As AI technology and its use cases for podcasting, audiobooks, and video creation evolve, so will the rules around the use of that technology. Major platforms are gradually rolling out new guidelines about using AI-generated content, but it can be a lot of work staying up to date with who’s banning what. 

If you’re a creator, you may be wondering things like whether AI-generated content is allowed, how you can make sure someone else doesn’t use your content to create a deepfake, and whether it’s OK to generate AI versions of existing artists and creators.

The answers to those questions are not exactly black and white; however, we’re at least starting to get some responses to how the major platforms will handle allowing and moderating synthetic and AI-modified content on their sites.

Let’s take a look at how Spotify, Audible, and YouTube are approaching AI content on their platforms.

Spotify’s rules around AI

Some AI-generated content is allowed at Spotify—but not all.

An example of an acceptable form of AI that Spotify would allow—according to Daniel Ek, Spotify’s CEO in an interview with BBC—are tools like auto-tune (and Studio Sound) that help to improve music and sound quality. 

On the other end of the spectrum, Ek says that music that impersonates artists is absolutely not acceptable. 

There’s still a murky middle: AI-generated music that was inspired by an artist that doesn’t mimic them completely. That’s not banned, but what qualifies as “inspired” is pretty unclear. 

It’s important to note that this challenge isn’t limited to Spotify, either. Many distribution platforms, like YouTube as you’ll read below, are having to determine where the imaginary line exists between deepfakes and a, shall we say, creative approach to artistry. Without any legal regulations, the answer is determined on a case-by-case and company-by-company basis.

It’s also important to note that Spotify has spoken out about letting AI train on the content that’s on their platform: It’s a big no. 

This makes legal sense, as other industries are currently fighting intellectual property lawsuits based on the alleged misuse of AI training by companies like OpenAI and Microsoft.

What this means for a podcaster or musician:

Spotify’s commitment to not allowing AI to train on its content means that putting your content on Spotify won’t risk it being used in a future AI model.

But be very cautious if you’re going to use AI tools to create inspired synthetic media. Spotify is committed to removing deepfakes and manipulated content—whether it’s an entire piece or just a clip. And with limited information about what constitutes inspired versus copied, it may not be worth the risk.

What this means for a Spotify listener:

Unlike the next two platforms, there is no clear tagging system on Spotify that indicates a part or the whole of what you’re listening to is synthetic. It’s important to think critically about what you’re hearing and do independent research to identify whether what you’re listening to is actually what it seems.

Audible’s rules around AI

In late 2023, Audible launched a beta program for self-published authors using Kindle Direct Publishing (KDP) to test their new tool Virtual Voice. At that time, there were approximately 7,000 audiobooks on the platform that were created using the new technology, and there were over 10,000 options available just two months later.

But the Audiobook Creation Exchange (ACX) has different rules. That’s the platform responsible for quality control before authors and publishers can put their audiobooks on Audible. It’s owned by Audible but it operates independently. The ACX requires that all content must be narrated by a human, unless otherwise approved.

So which is it?

It seems that Audible is moving into a future that allows synthetically narrated books. It just doesn’t have its child companies quite in sync about how this will happen. 

It’s also important to point out that Audible takes efforts to protect their creators’ content. According to an ACX help guide, “This includes the use of a variety of technologies including DRM, secure streaming, and other encryption technologies when delivering content to customers, which also helps to protect against unauthorized modification or distribution of their content.”

Note: Kindle Direct Publishing does allow AI-generated content beyond voices—i.e., the writing, the imagery, and/or translations—and just requires authors to disclose that. 

What this means for an author: 

If you’ve self-published on KDP and would love an audiobook to accompany your writing, you now have an opportunity to create one without having to record it yourself or hire a voice artist.

Just don’t use ACX to upload any synthetic content since it’s against their current guidelines.

What this means for an audiobook listener:

Your audiobook options are growing. With Audible testing out Virtual Voice, you now have the option to listen to AI-narrated audiobooks. 

YouTube’s rules around AI 

Unlike the other platforms, YouTube is clear on their approach to allowing and moderating AI-generated content on the platform. It seems they’re taking a strict stance in regard to their music industry partners and using more lax guidelines for their creators.

Similar to Spotify, some AI-generated or manipulated content is allowed on YouTube, but deepfakes aren’t tolerated. In addition, they require that any YouTube users who want to upload realistic-looking synthetic content will be required to disclose this. In doing so, that content will get a label that indicates to the end user that it’s altered or synthetic content.

The announcement on YouTube’s official blog indicates that this disclosure is especially important for sensitive topics, which they list as being things like public health crises, ongoing conflicts, elections, and anything about public officials. And these specific topics will receive labeling in multiple areas so there won’t be any confusion.

Creators who don’t disclose altered or synthetic content will be subject to penalties like content removal and suspension from the YouTube Partner Program.

If you’re worried about deepfakes, YouTube plans to allow viewers the ability to request removal of videos that “simulate an identifiable individual, including their face or voice.” 

But don’t get too excited yet. Removal of the content will be pending a YouTube review process. And this is where things get tricky.

Because the legal landscape surrounding AI and intellectual property is still being hashed out, it’s up to the YouTube team to determine what’s fair use and what isn’t. YouTube states that some of the factors that will go into their determination are things such as: 

• Whether the content is parody or satire, 
• Whether the person making the request can be uniquely identified, 
• Or whether it features a public official or well-known individual, in which case there may be a higher bar.

What this means for a YouTuber: 

If you plan on using any AI-generated or altered content on your YouTube channel, it’s your responsibility to not only understand their rules (and to pay attention as they evolve over time), but also to make sure that you’re following their guidelines, by doing things such as tagging altered or synthetic content when you upload it.

What this means for a YouTube viewer: 

It’s important to be discerning of the content you consume on any platform, and YouTube is no different. To start, keep an eye out for any of the platform tags that make viewers aware the content was not created by a human.

In addition to that, do your own due diligence. This may mean further researching any claims or stories you see on YouTube to be sure the information you’re receiving is accurate.

How will consumers know whether content is AI generated?

None of these major platforms want their users to feel unsure about the authenticity of the content they’re consuming. To the best of their ability, they’ll enforce requirements that all AI-generated or manipulated content must be tagged so the audience is aware that some or all of the content they’re consuming is synthetic.

Take Audible as an example. Consumers are made aware that an audiobook is virtually narrated by a tag underneath the book image that allows them to listen to a sample before purchasing. As long as you’re paying attention, it’s easy to identify that the narration on these audiobooks is not done by a real human.

What I like best about their approach is the option to listen into the synthetic narration before purchasing to decide whether or not they like the sound. I felt sure that I wouldn’t buy an audiobook that was narrated with the Virtual Voice tools. I’m picky enough about human narration of audiobooks! But I’ll admit, listening to the samples changed my mind.

The only downfall to tagging synthetic content—whether on Audible, YouTube, or any other platform—is that the responsibility is on the creator to actually disclose when the content is synthetic or modified in any way. It’s safe to say that there are people who will upload and not properly identify this content, which is why end users need to be discerning when they consume any content and do additional research if necessary.