Culture: The Real Costs of being Real

We collected 127 stories on ArtsJournal last week. Here’s what I learned.

A couple of weeks ago, I got a PR pitch from a theatre. It praised my writing in Diacritical, not just mentioning particular posts but engaging with the ideas and building on them. It was, I admit, flattering. Was it possible, the writer asked, that we could talk, and she could show me the exciting things her theatre was doing? We are aligned.

But there was something just slightly off about the pitch. It was a little too good, too detailed, too perfectly calibrated to understand what would pique my interest. I demurred.

Book critic Ron Charles wrote on his Substack this week about a solicitation he got from  Reese Witherspoon. She was, she wrote, a big fan of his books and wondered if they could work together. They weren’t his books. You can read the hilarious ersatz correspondence in his post. It’s highly entertaining.

Then there was the story this week in the Philadelphia Inquirer reporting that authors were being solicited by Fresh Air host Terry Gross, offering interviews for a fee. Fake of course.

It’s getting difficult to trust what (or who) is real in the age of AI.

Since May, a company by the name of Runtime Wire has published nearly 2000 stories by crawling court databases, online forums, company filings, social feeds, online press conferences and other news reporting. It isn’t aggregating, it’s breaking news. This is work reporters do. But machines are now reading the exhaust of other machines and breaking stories competing with traditional journalists.

Then there’s this report from Press Gazette that traffic to the top ten US news websites is down about a third from two years ago, and that across the top 50 sites, monthly visits have fallen 37 percent since July 2024. More than a third of web pages published since ChatGPT’s release three years ago show signs of AI authorship, according to Pew research. And the most recent audit of traffic on the web shows that 53 percent of all traffic online is now machine-generated (and rising quickly). Humans are now the minority audience. When AI search summaries answer your question directly, there’s no reason to visit actual sources or websites.

Count the layers between something real and the person who eventually is able to find it. There used to be essentially one—a reporter working for a news organization. Now there’s a crawler, a model, a syndication feed, a recommendation engine, and some bot that summarizes the result so a person never even has to click. Every layer adds distance and subtracts accountability.

And it’s happening from all sides. Content may or may not be human-made. But increasingly, the audience might not be real either. So we’ve got machines making content for other machines. And, if you believe the numbers, humans are being crowded out at both ends of the marketplace.

What it means to be the “author” of something is getting murky. An MIT study reports that at the scale of a large language model, you can remove any single image or even the collected work of an artist, and the generated output won’t change. Another study reports that 97 percent of listeners couldn’t tell whether a piece of music was created by humans or by AI. Prize-winning books have been “outed” as written by AI, and there are thousands of fake AI-produced biographies in the Amazon store.

Attribution, credit and compensation all assume definitive attribution. Diffuse “influence” isn’t provenance that proves anything. It means creative ownership has become increasingly difficult to trace or to prove.

 Consider what that does to attribution, to authorship. At scale, there is, essentially no definitive source anymore. Influence survives the removal of any and every individual who supplied it. That isn’t theft you can prove, it’s a style with no parent. News with no author. Art with no artist. The record of creation isn’t being erased so much as it’s being repossessed, relocated.

So we have disruption at the creation layer. But we also have it at the audience layer.

So what if the audience is fake, if the audience is bots and machines? Traditionally, the way we have measured the commercial value of something has been by the audience it attracts. In pre-internet times, that usually meant the audience was buyers—buy the cd, subscribe to the newspaper, buy tickets to the movie. Billboard, Nielsen and NYT Books Bestseller lists reported sales as a proxy for culture that was worth our attention. Of course commercial success has never been a reliable predictor of quality, but at base it served as the currency of popular taste.

In the internet era, where digital copies of culture are essentially free to make and easy to obtain, bestseller lists and streaming charts are even less justifiable indicators of what’s getting traction in the culture.

Now in the AI era, with machines talking to machines, the audience numbers are Alice-in-Wonderland manufactured. Which isn’t just a problem for artists whose work is online, it affects all of us in finding and shaping taste and directing what culture might get made.

Media critic Matt Pearce recently argued that platform algorithms have scrambled the meaning-making that the middleware layer of culture used to do. Stripped of context, content feeds offload the sorting and aesthetic judgments the middleware used to provide. He calls it a cognitive tax on the consumer overwhelmed by having to figure out what to pay attention to, let alone make sense of it.

The obvious hope is that all of this pushes people toward things they can verify with their own bodies, and yes, there’s plenty of evidence for this. Hadestown‘s filmed stage run was billed as five nights in movie theatre, then extended to four weeks, grossing $20.7 million in tickets. Ragtime broke Lincoln Center Theater’s box office record. Nate Bargatze grossed two million tickets across 128 venues in sixteen months. Portland’s newest dance company found tens of thousands of new fans by performing at WNBA games. Somebody built a national database of author readings because there’s now big demand for these live events. A writer in The Guardian described a “collective effervescence” to explain what happened to her at a Nick Jonas concert.

But being real doesn’t make you findable or verifiable in the new machine-infested marketplaces. There’s no question that the flood of AI slop has produced a backlash in the form of demand for “real” human experiences.  Audiences want the real thing. But can you be still be found when discovery runs through layers you don’t control and can’t audit? Google’s new Backstory tool lets a fact-checker ask where an image came from. But nothing lets you ask where an audience came from or, more crucially, if it was ever there at all.

Being an artist and being findable used to be a marketing problem. Now it’s what…  Sherlock Holmes hunting for documentation, verification and provenance? My theatre PR pitcher was next level good. It wasn’t just information or messaging. It aimed to persuade by getting inside my own words and aligning itself with my curiosity.

Ultimately? It doesn’t matter whether that email was written by a person, a bot, or a person using a bot. What matters is that I couldn’t really tell, and that finding out would have cost me more than the pitch was worth. That’s the shift. Not that the fakes are getting better, but that verification got expensive.

Most people won’t pay it. They’ll do what audiences are already doing; retreat to the few things that can’t be faked because you have to be in the room. That means that the only culture we can trust is the culture that can’t scale.

Also Worth Your Attention

The National Symphony announced a season with no Kennedy Center. From September to June the orchestra will stage some 60 performances at six venues across the region. This while the Kennedy Center board voted to restore Trump’s name, the tarp still hanging over the removed lettering two months on, and a Washington Post critic arguing that closing the building may be the best of a bad slate. The National Symphony will be okay. Other institutions challenged by withdrawal of support (public media after losing federal funding, the Washington National Opera) have seen a flood of support. In fact, this could be an opportunity to find new audiences, reinvigorate programming, and focus on testing its identity.

The argument has moved to the courts. ABC sued the FCC on First Amendment grounds, calling the demand for early license review an “existential threat.” Paramount asked a court to make plaintiff states and the Writers Guild post a $1.9 billion bond to keep suing — the cost of litigation deployed as the weapon. The National Trust told the Supreme Court the administration’s ballroom request was “astonishing.” And Utah’s statewide removal of Push happened automatically: state law bars a title everywhere once three county districts ban it. Procedure is cheap to deploy and expensive to resist, which makes legal capacity a class marker. Disney can spend. A museum that just cut half its staff cannot.

Editor’s Note: These weekly essays are meant to connect stories from the week to larger trends and ideas across the arts world. Want to support our work? Subscribe to ArtsJournal’s free newsletters. Or better yet, support us with a premium ArtsJournal subscription at $5/week or $52/year. Much appreciated.

Related