Produced by ElevenLabs and News Over Audio (NOA) using AI narration.
We live in the age of short attention spans. And yet, if you want to find a recipe in a blog post, you first have to scroll past a novella detailing the chef’s personal experience with the dish. Streaming shows are long and dragging into feature film territory. Episodes of The Joe Rogan Experience podcast are sometimes longer than Avatar. Even platforms once known for short-form media are pushing their boundaries: gone are the days of the 280-character tweet; on X, a user can now pay extra to post up to 25,000 characters (This article, for comparison, is under 6,000). YouTube videos once had a hard limit of 10 minutes; now they can (and will) reach 12 hours. Even TikTok is going long and is reportedly testing a new limit of up to 15 minutes for some creators.
No doubt some of this comes from genuine audience interest. After all, length is sometimes associated with quality. Read all thousand pages of Infinite jokeor watch all three hours Oppenheimeris considered a valuable achievement in a way that watching a 60-second TikTok dissection of shower mortar is not. Sometimes storytelling deserves prodigious length.
Other times that is not the case. Online media is often filled up, not because the subject requires it, but because creators are trying to game algorithms or make more money. On TikTok, people filibuster, delay their final point, or divide their videos into unnecessary “parts”—strategies to engage viewers and increase valuable engagement metrics. All of this behavior is a side effect of our algorithmically driven reality. These systems should, at the most basic level, recommend videos, text, and anything else people post online (without them it would be impossible to search everything). But along the way, they end up creating incentives for people to produce a lot of junk – and bloated junk at that. Anything shorter than a minute isn’t even eligible to monetize on TikTok.
Some of these apps seem to realize what they’ve done. TikTok and YouTube give users the ability to speed things up and watch videos at double speed if they want. But the solution only underlines the problem: everything takes too long. A lot of this comes from all those ads that appear before videos or between paragraphs – for laundry detergent, jewelry, tax software, whatever. Any additional real estate for these ads, whether it’s space on a page or time on a podcast, is an opportunity for platforms to make more money. Longer articles and videos have more space to be placed in ads, while avoiding the feeling of overwhelm that could come from cramming them into shorter material.
In 2019, a YouTuber found that videos longer than 10 minutes earned her three times as much revenue as shorter videos, according to The edge. “It’s a lot easier to monetize content when it’s longer-form,” Scott Kessler, head of technology at Third Bridge, a market research group, told CNN last month. A TikTok spokesperson told me that creators making longer-form content more than doubled their revenue year over year by 2023. (Google, which owns YouTube, did not respond to a request for comment.)
Algorithms promise to perform one basic task: pluck the supposedly best possible content from the deluge of the Internet and serve it to users. Some algorithms may even prioritize length data (such as time spent viewing) as an indicator of quality. But that’s hard to say for sure, because tech companies tend not to provide many details about their inner workings. Longer videos, like all other content, are recommended based on viewers’ preferences and interests, the TikTok spokesperson said.
In some ways, it matters less whether an algorithm is actually hardcoded to favor longer-form videos than whether the creators do think an algorithm favors longer-form videos. Folk theories about algorithms ultimately influence what types of content are created. If people come to believe that longer videos or podcast episodes do it ‘better’, they will make more of them.
Content creators do their best with limited data. From December 2020 to July 2022, Ashley Mears, professor of cultural sociology and new media at the University of Amsterdam, conducted research while embedded at a company that created videos specifically designed to please Facebook. “At one point it was three-minute videos,” she told me. “That was the minimum amount of time it took for a video to make money.” The company would script the content accordingly, adding cliffhangers or delaying the resolution until the second minute to keep people watching. Sometimes platforms fight back against this so-called engagement bait, penalizing those who do it. “Content creators are very smart, especially if they are professionals,” Mears told me. “They study the numbers and the statistics. And they will really give platforms whatever platforms want.” But what platforms want can change, leaving creators confused.
People are afraid that generative AI will pollute the internet. But more basic forms of the technology – social media and search algorithms – have been doing that for years. “In a sense, we’re already seeing people creating culture in response to machines and these machine learning codes,” Mears explains. Generative AI makes content even cheaper and easier to produce than it is today. All this behavior is just a taste of what is to come.
None of this is to say that good writing or video content can’t last a long time. But as any quality writer or YouTuber will tell you, a story should be its own length, and should only last as long as it needs to.
