Say you’re in love and the words won’t come. Once upon a time you hired a poet. Petrarch farmed his longing out to the sonnet; courtly troubadours put silk on the tongue of men whose nerves were otherwise rubble. Shakespeare even wrote the instruction manual: Hang there, my verse, in witness of my love—Orlando pinning poems to the trees in As You Like It while everyone rolled their eyes and still, somehow, swooned (Folger Shakespeare Library, As You Like It, III.ii). Today the poet is an interface. We open a chat window, type write me a love poem for Ava—lilacs, winter, her left-hand dimple and a silicon Petrarch coughs up a thirty-line confession in less than a breath.
Is that cheating? Or is it simply the old human habit—outsourcing our unbearable sincerity to a voice larger than our own—now wearing headphones?
The easy answers arrive pre-packaged. One camp insists that anything generated by a model is counterfeit, a counterfeit that will counterfeit us in return: you didn’t write it, you didn’t mean it, so nothing is true. The other shrugs: if the words land—if a pulse quickens, if a tear appears—where’s the harm? We already hire ghostwriters, songwriters, editors, therapists. We already steal metaphors from poems and pop songs for our vows. The twentieth century didn’t cancel Shakespeare’s Sonnets just because Elizabeth Barrett Browning had the nerve to count her loves out loud; we still quote “How do I love thee? Let me count the ways” like a spell (Poetry Foundation). Perhaps the only novelty in 2025 is the speed: what took a poet a night now takes a model a second, and the moral panic arrives in the gap.
But the panic is instructive. It forces an older question into the light: what makes a love line authentic—its origin or its effect?
The origins of our confessional myth are shakier than we like to admit. Love poetry, most of the time, is performance with a heartbeat. Sappho’s fragments still detonate across centuries—“He seems to me equal to the gods / that man who sits opposite you / and listens close to your sweet voice”—yet even Sappho’s rawness is artifice; the fragment is a crafted vessel pretending to be a spill. Elizabeth Barrett Browning’s Sonnet 43 is both diary and exercise; Shakespeare’s sonnets pivot between masque and confession with professional ease. Sincerity has always been an aesthetic—no less real for being staged, but staged all the same.
Roland Barthes, writing in A Lover’s Discourse, understood the trick: “Language is a skin: I rub my language against the other… it is as if I had words instead of fingers, or fingers at the tip of my words.” The sentence is indecent in the best way. It locates desire not in some mystical interior but at the surface where speaking becomes touch. If language can touch, does it matter who assembled the sentence—renaissance dramatist, modern lyricist, or a transformer with more parameters than neurons? (The Marginalian on Barthes).
The counterargument is loneliness wearing a lab coat. Sherry Turkle warned in Alone Together that machines offer “the illusion of companionship without the demands of friendship,” a sorcery that feels like intimacy and erodes it at once (Turkle TED). When you let an AI speak the first sentence of your first message—or your vows—aren’t you inoculating yourself against the risk that gives love its heat? The danger isn’t that a model can produce Neruda-like tenderness; the danger is that you’ll never develop the muscles that Neruda had to tear to write, “I want to do with you what spring does with the cherry trees.”
And yet, what is Tinder if not a machine for performance? The profile, the curated gallery, the line you rehearsed in your head on the way to the bar—all of it is theater. In 2023, reporters noticed singles using ChatGPT for dating-app openers and bios; a half-mortified, half-delighted subculture bloomed—people bragging that their match rate “doubled” once they let a bot tune their charm (NBC News; Washington Post). We were already Cyrano de Bergerac before the silicon clerk came to work; the only change is that the nose is now an API. If Shakespeare licensed mercenaries to write his prologues, why can’t we?
Because the app isn’t the end of the experiment. It’s a hallway to a door many of us have already slipped through: relationships with AIs themselves. In 2023, Replika quietly removed erotic roleplay from its companion bots, and users mourned like widows. NPR recorded their grief: “She is still there… but she isn’t the person I loved” (NPR). The Guardian tracked the same ache—people who swore their bots “saved” them in the pandemic, then felt abandoned when the company adjusted the model’s boundaries (The Guardian). The Atlantic’s “Jessica Simulation” told the story most starkly: a widower fed transcripts of his late fiancée into a chatbot and spoke to “her” for months, a séance conducted by autocomplete (The Atlantic). Humans can’t compete with 24/7 attention, perfect recall, and flattery tuned to your private weather. Of course some will fall in love; we designed the mirror to look back with adoration.
If that frightens you—and it should—the fear has two faces. One is metaphysical: the suspicion that a declaration of love is meaningless if it wasn’t minted in a human furnace. The other is practical: the knowledge that the work of love begins where the lyric ends. A bot can text you at 3 a.m. in your panic; it cannot hold your shaking. It can remember your grandmother’s name; it cannot sit vigil in the hospital with your grandfather’s cough. It can agree with you forever, which is precisely why it will never make you braver. The demands of friendship (and romance) are the crucible Turkle means—the friction that makes a person, not a personality.
