■ SAFE RISK ■ Creative Arts
No, though not because AI can't produce competent verse — it can, by the tanker-load. Poetry's economy runs on the reader's belief that a specific human meant it, and that's the one ingredient a model cannot supply.
“AI writes poems that rhyme. Poets write poems that mean something. Sometimes.”
Our AI replacement risk score — how we score jobs
Nobody is a poet the way one is an accountant. The working life is a patchwork: writing (unpaid), submitting to journals (often paying reading fees for rejection), readings and slams, teaching workshops or MFA courses, applying for grants and residencies, occasional commissions — weddings, memorials, corporate anniversaries in a brave mood. The median poet's poetry income would embarrass a lemonade stand; the job survives on teaching, day jobs, and stubbornness. This matters for the AI question: you can't automate away an income that mostly wasn't there.
Language models write verse fluently — meter, rhyme, imitations of Dickinson or Neruda on demand. Studies have found casual readers sometimes rate AI poems as highly as human ones, a result that says as much about how people read as about the machines. The commercial fringes are genuinely exposed: greeting-card verse, song-lyric drafts, Instagram-caption poetry, and content-mill inspirational quotes are already semi-automated. Literary journals now wrestle with floods of AI-generated submissions, which mostly punishes editors, not poets.
But poetry's core transaction resists in a way that's almost structural. A poem is read as testimony — this person saw this, felt this, chose these words. Strip the author and you strip the meaning; an elegy generated by a server mourns no one. The live scene (slams, readings, open mics) is flourishing and is inherently embodied. If anything, AI's fluency is clarifying poetry's value proposition: when competent language is free, witnessed language is what's left to sell.
Automatability: our editorial assessment of current and near-term AI capability
Resilient for the foreseeable future, partly because there was never much of an economy to disrupt. The commercial fringe — greeting cards, caption verse, lyric drafting — is automating now. Journal slush piles are drowning in AI submissions this decade, changing how editors work. But poetry as a human practice, and the teaching-performing-grant economy around it, shows no sign of displacement through the 2030s and beyond.
It writes technically competent poetry — correct meter, serviceable imagery, convincing pastiche — and some studies show casual readers can't reliably tell the difference. What it can't do is mean it. Poetry is read as one human's testimony, and a poem with no one behind it is a costume with no body. Experienced readers and editors still tend to notice the hollowness; casual scrollers often don't.
The joke answers itself — what income? Poetry has almost never paid through poems; it pays through teaching, performances, grants, and commissions. Of those, only commercial verse (cards, captions, lyric drafts) is being automated. Teaching and live performance are stable or growing. The real new problem is AI-flooded submission piles making it harder for unknown poets to get read by journals.
If you're pursuing it for money, you were already in trouble, and AI didn't cause that. If you're pursuing it as a vocation with a sustainable day job or teaching path, nothing about the current moment stops you — and the live scene is arguably healthier than it's been in decades. Build a body of work, a community, and a public voice; those compound and can't be generated.
Painfully. Several well-known magazines have temporarily closed submissions after floods of machine-generated work, and most now publish explicit AI policies. Expect more verification: cover letters about process, relationships with editors mattering more, and prizes requiring declarations. For genuine poets, the practical fallout is slower response times and a rising premium on being a known, trusted name — which makes community involvement more valuable, not less.