The internet is, or at least perhaps was, a sacred thing. The hackers and geeks who popularized the protocols envisioned an open, free place, where authentic, free speech would elevate the human condition. AI could spoil that promise by filling this place with a tidal wave of garbage. I went looking for a writer who could help us understand this current moment of online distrust and thought Jessi Hempel was the perfect person.
She spent nearly two decades reporting and writing about the business of technology for BusinessWeek, Fortune, and Wired. In recent years, she’s shifted her focus to the ways in which technology is reshaping our careers and working lives. Now she is the senior editor-at-large at LinkedIn, where she hosts the award-winning podcast Hello Monday. Her weekly newsletter is worth reading, and I recommend it.
As a reminder, I’m out on paternity leave till November, and commissioned a series of original essays from the world’s best technology writers while I’m out. Thanks for reading!
First though, this newsletter is brought to you by Lovable.
I’m obsessed recently with the idea of disposable apps. These are little pieces of software that are valuable for a day or a month, but after that, are never opened again.
Nobody builds these because the coding used to be hard (not anymore, thanks AI) and maintaining apps was annoying. That second bit is what Lovable solves. All the infrastructure and security concerns surrounding an app can be automated away. Now if you have an idea, 30 minutes later you can have a functional, simple app for less than the cost of a Big Mac Combo.
What would it be like if every idea you had became a functioning MVP with essentially zero effort? Find out and try Lovable today.
And with that, here’s Jessi.
The internet has become a test that none of us agreed to take. As we scroll and read, we constantly find ourselves asking: Human or machine?
I can’t overstate the mix of emotions I feel when I discover I’ve been duped. It happens all the time. I’m drawn in by a sappy LinkedIn post in which a friend talks about her experience with a bullying boss when I realize the anecdote seems a little too…canned. I see that tell-tale “It’s not this, it’s that” construction and my stomach drops. Shakespeare’s phrase reverberates through my mind: Et tu, Brute?
All signs point to the fact that my paranoid second-guessing is justified; in a recent analysis of more than a million social media posts, the AI detection software Pangram found that one in four longform content posts is fully AI-generated.
This is the alienating milieu into which Stanley Druckenmiller emerged in August. He’s the hedge fund billionaire who published an op-ed in The Wall Street Journal that was very obviously crafted by AI. Readers called foul. Druckenmiller acknowledged it and compared the process to using a calculator. He promised the ideas were his own, and The Wall Street Journal even defended the work.
And yet something about the whole affair still felt wrong to a lot of us. Nick Thompson, CEO of The Atlantic, thought something more fundamental had happened. One of the problems was “authenticity,” he said in a video on LinkedIn. “Your name’s on it. You should have written all the words and you should stand by it.”
That second part of Nick’s observation is obvious: an author should stand behind her work. But must she have written all the words? Our discomfort arises from the fact that we assume that a public voice belongs to a person, and we’ve spent decades learning to recognize that person through signals embedded in their work. We assume we can judge their authenticity. Now that assumption is breaking down.
AI makes it possible for anyone with something to say or sell to manufacture a convincing voice, and the technology keeps getting better. We’re entering an internet on which the voice itself can no longer tell us very much about the person behind it. And just as a Fendi bag would not be worth $4,000 in a world where no one could distinguish it from a perfect counterfeit, a public voice derives much of its value from our ability to connect it reliably to the person whose name is on it.
So how do we convey authenticity in an AI-driven future?
To understand where all this is going, we have to begin with a working definition for authenticity. To do that, I called Alice Marwick, who is now the director of research at the nonprofit research institute Data & Society. I met Alice when I was asking similar questions of our media ecosystem back in 2005; back then, she was still deep in her PhD work, looking at how people performed identity online. Alice will tell you that authenticity is a judgment made by an audience. We encounter a public performance and decide whether it corresponds to the person behind it. Crucially, she points out that it is a relative judgement. “Authenticity is always something we see in opposition to something else,” she explains. We arrive at authentic by calling out what is inauthentic.
Starting long before social, every new medium had its signals of authenticity, manufactured to produce a feeling of presence. My favorite example of this is reality television. You, the viewer, knew that to a large degree, the material was scripted. Of course it was fake! And yet in watching, you tried to discern where the script faltered and our characters were just themselves. The sport of it was finding the authenticity!
