Case Study: How a Worker in Their 50s Produces and Monetizes AI YouTube Videos as a Side Hustle

AI can be used to quickly create scripts, narration, images, and rough edits. However, “producing a video in 2 hours” and “earning 5 million won per month” are separate issues. The former concerns production efficiency, while the latter is a business outcome determined by views, advertising rates, viewer countries, video length, costs, taxes, and other factors.

This article reconstructs the production process using interview material from an IT employee in his 50s as a case study. The earnings figures are self-reported by the interviewee and are neither externally verified averages nor reproducible income guarantees.

Case Overview

The interviewee said that while considering his income and lifestyle after retirement, he came across a free YouTube-related course and started a side business. He initially considered economics and science as potential topics, then focused on the long-form science channel that achieved monetization results first.

The material states that video uploads began in April, but the year is not specified. Therefore, the results should not be assumed to belong to a particular year.

How to Interpret the “5 Million Won per Month” Figure

Revenue, Payouts, and Net Profit Are Different

Estimated revenue displayed in YouTube Studio may not be the same as the final payout or net business profit. To compare actual performance, the following items must be separated.

• Estimated advertising revenue: The amount provisionally displayed in YouTube Analytics
• Finalized earnings: The amount confirmed as payable after adjustments are applied
• Platform payout: The amount actually deposited after withholding or payment adjustments
• Net business profit: The payout minus AI subscription fees, voice and video generation API costs, music and source material costs, equipment expenses, and taxes

Based on the material alone, it is impossible to determine whether the “4.5–5 million won per month after taxes” mentioned in the interview refers to the amount after platform withholding or net profit after domestic comprehensive income tax. The statements about cumulative earnings of 17 million won and approximately $3,000 each month are also difficult to reconcile precisely without view counts and exchange-rate data for each period.

Data Needed to Verify Performance

At a minimum, the following metrics are needed to evaluate an earnings case.

1. The exact period during which the earnings were generated
2. Monthly views and valid playbacks
3. RPM and playback-based CPM
4. Viewers’ primary countries and languages
5. The share of revenue by source, including long-form advertising revenue, YouTube Premium, and affiliate revenue
6. Estimated revenue before cancellations and adjustments, and the actual payout
7. Costs for AI tools, APIs, music, and outsourced work
8. A distinction between pre-tax and after-tax profit

A channel experiencing a rapid short-term surge in views may record high peak daily earnings, but there is no guarantee that the same level will continue afterward.

Reconstructing the AI Video Production Process

The process described by the interviewee is closer to a semi-automated workflow that connects multiple generation tools to automate rough production while leaving final decisions to a person, rather than an approach in which AI operates the entire channel without human involvement.

Step 1: Finding Potential Topics

The interviewee explained that he subscribes to multiple science channels and looks for videos whose recent views are increasing rapidly relative to their subscriber counts. This can be used as a signal of viewers’ current interests, but it does not mean that the structure and expression of the original video may be copied.

Safer research methods include the following.

• Find questions or phenomena that have recently gained attention.
• Review papers, public agency materials, and multiple explanatory sources rather than relying on a single video.
• Design a core question, an original explanatory sequence, and a visual approach that differ from the original.
• Confirm that the title and thumbnail match the actual content of the video.
• Record the source and verification date for each referenced claim in the working document.

A “video with relatively few subscribers but many views” is only a clue for discovering topic demand. Imitating the title, script progression, and thumbnail composition of a particular video may create copyright and reused-content issues.

Step 2: Compiling Source Materials and Writing the Script

The interviewee said that he provides a reference video URL to Claude and requests a script approximately 10 minutes long to create the basic structure. However, it should not be assumed that AI can always accurately read the full video, subtitles, or latest information from a URL. The scope of processing may vary depending on access permissions, subtitle availability, and service functionality.

A more reliable method is to organize and provide verified materials first.

• The core question to explain and the target audience
• Facts confirmed through authoritative sources
• Speculation and unverified figures that must not be used
• The video’s original perspective or analogy
• The desired length, writing style, and scene structure
• Instructions that each fact must be rechecked

Once the AI draft is complete, a person must recheck proper nouns, dates, figures, causal relationships, and the level of scientific consensus. Any quotations or research findings not found in the sources must be removed.

Step 3: Generating Voice and Visual Materials

In this case, voice, image, and video generation services were connected to an integrated production solution. The process involved creating narration with a voice API such as ElevenLabs and generating photorealistic images and short intro videos.

The following risks must be reviewed at this stage.

• Whether a real person’s voice or face has been imitated without permission
• Whether generated images distort scientific facts
• Whether actual observational images could be confused with conceptual illustrations
• Whether the rights to logos, characters, photos, or video sources are being infringed
• Whether realistic-looking synthetic scenes require disclosure
• Whether the conditions for commercial use of music and sound effects are met

In science videos, generated images may serve as explanatory visualizations. If they are presented as though they were actual microscope images, astronomical observation data, or historical records, it is safer to clearly explain within the video that they are synthetic images.

Step 4: Reviewing the Rough Edit in CapCut

Once the integrated tool creates a project file with the script, narration, images, and video placed on the timeline, subtitles, background music, scene lengths, and other elements are adjusted in CapCut. Even after automatic placement is complete, a person must review the following from beginning to end.

