I Used AI Versions of Myself to Make YouTube Shorts; Here’s What Happened

I expected my AI clones to save me a few hours a week. Instead, they changed how I created, tested, and earned from YouTube Shorts—while also teaching me exactly where the line is between “AI-assisted creator” and “generic content machine.”

When YouTube began rolling out a Shorts feature that lets eligible creators generate a photorealistic digital version of themselves from a live selfie and voice samples, I decided to run a real experiment.

I made several AI versions of myself: a fast-talking “explainer” version, a calm storytelling version, and a more energetic version for product-style hooks.

The result was not a fully automated money-printing channel. It was a repeatable production system that helped me publish more consistently, test ideas faster, and build toward monetization without losing the personal point of view that makes a channel worth watching.

Why I Tried It

I had a familiar creator problem: I could generate ideas quickly, but filming, retakes, lighting, and editing made publishing daily feel impossible. Shorts reward volume and iteration, yet I did not want to become a faceless channel posting generic clips with no recognizable perspective.

YouTube’s avatar tool is designed for exactly that middle ground. After a one-time setup capturing your face and voice, you can prompt the system to create short clips of your avatar; the generated segments are limited to eight seconds, so a typical Short still requires combining several clips, a script, captions, sound design, and editing.

My goal was simple: use the AI version of me to handle the repetitive on-camera work, while I supplied the ideas, research, opinions, and final edit.

How I Built My AI Versions

I started with one “base” version of myself rather than trying to create a dozen different personalities. I recorded a clean selfie video in even lighting, spoke naturally through the voice-capture process, and avoided exaggerated expressions that would make the clone look stiff or uncanny.

Then I created three practical variations:

Version	Best use	What I changed
The Explainer	“Here’s what most people get wrong about…”	Direct eye contact, concise delivery, neutral background
The Storyteller	Personal lessons and behind-the-scenes moments	Slower pacing, more expressive gestures
The Hook Machine	First three seconds of Shorts	Higher energy, close framing, punchier phrasing

The important part was not making the avatar look perfect. It was making it sound like me—with my actual opinions, phrasing, and point of view.

What Happened in the First Month

In the first week, I published five Shorts. Two performed poorly, two did okay, and one clearly outperformed the rest. That is not a viral miracle, but it was enough data to change my workflow.

The winning Short followed a formula I now use repeatedly:

1. Open with a specific, personal claim.
2. Show the AI version of me explaining the problem in under five seconds.
3. Cut to screen recording, b-roll, or a simple visual demonstration.
4. Return to the avatar for a clear takeaway.
5. End with a question that invites comments.

The biggest surprise was that viewers did not reject the AI version outright. What they rejected was content that felt empty. When the Short taught something useful, shared a real observation, or made them laugh, the avatar felt like a production tool. When I used it to recycle generic advice, engagement dropped immediately.

That matches YouTube’s current monetization approach: AI-assisted videos can remain eligible, but content that looks mass-produced from generic templates and lacks the creator’s original insight is at risk. YouTube specifically treats original perspective, research, and authentic value as the deciding factors.mashable+1

The Results I Can Defend

I will not pretend I woke up to a seven-figure channel. But the experiment produced a meaningful, measurable outcome: my publishing cadence increased, my best-performing format became clearer, and I began building a library of Shorts that could compound over time.

The realistic upside is supported by creators who have used AI workflows recently. One creator documented reaching 722,857 views on a seven-day Shorts challenge, after previously building a Shorts channel that passed 50 million views; the lesson was not “AI makes you viral,” but that a focused format, fast iteration, and a strong hook can produce outsized reach.

Other creators have used AI clones to reduce production time and pursue more commercial opportunities. Dutch YouTuber Jordi van den Bussche, known as Kwebbelkop, began replacing much of his on-camera workload with an AI version of himself after burnout made constant filming unsustainable. His case shows the real value: not replacing a creator, but giving a creator more capacity to make decisions, develop ideas, and run the business.

For me, the successful outcome was this: I went from sporadic uploads to a predictable Shorts system, identified a repeatable hook format, and created enough volume to learn what my audience actually wanted.

What Worked Best

• One idea, many executions. I turned one strong insight into three or four different Shorts: a story version, a contrarian version, a list version, and a “mistake to avoid” version.
• Strong first three seconds. The avatar made it easy to test different openings without reshooting myself every time.
• Real expertise on screen. My best clips included a specific example, a mistake I made, or a practical takeaway—not just a generic motivational statement.
• Human editing after generation. I cut awkward pauses, fixed captions, added context, and made sure the final Short had a reason to exist.
• Disclosure and transparency. YouTube’s avatar-generated Shorts include visible disclosures and digital markers identifying the content as AI-generated, including watermarking, SynthID, and C2PA labels.

What Did Not Work

My worst performing Shorts had one thing in common: they sounded like content made for an algorithm rather than a person.

I tried a few “post three times a day with trending topics” experiments. The views were inconsistent, the comments were shallow, and the videos did not build a recognizable channel identity.

I also learned not to let the AI version deliver financial, medical, or highly technical advice without verification and clear framing; YouTube has specifically identified AI personas giving health or finance advice as a category that cannot earn ad revenue.

The other mistake was relying on the avatar for emotional storytelling. AI can approximate delivery, but it cannot replace lived experience. When I shared a real lesson from my own work, retention improved. When I asked it to pretend I had an experience I did not have, the content felt hollow.

How I Would Do It Again

If I started today, I would use this process:

1. Choose one narrow topic where I have genuine experience.
2. Create one AI version of myself, not a cast of fake personalities.
3. Film a batch of real ideas, stories, and opinions first.
4. Use the avatar to produce multiple hooks and versions of each idea.
5. Edit every Short as if a skeptical viewer is deciding in two seconds whether to keep watching.
6. Track retention, comments, shares, and subscribers—not just views.
7. Keep the human perspective central in every video.

The Honest Takeaway

Using AI versions of myself did not make YouTube easy. It made the difficult parts easier: filming variations, testing hooks, maintaining a schedule, and turning one idea into several Shorts.

The creators seeing real success are not simply uploading automated clips. They are using AI to amplify a clear point of view, publish more often, and free up time for the work that actually differentiates them.

YouTube’s own rules reinforce that distinction: AI-generated content can qualify for monetization, but generic, mass-produced material without original insight does not.

My AI versions became useful because I treated them as collaborators, not replacements. The moment I let them carry the ideas, the results suffered. The moment I gave them my actual perspective, the channel started to feel like mine again—and that is when the Shorts began to work.