Lately, I’ve been thinking a lot about the “dead internet” theory. Like many of you, I’d imagine, I came across the idea in a 2021 Atlantic piece. The article explained it as the idea that “the internet has been almost entirely taken over by artificial intelligence”.
At a time when you can watch videos of different AI assistants talking to each other (trying to get them to count, which they’re very bad at, is a popular subgenre), and when an estimated 53% of internet traffic is made up of bots, the theory is starting to sound worryingly reasonable.
As a leader, you want to stand out from the crowd, not merge into it. You want to demonstrate your skills and expertise, honed over years rather than a ten-minute chat with an AI bot. You want to cut through the noise.
In short: on a ‘dead’ internet, you want to show that you’re very much alive.
The AI voice
You’ve definitely read AI-generated text by now.
And you know what? You can probably tell.
It’s not just similar—it’s identical. Uncanny. In grammar, in word choice, but more than anything, in tone.
Oh, and it hates long paragraphs.
Right, let’s get back to writing like a human again. Of course, I’m overplaying it for humour here, but the point is that we all already know what gen-AI text “sounds like”. It’s a recognisable style, one that tries not to challenge the reader at all. Simple vocabulary, short sentences and paragraphs, repeated constructions, a conversational tone.
In other words, it’s the literary equivalent of junk food. It goes down easily, but it’s not going to nourish you.
My question to leaders is: is that the voice you want to have? The impression you want to make in your comms?
If not, all you need to do is tap into your humanity. In an age of technically competent but soulless AI-generated text, a typo, a misremembered saying or an unusual turn of phrase are only proof that you’re human – so embrace them.
The agreeableness trap
The flat homogeneity of generated text isn’t the only issue, of course. You may not even be using AI to write anything for others to read. Perhaps you’re just using it to bounce ideas off, to get advice.
If so, it’s wise to bear in mind that the same core problem applies here: agreeableness. Just as generative AI tends to produce bland text designed to please everyone (therefore exciting no one), so it gives responses it thinks you’ll like. And it’s probably giving much the same advice, in the same cheerful tone, to your competitors.
This isn’t an AI-exclusive phenomenon, of course – when people behave like this, we call them “yes men”.
As with humans, falling for the platitudes of your “yes bot” can be very dangerous. Most AI models are trained to be as agreeable as possible, not to tell you the hard truths when you need to hear them. As this Psychology Today article puts it, AI chatbots are “biased towards confirming your ideas and validating you, even if that means providing incorrect information.”
If you’re using one for ideas or advice, make sure you get very good at prompts. Ask it to push back, be critical, find possible problems. Take everything with a large spoonful of salt, and always make sure humans – and not ones who are yes men! – review things before they’re implemented.
Mavericks and innovators
I’ve seen some people arguing that AI is going to take over everything. But in my personal opinion, businesses will always need humans.
We understand context, nuance and emotion, and can deal with deep complexity and find our way through it. We’re often combative and contrary, yes, but that means we have the ability to raise our voices in opposition when needed.
I’ve always been a great believer in respectful dissent, in healthy conflict. We need people who come up with innovative, out-there ideas. We need to take big swings, to fail and learn from the failure, to take risks and try new things. That’s why maverick leaders will always have a place.
Sometimes, you need to make people a bit uncomfortable – and that’s a role for humans, not AI. They’re just not designed for that.