Originally published here.

One of Germany’s leading news outlets, Der Spiegel, ran into trouble last week when they accidentally published a ChatGPT note beneath a major scoop: “If you like, I can adjust the tone and level of detail (e.g., a more sober news style vs. more magazine-like) or highlight the specific changes for you compared with the original.” Oops. The outcry was big. Many readers argued that using ChatGPT has nothing to do with quality journalism.

When reading about Der Spiegel’s hiccup, it struck me again how differently we praise or stigmatise AI depending on the use case. Inside companies, AI adoption is seen as heroism. Leaderboards track usage across departments, celebrating those who adopt the fastest and gain the most productivity. Performance reviews now classify AI adoption as poor, acceptable, or transformative. Entire careers are becoming firmly tied to AI adoption. Across sales, marketing, operations, or R&D, more AI usage equals better performance.

So why do we set such different standards for using AI to publicly write or talk? I believe it is because this work is identity-bearing, rather than productivity-bearing.

In a company, the social contract is simple: “I can produce”. You are here to produce output. The more output you can produce, the better. If you can ship faster, you are a hero.

By contrast, in the broad category of publicly speaking or writing, let’s call it authorship, the point is not to produce efficiently, but to express who you are. You are the work. The social contract is “this is me”. In fact, this is baked into the very origin of the words author and authenticity:

In authorship, we expect humans to be who they say they are - the originator - as well as to be acting on their own. Otherwise, we feel cheated on and we feel disdain towards the cheater for pretending to be something that they are not. Cheating breaches trust. It violates the rules we think we all agreed on. In Der Spiegel’s case, the author was caught using AI, like a child with a hand in the biscuit tin.

To be brutally honest, even though I use ChatGPT every day, whenever I spot an obviously ChatGPT-generated post, I struggle to take the author intellectually seriously. I think of them as a fool.

However, these current norms around AI usage likely won’t last forever. History provides two analogies for how AI-augmented authorship and authenticity may play out going forward: playback in pop music and beauty standards.

Pop stars lip-syncing to playback music rather than singing live now make us cringe. However, in the 1980s and 90s, full playback was common. The advent of TV came with complex dance choreographies that often left singers breathless, causing playback to be a good solution for fast-moving pop music shows. But eventually, several high-profile lip-sync performances drew backlash for not being transparently disclosed, including from Milli Vanilli and Ashlee Simpson. Simpson left the stage in shame after a failed vocal track on live TV in 2004, and the earlier Milli Vanilli scandal led to their Grammy being revoked and effectively ended their career. Over time, as live sound technology improved, audience expectations shifted toward preferring authenticity over faked perfection.

It used to be unacceptable to admit to getting cosmetic surgery. Women who were found out were shamed and tabloids regularly targeted celebrities over alleged procedures. Society demanded beauty and the defiance of ageing, yet any woman who openly admitted to a nose job was penalised. The contract was: you must be beautiful, effortlessly and naturally.

Over time that norm changed. Beauty became malleable and the hard work behind it more transparent. Heavy make-up applied live on social media offered a behind-the-scenes look at what is essentially engineered beauty. Full lips, high cheeks, and narrow noses became a standard, aided by scalpels and syringes. Women now openly acknowledge surgery, and proudly wear faces that obviously had work done. This signals a new social contract - the Kim Kardashian effect: from “this is the authentic me” to “this is the engineered me”. Unlike Ashlee Simpson and Milli Vanilli, Kim Kardashian doesn’t pretend to be all natural beauty. By embracing the artificiality, she is claiming a new authenticity, and thereby able to turn what was previously seen as a weakness into incredible success.

The risk of this shift in beauty standards from something that is given to something that can be formed is sameness. Many heavily sculpted faces now look alike, producing an endless scroll of look-alikes on social media. And so once again a counter-movement is rising in elite fashion and beauty circles: it despises any visible enhancements, embraces minimal or no make-up, natural hair colour, and bare nails. As the masses adopt beauty that can be formed, the inner circle rejects it to cement its status.

I expect a similar shift to happen in AI-assisted authorship. We are still in the early adopter phase, where authors getting caught using AI face intense backlash. Like Milli Vanilli in the 90s, undisclosed ChatGPT use passed off as authentic authorship may even cost some careers, while in business contexts, failing to adopt AI may do the same. But over time, we will normalise the tool and confidently shift the authorship contract from “this is the natural me” to “this is the augmented me”, much as we did with nose jobs. AI-assisted voices will become the norm and there will be a Kim Kardashian of this new form of expression, setting a new bar for successful AI-authorship. But as with beauty, the risk of AI authorship is sameness. The most prestigious voices may respond with quiet revolt, rejecting lazy perfection and embracing deliberate imperfection. We already see people roughening their ChatGPT outputs, adding back mistakes and messiness. Eventually, despite ever more powerful AI models, authenticity - “one acting on one’s own” - is where status lies.