Social media made personal authenticity especially valuable as institutional authority weakened. Ordinary people could speak directly to audiences. Blogs went mainstream; Facebook spread beyond college campuses. It’s impossible to overstate the significance of this shift in the way we consumed media. When Fortune published my first cover story on Facebook in 2010, several readers called me directly through the magazine’s main switchboard for user instructions. “There’s no customer service phone number on this site!” they would complain, and then they’d ask for help setting up a profile. Beyond using tools differently, readers were shifting their perceptions of what constituted authority. The institution no longer did all the work of establishing credibility; increasingly, audiences were also asking the individual speaking to them: Can I trust you?
This is about when the word “creator” arose to encompass people, from people like me to fashion bloggers to the children’s YouTube star Blippi, who build an audience around a body of work on the web. The successful among us have become extremely sophisticated at performing the signals of “this is really me.” On the social web, a vocabulary of authenticity is constantly evolving; audiences seek informality, intimacy, spontaneity, direct address, disclosure. Audiences have always understood that these were conventions, but that didn’t necessarily make the performance fraudulent. It was the art of the form. We learned to read these signals as evidence that behind the mediated product was a real person giving us something of herself.
These days AI can manufacture enough of the signals we’ve learned to associate with authenticity that even those of us who study language on the internet for a living can’t always tell what’s what. There’s emerging evidence to support this. I was captivated by a February 2025 study published in Scientific Reports in which researchers asked people to evaluate personal advice without telling them whether a human or ChatGPT had written it. The ChatGPT advice won: on a seven-point scale, participants rated it a full point higher in quality, about two-thirds of a point higher in effectiveness, and 0.39 points higher in authenticity.
Of course, when participants learned the advice came from ChatGPT, they devalued it; that’s the confusion inherent in decoding authenticity right now. We’re trying to understand tomorrow with yesterday’s tools; in that way, we’re not so different than those Fortune readers looking for the 1-800 number on the Facebook website.
Audiences will continue to seek out authenticity as the determining factor for whether they will trust a creator, be it Mark Rober or Mel Robbins, a journalist like me or a brand, but the signals will change. The people who nail this next version of media, harnessing its tools to deepen trust with audiences, will discern these signals early and design into them.
A strong reality here is that it might get a lot harder to find those trustworthy, authentic-seeming people. As I was thinking through all of this, I called up Max Spero, the founder of Pangram, to get his take on where authenticity moves in the future. We brainstormed for a while on what our social feeds might look like in a year, and whether people might still be complaining about AI slop. He described one particularly dystopian vision for a future in which, instead of using ChatGPT to pen our Substack notes and Instagram captions, we deploy our agents to do it. But a cacophonic, alienating world in which machines spit out all of our social exhaust will only make us cling even more tightly to the creators we are able to discover. In an even noisier, messier world, knowing that work is created by a particular person with a point of view and something at stake may render the work more valuable.
Which brings us back to Druckenmiller. Does it really matter whether he typed every word? We’ve always tolerated mediation between a public person and an audience. Before you read this essay, I will send it to Evan, who will edit it. What audiences want to know is whether the expression meaningfully belongs to the person claiming it. Did I make the important choices? Does it represent what I actually think?
Authenticity, Alice reminded me, is a judgment made by an audience. And increasingly, I think that judgment will have to be made from what she calls “co-presence over time.” This is the accumulation of encounters through which we come to know how someone thinks, what she cares about, how she changes her mind, where her judgment is good and bad, and whether her words and actions hang together. There are no shortcuts for this. We’ll have to get to know each other online just as we do in the real world, one interaction at a time.
This emerging understanding of how to manufacture authenticity will require new labor on behalf of both the creator and the audience. The person speaking has to demonstrate intention, judgment and accountability over time. And the audience has to pay attention over time.
If this is starting to sound suspiciously like a real relationship rather than the parasocial relationships that have defined the social web, well, it is. Perhaps this is the opportunity buried inside the authenticity crisis. Increasingly, AI will make it impossible to prove, sentence by sentence, that a human being was there. But it may force us toward a more demanding definition of authenticity: a relationship in which a person has become knowable enough to be trusted.