• Whether the narration and visuals explain the same facts
• Whether numbers, units, and proper nouns in subtitles are accurate
• Whether scene transitions are excessively repetitive
• Whether voice pronunciation and sentence pacing sound natural
• Whether the background music conflicts with the narration
• Whether materials requiring attribution are presented in a misleading way
• Whether each video offers substantive commentary and new value

Step 5: Final Human Approval Before Uploading

The interviewee said that even when an automatic upload feature is available, he reviews the video himself before publishing it. This is an important control procedure for reducing errors, rights violations, and incorrect publication settings.

Before uploading, confirm the following.

1. Whether the title and thumbnail exaggerate or misrepresent the video’s content
2. Whether the necessary source and rights information has been included in the description
3. Whether the made-for-kids setting has been configured correctly
4. Whether realistic altered or synthetic content must be disclosed
5. Whether there are any issues with copyright check results or music licenses
6. Whether someone is responsible for reviewing comments and viewer reports after publication

Conditions That Make 1–2 Hour Production Possible

Producing one video in 1–2 hours is not a standard time that every creator can achieve immediately. It is more appropriate to view this as a compressed editing time made possible when the following prerequisites are in place.

• A reusable project template is available.
• Voice, image, and editing APIs have already been integrated.
• The channel’s writing style and visual style have been established.
• The creator understands the tools’ error patterns and how to correct them.
• Time for topic research and collecting reference materials is handled separately.
• Complex animation or original filming is not required.

Accurate science content may require more time for research and verification than for production itself. Omitting fact-checking to save time can undermine trust and monetization potential in the long run.

Points to Consider Under YouTube Monetization Policies

The Nature of the Result Matters More Than AI Use Itself

YouTube does not exclude every video from monetization solely because it was created with AI. However, content that uses templates to mass-produce nearly identical videos or lacks substantive differences between videos may be considered repetitive or mass-produced content.

Reviewers may examine not only individual videos but also the channel’s main themes, most-viewed videos, newest videos, share of watch time, and metadata such as titles, thumbnails, and descriptions. Therefore, simply adding a logo or changing the subtitle font does not guarantee originality.

Reused Content and Copyright Are Separate Issues

Even if no copyright claim is filed or permission has been obtained when using edited footage from another creator, the video may still fail to meet YouTube’s reused-content criteria. Conversely, claiming to have added original commentary does not automatically grant the right to use the material.

A video containing reused materials should be able to clearly answer the following questions.

• What new explanation or analysis does it provide compared with the original?
• Is it evident that the creator personally planned and narrated the video?
• Do viewers have an independent reason to watch this video instead of the original?
• Are the necessary rights secured for the video, images, and music used?

Disclosure of Synthetic Content

If realistic-looking synthetic content alters the words or actions of a real person, modifies a real place or event, or depicts a realistic scene that never occurred, it may need to be disclosed as altered or synthetic content during the upload process. Simple script assistance, subtitle generation, and clearly unrealistic visual effects may not be treated at the same level, so the official criteria and specific context must be reviewed.

Making a disclosure does not automatically restrict monetization. However, repeatedly failing to make required disclosures may cause YouTube to apply labels or impose penalties.

Reviewing the Claim That “Improving Quality Prevents Channel Deletion”

The interview material includes advice suggesting that adding personal opinions and a logo to improve quality nearly eliminates the risk of penalties. This should not be accepted as a guaranteed rule.

A channel’s status is determined by considering the following factors together.

• Whether the Community Guidelines have been violated
• Whether copyright infringement has occurred and warnings have accumulated
• Whether spam, deception, or manipulation has occurred
• Whether the content is repetitive or reused under monetization policies
• Whether synthetic-content disclosure requirements have been followed
• The overall context of the videos, titles, thumbnails, and descriptions

Superficial changes such as adding a logo, background music, or one or two sentences of personal opinion are not sufficient. The key factors are original research, independent structure, meaningful commentary, accurate information, and human editorial judgment.

Sustainable Operating Metrics

Looking only at upload volume and revenue may obscure a channel’s vulnerabilities. It is advisable to track the following metrics together.

Rather than repeating the superficial topics of profitable videos, creators should analyze the explanatory approaches that produced high retention and the questions viewers asked.

Actionable Safety-Focused Workflow

Weekly Planning

• Gather 10 potential topics and check search demand and recency.
• Confirm that each potential topic has sufficient reliable sources.
• Choose a question or explanatory approach that differs from existing popular videos.

Production for Each Video

• Read the materials and create a list of facts and a source record.
• Use AI to generate an outline and draft.
• Have a person verify claims, figures, and wording, and add perspective.
• Add voice, images, video, and music for which usage rights have been confirmed.
• Review the automatically edited result from beginning to end.
• Add any required synthetic-content disclosures and information in the description.

Post-Publication Management

• Do not make thumbnails provocative based solely on the initial click-through rate.
• Review errors identified by viewers and make corrections if necessary.
• Record both the cost and finalized earnings for each video.
• Review the channel once per month to determine whether it is becoming excessively repetitive.

Realistic Conclusions from the Case

The central point of this case is not that AI automatically generates 5 million won per month. Rather, automation reduced the time spent on script drafts, voice generation, image creation, and rough editing, allowing the creator to devote the remaining time to topic selection, thumbnails, and final review.

However, the interview earnings were not independently verified, and using the same tools and upload frequency does not guarantee the same results. The factors that increase the likelihood of success are not simply enduring for 3 months, but conducting accurate research, providing original commentary, complying with policies, managing costs, and continuously improving based on viewer data.