It finally happened. Artificial Intelligence wrote advertising copy that performed better than that written by professional copywriters. Feels like that moment when chess master Garry Kasparov was beaten at the game by a computer. We were all kind of bummed but also surprised that it hadn’t happened sooner.

The moment for copywriters came last month during a trial run of an AI writing software from tech company Persado for JPMorgan Chase. They pitted human-written ad lines versus software-written ad lines to see which one got the most clicks.

It was a resounding defeat for the human copywriters with AI headlines performing 2-5 times better. The results were so conclusive that JPMorgan inked a five-year-deal with Persado. According to the companies, the idea of the deal isn’t to replace copywriters, but to make what copywriters create “more powerful.” So not AI writers, but AI editors. Literal robot overlords.

But then you look at the copy itself from the trial:

Human Copywriter: Access Cash from the Equity in Your Home

AI Copywriter: It’s true—You can unlock cash from the equity in your home

---

Human Copywriter: HURRY, IT ENDS DECEMBER 31 Earn 5% Cash Back at Department Stores, Wholesale Clubs

AI Copywriter: REGARDING YOUR CARD: 5% Cash Back is Waiting For You

---

Human Copywriter: Go paperless and earn $5 Cash Back.

AI Copywriter: LIMITED TIME OFFER: We’ll reward you with $5 Cash Back when you go paperless

First, you could argue whether this represents actual writing, but that aside, the AI headlines weren’t anything revelatory. The copy is cliché ad forms the success of which could have been parsed using human writers and A/B testing. But, more interesting, the AI versions were overall…cheesier, and in at least one of those cases, deceptive. But that’s what makes them work better. Media has already learned that click-bait headlines beat accuracy, sincerity, and cleverness. They work because humans suck. And most copywriters, even the most jaded, soul-shorn among us, still blanche against being too cheesy and deceptive. AI has no such shame. HAL 9000 voice: “You told me to get clicks, Dave, and this is what you homo sapiens click.”

Don’t get me wrong. I’m not sentimental about copy. I think that eventually AI will be great at not just all ad copy, but novels and screenplays and legislation, too. That doesn’t depress me (there’s too much else about being a writer that is more depressing, so get in line, robots.). But what happens when AI is catering to our worst instincts, as it will inevitably do if it’s programmed to create ad copy to “get clicks” or a novel that “people will buy.”

Science fiction authors have been writing about this exact phenomenon for years, that our creation of artificial intelligence will end up revealing the worst things about ourselves. Like Westworld. Or anything Asimov wrote. Short Circuit II.  And, of course, it’s already happened elsewhere here in the present. Like when Microsoft launched that AI chatbot onto Twitter to learn and grow as it conversed with other people. It become a racist, misogynist creep.

Granted, this might be an overly grand conclusion based on AI sticking “It’s true” in front of an email subject line. And this kind of task is something most copywriters would be glad to hand off, anyway. But the implications are definitely…cheesy and possibly deceptive. Look at this quote from the CMO of JPMorgan, Kristin Lemkau, about the test. In an internal memo to, I assume, stop all her copywriters from jumping out their office windows, she wrote “Machine-learning is the path to more humanity in marketing.”

I…can’t even parse that. She should’ve had AI software write it.