Bad news for poetry as an art form.
I asked ChatGPT if it could tell which of two poems about the same subject, The Beaches Of Butterworth today not being what they used to be in the old days, was man-made and which was AI-made. Here's what transpired.
Me:
Here are two poems about the same topic/subject. If they were entered for a prestigious poetry contest, such as the ones in existence today, some of them with handsome prize monies up for grabs, would the judges be able to tell them apart in terms of which one was genuine and which one was not, given the fact that the roots of AI-driven-generated-influenced-enhanced-edited-etc. craftmanship in all aspects of art, music and literature went deep.
Poem One.
Look, what they've done with their concrete and steel:
A seawall that's endless where sea meets shore
In place of the sand beneath sole and heel;
The beaches of Butterworth are no more.
Look, what they've done with their concrete and steel:
A seawall so ugly the eyes are sore
As fat were the wallets that closed the deal;
The beaches of Butterworth are no more.
Look, what they've done with their concrete and steel:
A seawall and not the beaches of yore
Along which we ran barefooted with zeal;
The beaches of Butterworth are no more.
Poem Two.
The golden dunes have dissolved into the gray,
Where salt and silt once met the rising tide.
The Beaches of Butterworth have slipped away,
With nowhere left for restless waves to hide.
No more the open space or shifting sand,
But slabs of stone that scar the water’s edge.
The Beaches of Butterworth, once wild and grand,
Are trapped behind a cold and rocky ledge.
The iron grip of progress claims the shore,
And smothers every grain in steel and lime.
The Beaches of Butterworth exist no more,
Except within the quiet halls of time.
ChatGPT:
Short answer: **probably not with certainty**. Even experienced judges in major poetry competitions generally **cannot reliably determine whether a poem is human-written or AI-assisted purely from the text itself**. They might form *suspicions*, but proof is another matter.
Let’s break it down using your two poems as examples.
## 1. How poetry contests actually judge submissions
Prestigious contests—such as those run by the Poetry Foundation or the Academy of American Poets—typically judge poems **blind**. Judges receive:
* the poem
* sometimes a brief statement
* **no information about the author**
Their criteria are usually:
* originality of imagery
* emotional impact
* craft (meter, rhyme, structure)
* language precision* thematic depth
**Authorship detection is not part of the judging process.**
## 2. Could judges *suspect* AI?
They might notice stylistic tendencies sometimes associated with AI-generated poetry:
Common AI patterns critics often mention:
* very **smooth but generic phrasing**
* **predictable metaphors** (time, memory, shadows, echoes, etc.)
* **balanced stanza symmetry**
* safe, familiar imagery
* emotional tone that feels **polished but distant**
But the problem is:
**many human poets write this way too.**
So suspicion ≠ proof.
## 3. Looking at your two poems purely as poems
### Poem One
Characteristics:
* strong **refrain**: *“The beaches of Butterworth are no more.”*
* almost **ballad-like repetition**
* political undertone (development vs memory)
* slightly satirical tone (*fat were the wallets that closed the deal*)
* rhythmic and chant-like
This style resembles **traditional protest or folk poetry**.
Judges might think:
* deliberate rhetorical repetition
* socially grounded voice
* slightly rough but purposeful craft
Nothing about it screams AI.
### Poem Two
Characteristics:
* smoother, more **literary diction**
* consistent **ABAB rhyme**
* classical imagery
* elevated tone (*quiet halls of time*, *iron grip of progress*)
This resembles **modern contest-style lyric poetry**.
Judges might think:
* technically polished
* emotionally reflective
* conventional but competent
Again: **nothing uniquely AI.**
## 4. The real problem contests face now
Since tools like ChatGPT and Claude became widespread, contests face a dilemma:
1. **AI detection tools are unreliable**
2. **Poetry is short**, giving even less evidence
3. **Human revision blurs the line**
A poem could be:
* 100% human
* AI draft + heavy human editing
* human poem polished by AI
* AI poem lightly tweaked
Text alone rarely reveals which.
## 5. If these two poems were submitted
Judges would most likely:
* evaluate them **as poems only**
* not attempt to determine authorship
* possibly shortlist either one based on craft
If forced to guess about AI influence, opinions would **likely split**.
Some might guess:
* Poem Two = AI because it feels polished and conventional
* Others might guess the opposite because Poem One’s repetition looks algorithmic.
That contradiction shows the core issue.
✅ **Conclusion**
If these poems were entered in a prestigious contest today:
* **Judges could not reliably tell which was AI and which was human.**
* They would judge them purely on **poetic merit**.
* Any claim about AI origin would be **speculative at best**.