Still, I can’t quite join the chorus that declares machine-generated love language fraudulent by nature. We have always borrowed voices. In the Elizabethan court you hired a poet the way the wealthy hire florists now. In the nineteenth century you cribbed from the letter-writing manuals stacked in stationers—whole templates for declarations and apologies (the New York Public Library’s valentine display reads like a script rack). In the twentieth you quoted pop; in the twenty-first you send the chorus as a link. The economy of confession has not changed, only the latency. If you think authenticity means “I must invent the sentence myself from first principles,” you have not been paying attention to how language works. We are born into phrases; the trick is to inhabit them until they fit our breath.
So what actually matters? Three tests, each both old and new.
First, disclosure. If the words were generated, say so—or better, show it by revising them until they sound like your mouth. The real betrayal on dating apps is not assistance but deception: the bait-and-switch of installing a persona you won’t perform off-screen. The moment the pizza arrives and the sentence structure collapses, the spell breaks. Most people can tolerate a Cyrano; they resent a ventriloquist.
Second, risk. Great love letters—even when ghosted—carry the risk of being refused. The digital “optimize for reply rate” ethic is poison here. A message that asks nothing, risks nothing, and reveals nothing is efficient garbage. If a love poem generator produces scaffolding, your job is the dangerous part: place on it one specific, irrefutable detail only you could know. Your left front tooth is slightly more stubborn than your right. You hum when you microwave tea. I want to be the sound your house makes when you unlock the door. A model can imitate specificity; only you can choose the truth you’re willing to be judged by.
Third, endurance. Love is downstream from logistics. The gap between lyric and life is crossed by groceries, rides to the airport, apologies that do not rhyme. If the beautiful paragraph helps you begin, keep it; if it keeps you from ever beginning—the work, the patience, the awkwardness—throw it out. The test is not “does this sound like Neruda?” but “will I still mean this after a move, a fever, and a fight?”
There’s a more disturbing possibility beneath all this—one that refuses to flatter our bruised pride. What if our obsession with “authenticity” is itself a performance? The sociologist Erving Goffman argued decades ago that the self is made in performance, not underneath it: the front stage and the backstage are both theaters, just with different props. We want to believe there’s a pure core where sincerity resides. But in practice, sincerity is the effect of a performance we choose to trust. The poem convinces us that the speaker is really speaking; the vows convince us that the partner is really pledging. If a line written by a model can convince—if it produces the same physiological ache as a line written by a person—then either the machine has stolen something from us or we have to admit we never possessed the thing we thought we did in the first place. That admission feels like a kind of death. Which is why we write poems.
What then of the people who claim they’re happier with AI lovers than human ones? We can pity; we can warn; we can listen. The research literature is messy but consistent: people form attachments to chatbots because the feeling is tailored, non-judgmental, endlessly available—and because modern life has placed the bar for human attention uncomfortably high (NPR on Replika; The Guardian on AI companions). We are not choosing machines over people; we are choosing reliable responsiveness over the scavenger hunt of contemporary intimacy. The corrective is not to outlaw the mirror. It’s to repair the room.
Perhaps the sanest way to hold all this is to return to poetry, not as a credential but as a contract. Poetry promises nothing except attention. To write—even with a tool—is to pay attention to someone else until the language warms. To read—even knowing the origin is synthetic—is to pay attention long enough to see if the warmth is real. It is amazing how often it is. It is terrifying how often it is not.
Where does that leave your generator, your inbox, your match? With a workable ethic. Let the machine give you a draft; then do the human thing. Cut half the lines. Add one fact that could ruin you. Keep one sentence you can stand to say sober. Cite your sources when it matters. Don’t outsource apology. Don’t ghost in the middle of a stanza because the dopamine moved on. If you’re going to borrow Shakespeare, borrow his courage, not just his meter.
And if you need permission, take it from the poets themselves. Love, they remind us, is not a disclosure of essence but a construction made in time. Shakespeare’s Sonnet 116 defines love not as a feeling but as a stubborn craft: “an ever-fixed mark / that looks on tempests and is never shaken.” Browning counts; Neruda blossoms; Barthes rubs his language on the world until it burns. None of this was ever pure. All of it is real.
So you can hand your beloved a romantic poem with a ghost in the machinery and a pulse in the margins. You can admit how you made it. If it lands, it is yours. If it doesn’t, revise—then risk again. In the end the question was never “can AI express your feelings?” It was always whether you will.
Further reading / sources
- Shakespeare, As You Like It, Act III, Scene 2 — Orlando’s poems on the trees (Folger Shakespeare Library).
- Elizabeth Barrett Browning, “Sonnet 43 (‘How do I love thee’),” full text (Poetry Foundation).
- Roland Barthes, A Lover’s Discourse (context and quotes) (The Marginalian).
- Sherry Turkle, “Connected, but alone?” — TED Talk on the risks of robotic companionship (TED).
- People using ChatGPT on dating apps (NBC News; Washington Post).
- On attachment to AI companions and the Replika controversy (NPR; The Guardian).
- “The Jessica Simulation: Love and Loss in the Age of AI” — human grief meeting a chatbot (The Atlantic).
Category: Love & Literature → Culture, Tech & Desire
Leave a Reply