AI in wine writing: is it that bad, or does it “just hit different”?

The missing middle ground

The wine writing world is blowing up about a piece that won an award, despite having clearly been AI generated. But for those of you who don’t already subscribe to 30 wine Substacks all banging the same drum: the essence is that there’s such a thing as the 67 Pall Mall Global Wine Communicator Awards with a prestigious cast of judges. About a week ago, they handed out half a dozen awards, including one for short-form writing that went to a little-known author who compared a 1969 Bollinger VVF (that’s Champagne to you and I) to a Stradivarius Violin. Compared to, not compared with – as in, shall I compare thee to a summer’s day? The premise of the piece, if I’m not oversimplifying, is that they’re both rare and collectible and can no longer be recreated because the conditions that made them no longer exist (which I suppose goes for just about everything – especially when it comes to wine and wood instruments).

To anyone who hasn’t been living under a literary rock, the writing presents many of the telltale signs of AI: the short, staccato sentences that now seem to be all over my LinkedIn feed (apparently people’s attention spans are so short we can no longer handle paragraphs); the “not this, but that” constructions; the rhetorical tic of every observation arriving in threes. Understandably, plenty of people are angry that the judges somehow missed all of this and awarded the prize anyway.

But this particular situation has also led to much discussion about the rise of AI in wine writing in general. The argument that Substack’s wine community seems to be making is that AI cannot taste a wine, nor have a lived experience triggered by a wine – ergo AI is bad for wine writing. And if it were as simple as that, I would wholeheartedly agree.

If the prompt is only: “generate an entertaining article about Burgundy Pinot Noir” then of course the machine is just going to spew out 800 generic, impersonal words that are of limited value to anyone. But I’d argue that’s often far from the case.

I have a professional acquaintance who is unquestionably a wine expert. They have numerous qualifications, they teach classes and, having sat next to them at tastings, they have an excellent palate. They also happen to be a non-native speaker of English and their language skills aren’t particularly good – whether spoken or written. This isn’t a criticism: my messages in Dutch aren’t too hot either. It’s a fact of learning a second language as an adult.

In the past year or two, however, they’ve started posting articles in perfect English. The research looks to be impeccable (honestly, I wouldn’t know – they’re far more qualified than I am) and there are genuine, personal opinions that make me confident they’ve tasted and assessed everything in the articles. But the writing itself has clearly had more than a little help from AI. I don’t just mean proofreading, or even fixing shaky English grammar. Everything from the sentence construction to the word choice screams AI – despite the fact that there are plenty of parts written in the first person.

This isn’t “AI slop”. Neither is it a light edit of a human’s work. A human thought about it, but let the robot do the writing. It’s a grey area that people on both sides of the argument seem to overlook – mostly or entirely. We see a piece that has the rhetorical hallmarks of AI all over it, and we throw it into our mental bin as being “AI generated”. It’s off-putting, knowing that a machine wrote the thing.

But the question is: does it matter? Sure – it matters to writers like me (and the rest of WineStack). I used to compete only with other writers; now I compete with anyone who has something to say. But I’m not sure it matters to the intended audience, to the wine consumer. In the cases I’m talking about, the reader is still receiving well-researched information and a real person’s tasting notes – albeit with far more non-literal uses of words like “quietly” and “intentional” than can ever be justified.

Reading something that’s been written by a robot – so long as the input that informed it came from a person’s knowledge and experience – is certainly no worse (and is frequently better) than reading the poorly constructed Dutlish I so often encountered a few years ago. Although if I’m forced to concede one more time that something “just hit different,” I might be minded to throw my phone into the nearest canal. (Also: it’s differentLY, for the grammatically challenged machines reading this.)

Random picture of the human tasting all the wine that goes into the writing of this Substack (neither quietly nor intentionally)

Click a button, any button!