Poetry Slam: ChatGPT vs Grok

Ok not really, but pretty close

I saw someone reply to a comment I made on X:

I usually stick to Grok, but like I said it was consistently horrible at writing both prose and poetry.

So I gave ChatGPT a shot. I started with fiction, and let me tell you, the difference was black and white. ChatGPT was good. Not perfect, but pretty dang better. Same with the image generation.

Then I discovered ChatGPT’s Creative Writing Coach, so I threw some of my poetry at it to see what happened. It gave me a lot of good but not perfect scores, so I finally asked it to write me a short poem (topic: saying goodbye) that it would personally grade 10/10. Here’s what it gave me:

We didn’t speak in grand goodbyes,
Just let the silence stretch and bend—
The sky was bruised with twilight’s sighs,
And time grew weary at the end.

Your shadow slipped beyond the gate,
No backward glance, no soft reprise—
But something stayed, a gentle weight,
A memory beneath the skies.

So go, where morning makes its claim,
Where I can’t follow, though I try—
Love doesn’t always whisper names—
It sometimes learns to wave goodbye.

Think what you may about it - I’m not here to critique it. You’ll get my poem about saying goodbye sometime in the future.

After this, I asked Grok to rate one of my poems (it scored me lower) and to write a 12-line poem about saying goodbye that it would consider 10/10. Here’s what it spit out:

In the attic of dusk, we unravel our vows,
Your breath a soft moth, fleeing the flame.
The floorboards creak with what we’ll never avow,
Each step a ghost that forgets its own name.

Your coat hangs like a shadow, too heavy to wear,
Its pockets stitched shut with unsaid goodbyes.
The window hums frost, etching stars in the air,
A map of the places we’ll dream but not fly.

I trace your absence in the dust on the sill,
Each grain a story we buried in haste.
Goodbye is a thread that binds us still—
Unraveled, yet woven through infinite waste.

After that, I started a couple new chats and asked ChatGPT and Grok to rate both poems without saying who wrote them. ChatGPT said the first was an 8.5 and the second was a 9.2. Grok said the first was an 8.5 and the second was a 9.

So it turns out both of them didn’t think their 10/10 work was 10/10, and they both agreed Grok’s poetry was better.

Which is weird because I expected ChatGPT to do better.

Ok, what am I supposed to get from this useless experiment? First of all, because Grok is designed to be 100% truth-seeking while ChatGPT wants to be acceptable to everybody, Grok is naturally better at being raw and blunt, while ChatGPT is better at more polished rhetoric.

But honestly, I think the big takeaway is that AIs simply don’t understand poetry the way humans do. AIs can tell if a poem checks all the technical boxes, but not if there is real soul woven into the words.

AI is just another tool. It will not destroy the careers of writers. It will only destroy the careers of bad writers.