The Voight-Kampff Test for Novelists

“Artifical Eye” Created in Midjourney

Last week, a minor literary scandal popped up around a novel called Shy Girl. Advance copies went out. Readers read. Some liked it. Then someone made the worst accusation one can make these days: it was written with AI.

Cue the pitchfork wielding mobs to poured into the digital streets.

The publisher pulled the book. Readers felt betrayed. Writers felt vindicated. Editors insisted they could spot AI prose from a mile away. Meanwhile, AI detection tools flagged human-written text as machine-made, like a smoke alarm that going off when you’re pan searing a steak.

It was, in short, a mess. But a very interesting mess.

Because it got me thinking about Do Androids Dream of Electric Sheep? and its film adaptation, Blade Runner.

In that imagined world, the problem isn’t that artificial humans exist. The problem is that they’re too good. They pass. Not perfectly, not always, but often enough to make everyone nervous. So nervous, in fact, that they invent a test to tell the difference. The Voight-Kampff test measures empathy, because empathy is supposed to be the one thing machines can’t fake.

Publishing, it turns out, has been operating under the assumption that it has its own version of this test.

It’s called “voice.”

Editors talk about it the way sommeliers talk about tannins. They can feel it. They can detect it. A human voice has texture, intention, a certain ineffable hum. AI, by contrast, is supposedly smooth, empty, a wax fruit bowl of prose.

Except now the fruit is fooling people.

The Shy Girl situation suggests that whatever test we thought we had… we don’t. Or if we do, it doesn’t work very well. The detectors are unreliable. The experts disagree. And the crowd, as always, is easily swayed by a good narrative.

Which brings us to the most interesting wrinkle: some readers who initially gave the book high ratings went back and revised them downward after the AI allegations surfaced.

Nothing about the text had changed. Not a comma. Not a metaphor. Just the story about the text.

And suddenly the book tasted different.

It’s like drinking a glass of wine, loving it, and then being told it came from a box. The flavor doesn’t change, but your relationship to the flavor does. You’re no longer just tasting the wine. You’re tasting what the wine means about you.

Books, it turns out, work the same way.

We like to think we’re responding to sentences, to craft, to the arrangement of words on a page. But we’re also responding to the invisible presence behind those words. The person in the black and white book jacket photo who struggled, suffered, revised, deleted, and cared enough to get it on the page.

Reading is not just an aesthetic experience. It’s a social one. Maybe even a moral one.

So when that human presence is called into question, the experience wobbles. The floor tilts. The magic trick is revealed, and instead of applauding, the audience feels cheated.

But here’s the uncomfortable question: if you loved the book before you knew, what exactly were you loving?

If a novel moves you, makes you laugh, keeps you turning pages, does it matter whether the mind behind it was biological or silicon?

This is where the comparison to Blade Runner stops being clever and starts being useful.

The replicants in that story aren’t just convincing. They’re emotionally effective. People fall in love with them. They form attachments. The distinction between real and artificial becomes less about behavior and more about origin. Not “what does it do?” but “where did it come from?”

And origin, as it turns out, carries a lot of weight.

The outrage over AI-written fiction isn’t just about quality. If it were, there’d be no panic. Bad writing has never needed a moral framework. It’s simply ignored. The intensity of the reaction suggests something deeper: a fear that the thing we value — human creativity — might be more replicable than we’d like.

Writers, understandably, have skin in this game. If a machine can produce something indistinguishable from their work, then the floor falls out from under them. What was once a hard-won skill starts to look like a pattern that can be learned, imitated, and scaled.

So the argument becomes moral. AI writing isn’t just bad. It’s wrong. It’s deceptive. It’s theft.

And maybe some of that is true.

But it also might be beside the point.

Because the real problem with AI writing isn’t that it fails the test. It’s that it sometimes passes.

It produces something that people enjoy, and only later do they discover a reason not to.

Which suggests that our evaluation of art is not nearly as objective as we pretend. It’s entangled with authorship, with intention, with the story we tell ourselves about where art comes from and what it represents.

In Do Androids Dream of Electric Sheep?, empathy is supposed to separate humans from machines. But even there, the line blurs. The test is fallible. The people administering it are biased. And the beings on the other side are more complicated than the categories allow.

Sound familiar?

We don’t actually measure empathy in a novel. We measure our belief that empathy is present. And belief, as the internet has repeatedly demonstrated, is extremely easy to manipulate.

So we’re left with a question that’s less technological than philosophical:

Can we accept that something meaningful, even beautiful, might come from a source we don’t consider fully human?

Or do we need the illusion of a human behind the curtain to make the experience count?

I’m inclined to think that if you like a book, it shouldn’t matter how it was written.

But I’m also aware that the moment you tell someone the magician used a trapdoor, the applause abates.

And no one likes to feel like they’ve been clapping for the wrong reason.