The Days When "Made with AI" Impressed People Are Over

More and more people are posting images and videos made with AI on social media. Two years ago, that alone might have turned heads. These days, I sometimes feel that the only person enjoying an obviously AI-made post is the person who posted it.

And yet some AI work is interesting even when I know AI made it. I wanted to work out what separates the two.

The posts I find cringe

Someone has AI make an image or some text, posts it as it is, acts as if they came up with it and wrote it themselves, and says, “This is the latest thing.” When I see a post like that, it makes me wince.

Using AI isn’t the problem. All you have to do is write “made with AI.” What’s cringe is hiding it and taking the credit.

Research points the same way. In a Duke University study, people who used AI were seen as lazier and less competent. The users knew it, and were less willing to say they had used AI (as described by PsyPost; those were workplace evaluations, not social media). In a Columbia University experiment, work that people could not tell apart from human work was rated lower for creativity and value once it carried an “AI-made” label (paper). About nine in ten US consumers say they want AI labels on images and video (Fractl).

I get the feeling that older guys do this more.

It’s been spotted, and people are bored of it

I think work that is obviously AI has already been spotted, and people are tired of it.

In a US survey of 18-to-27-year-olds, 41% actively dislike AI-generated content on social media, 31% are wary because it’s hard to tell what’s real, and 28% find it entertaining (Attest). If everyone uses similar AI, similar things line up. In one experiment, short stories written with AI ideas were rated higher one writer at a time, but the stories ended up resembling one another (as described by PsyPost).

But I suspect only the work with strong AI tells is being spotted. In a Yale poll, most 18-to-34-year-olds were confident they could tell, yet they got headshots right only about half the time (reported by The Nation). The AI work nobody can spot flows by without anyone knowing.

What feels most worn out to me is past human work remade with AI. When the Ghibli-style images took off, opinion split (AFP, in Japanese). But the criticism centered on style and copyright. What bores me isn’t that. It’s that there is nothing new in it.

The AI work I still watch

Even so, some work is interesting even though I know AI made it. History videos are one example. Some of them make me think, “Oh, there’s an angle like that.”

It’s certain that AI made them. But in the sense that they create a feeling that wasn’t there before, I think they’re creative. That’s the creativity of the person who thought of the angle.

With history, though, there’s something to be careful about. A study of history videos on TikTok found plenty of historical errors in the captions, and reactions in the comments split sharply by subject. Videos about the Holocaust drew especially large amounts of hate and disinformation (paper). Even with a good angle, the responsibility for checking the facts stays with the person who made it.

Creativity goes against the statistics

AI builds plausible answers from a huge amount of data. In other words, its output comes from the statistics of what most people consider right. Creativity sometimes has to do the opposite. You go against what most people think of as success. If all you do is make, neatly, what everyone says is good, it’s already inside the statistics. Stepping outside and deciding, “This might be interesting,” is human intuition, I believe. It’s hard to capture that intuition in numbers.

I think that’s why AI work gets old on social media. Whatever sits in the middle of the statistics becomes familiar. If a human’s angle has been placed off to the side, people will watch it even when they know AI made it.

If you’d like to talk about running social media or making things with AI, get in touch below.