch of e-com brands/shops) showed it in the wild:
LingoLooper backed it up:
Yet, still very visible to the trained eye. They leaned into it to juice engagement:
The comments tell you that this gimmick might age fast.
Another case is this study app:
They edited a human‑AI avatar in front of a classroom, then switched to the classic Studytok laptop view:
The “human” AI avatar is only a media asset piece that usually takes < 10% of the edit within a heavy (short) story telling driven format.
Get yourself some standout features
Differentiated, unique features are what you need to crack content‑market fit.
It sounds obvious, but every week we talk to people building utility apps who want to lean on TikTok marketing.
These apps usually live in crowded markets, where ASO and paid ads do the heavy lifting.
It’s hard to re‑invent the wheel when the product isn’t flashy. Here are a few patterns we’ve spotted that make those apps feel splashy and go more viral:
The BFF Lock Screen Widget
Tolan’s AI Avatar Character
Ditto Shareable List Social – straight from a viral format
Eyepic Zommed-in Beauty Enhancer (cc Color Analysis, Glam, …)
Liftoff Competitive Rank
Just a few of the many examples…
Sometimes ONE unique, or on‑trend, feature is all it takes to pull users into your app.
—
That’s a wrap.
See you on Monday for a recap of the most viral strategies we covered this week.
Best,
The Social Growth Engineers team.
Leave a Reply