Nearly Half of New Songs Are AI-Generated. Can You Tell the Difference?

Data released by the music streaming service Deezer reveals that 44% of newly uploaded songs are AI-generated. Moreover, 97% of listeners could not distinguish between human-created songs and AI-generated ones.

This is a discussion about music, but it extends beyond that.

Estimates suggest that around 15% of posts on Reddit are AI-generated. The “AI ratio” of content—text, images, videos, music—is rapidly increasing. In other words, the proportion of human-created content in the information we encounter daily is decreasing.

This structural change is hitting the information dissemination of local small and medium enterprises (SMEs) hard.

What Happens When Content Production Costs Approach “Almost Zero”?

Let’s start by looking at the numbers.

Three years ago, outsourcing a single blog post would cost between 30,000 to 50,000 yen. Creating a short video for social media would set you back 100,000 to 300,000 yen. What’s the situation now?

• One text article: Drafted with ChatGPT → Edited by a human. Effective cost under 5,000 yen.
• Images for social media: Canva + AI image generation. Effective cost almost 0 yen.
• Promotional video with background music: AI music from Suno + CapCut. Effective cost a few thousand yen.

Content production that once cost 300,000 yen a month can now be done for 30,000 yen. Costs have been reduced to one-tenth.

This is good news for SMEs. They can produce content on the same level as large corporations. However, there’s something important to consider here.

If everyone uses the same tools, everyone will produce similar content.

What’s happening on Deezer exemplifies this. Tens of thousands of AI-generated songs are uploaded daily, flooding the platform with “noise.” Unheard songs are increasing infinitely. As a result, the songs that truly deserve to be heard get buried.

The same phenomenon occurs in corporate information dissemination. AI-generated blog posts, AI-generated social media posts, AI-created newsletters—all of them of “mediocre quality” and “somewhat familiar”.

As the cost of content production approaches zero, the value of that content also approaches zero.

This is a structural problem.

“Human-Created” Becomes Valuable—But with Conditions

So, does simply declaring “This was created by a human” add value? It’s not that simple.

Consider this: Would you buy a pre-packaged dish labeled “homemade” at the supermarket just because of that label? No. Only when you can see “who made it,” “why they made it,” and “what feelings went into it” does homemade gain meaning.

The same applies to content. The fact that it was created by a human has no inherent value. The value lies in “this human, in this context, from this experience”.

Herein lies the winning strategy for SMEs.

Content from large corporations, no matter how much human involvement there is, is disseminated as “the corporate face.” There are brand guidelines, legal checks, and the language becomes polished. As a result, it becomes indistinguishable from AI-generated content.

On the other hand, consider SMEs:

• Photos of the CEO sweating on the front lines
• Videos of factory artisans saying, “This process cannot be entrusted to machines”
• A history of participating in local festivals every year
• The ability to remember customer names and send personalized messages

These cannot be generated by AI. Because they are “facts.”

AI can produce “plausible text,” but not “what actually happened.” The greatest weapon that SMEs possess is a verifiable real story that is not fictional.

What Specifically Should Be Done: Three Practices

1. Use AI to Ensure “Quantity” and Humans to Create the “Core”

There’s no need for everything to be created by humans. In fact, that would be inefficient.

A practical distribution would look like this:

For 50,000 to 80,000 yen a month, both AI-generated content and core content that only humans can create can be produced. This would have cost 300,000 to 500,000 yen three years ago.

The key point is to clearly separate the “core content”. It’s not about leaving everything to AI or having everything done by humans. Draw a line for “this part must be done by humans.”

2. Switch to Content That Leaves “Evidence”

Content that earns trust in the AI era has a common trait: it has evidence.

• Photos of the manufacturing process (unedited)
• Actual interactions with customers (with permission)
• Handwritten notes from employees
• Stories of failures and subsequent improvement processes
• Episodes with specific dates and locations

These serve as evidence of “what really happened,” rather than just “plausibility.” This is the biggest differentiating factor from AI-generated content.

A local food manufacturer captures the morning scenes in their factory every day with a smartphone and posts them on Instagram. No edits, no filters. The cost per post is almost zero. Yet, there is “today’s reality” that cannot be created by AI. Their follower count may be lower than that of major food manufacturers, but their engagement rate is reportedly over five times higher.

3. Intentionally Design “Traces of Human Involvement”

This may sound paradoxical, but we live in an era where human-like qualities must be “designed” to be conveyed.

This is because AI-generated content can be “somewhat human-like.” In some cases, AI-generated text is even easier to read than text written by humans without thought.

That’s why it’s essential to intentionally include elements that make it clear, “this was thought up by a human.”

• Proper nouns: “Last week, I heard a story from Mr. △△ in ○○ City.”
• Granularity of numbers: Instead of saying “sales increased,” say “March orders increased by 23 compared to the previous year.”
• Emotional fluctuations: “Honestly, I thought it would fail at first.”
• Imperfections: A slightly rougher text feels more human than a perfectly polished one.

AI excels at writing “generally correct things.” Conversely, “specific, individual, and imperfect things” are areas where AI struggles, and where humans shine.

The Real Question Is Not “What to Disseminate”

The figure of 44% on Deezer indicates not just that “AI is impressive.”

When the cost of creating content approaches zero, the value of “creating” itself diminishes.

What remains valuable?

• Judgment on what to create
• The will to create it
• Trust in who is saying it

These three cannot be replaced by AI. And these three are stronger in SMEs than in large corporations. This is because the decision-makers in SMEs are visible. The CEO can decide, “We will do this,” and disseminate it under their name. This is a feat that large corporations cannot achieve.

What SMEs should do in the AI era is neither to fear AI nor to ignore it.

Delegate what AI can do to AI, and spend time on what only humans can do.

And what only humans can do is not just writing beautiful sentences. It is narrating what happened in their own field, in their own words, under their own name.

That will be the only form of dissemination that does not sink in the flood of AI-generated content.