It was the Romantic poets, some argue, who first worshipped originality. “Poetry is the spontaneous overflow of powerful feelings,” writes William Wordsworth in his foundational text, Lyrical Ballads, “It takes its origin from emotion recollected in tranquility.” For these thinkers, poems were the finely-wrought symptoms of our passions, that stormy sea of feeling that spills into our words if only we build the proper channels. Loving spontaneity and cherishing the process of recovering feeling from within, these writers naturally feared imitation, which removed the need for both. With originality came a promise of sincerity, the arbiter of an expression’s worth. Friedrich Schlegel, the German Romantic poet, articulates this requirement in particularly harsh terms: “If lyrical poems are not completely unique, free, and true … they’re worthless.”
The demand for spontaneity persisted into the nineteenth century, taking on even higher stakes. Political economist John Stuart Mill, in 1833, described poetry as not only “the thoughts and words in which emotion spontaneously embodies itself,” but also “the exhibition of a state … of human sensibility.” While the former half of the sentence merely reiterates Wordsworth, the choice to include humanity in the latter half is interesting. Why make the distinction? Against whom, or what, does Mill stake his claim?
Perhaps in part, he wished to distinguish us from the other animal species—and indeed, the urge to prove that we are the only creatures who can express ourselves in writing continues, manifesting in cautious studies and breathless reportage. But the more threatening category, the one which most urgently needed distinction from humans, was technological. With the onset of steam powered production, we felt the existential crunch of solid, steady machinery, capable of making items en masse and with great efficiency. Fearing the corruption or depreciation of the human soul, writers and thinkers scrambled to eke out an identity for it. And so Romantic poetry became a stalwart against intellectual and spiritual death, protected by the amulet of passion, or love—that which we believe machines lack.
The idea that machines cannot feel, and thereby cannot participate in the “human sensibility” of true lyricism, is less popular today than it was a few decades ago. Back then, movies like The Terminator (1984) and Star Trek (2009) defined robots by their inability to feel, while journalism and scholarship confirmed public interest. Nowadays, we have forgotten or moved on from that central handicap in light of our robots’ immense utility. We have gone so far as to untether originality from sincerity, believing AI can simulate feeling in writing. These requests can be fulfilled by anyone with access to a large language model, but it has also crept into larger-scale practices like Hollywood screenwriting and Spotify music production. Just last year, Songs of Love, a nonprofit dedicated to creating heartfelt and personalized music for children with terminal illnesses, partnered with the AI company Suno to write and produce their songs.
The embrace of these tools in emotionally significant environments signals a shift in our understanding of the “human sensibility” that Mill spoke of. Were he and the Romantic poets mistaken about the importance of originality? What do we lose when we content ourselves with the creative output of machines, fundamentally derivative and plagiaristic, incapable of being “completely unique, free, and true”?
To answer these questions, let us compare (lowercase r) romantic writing by humans and machines. Instead of contemporary poetry, which lacks the cultural significance of its nineteenth-century counterpart, we will turn to love songs. Love songs are the best analogy to Romantic poetry, both because this music genre is immensely popular—look at any billboard and you will find few that do not fall within the category—and because its lyrics provide a shared reference point for potent feeling across generations—think Whitney Houston and Rick Astley. As for machine-made writing, I entered descriptive prompts into OpenMusic AI and Freshbots, two lyric-generating services, to create comparable equivalents.
Our first case study is Janis Ian’s 1975 ballad, “When the Party’s Over.” A longtime favorite, I find the song enchanting, its phrases earnest and delightful. Yet, Ian’s lyrics fail our standard tests for quality, and may very well be considered bad writing.
The song itself is an invitation to a courtship built on music: “I’ll teach you how to sing and dance,” Ian croons in the chorus, “with a song and dance routine. / And when the party’s over, you can fall in love with me.” Technical imperfections proliferate. Repeating “song and dance,” for instance, violates the principle of concision, a redundancy that would not pass inspection in a high school English class. Later, Ian sings, “This is my song to bring you back again,” the unsubtle admission recalling another principle of early writing education that cautions against “In this essay I will…” theses. Ian’s use of “fandango” flies in the face of George Orwell’s distaste for jargon, and the passive tense crops up on several occasions. Even the song’s central conceit makes little sense—after all, who waits until after the party to fall in love?
Yet it is precisely these technical blunders that make Ian’s ballad charming and compelling. Redundancy becomes rhythmic in a satisfying progression from “sing” to “song” to “sung”; bluntness comes off as earnestness; “fandango” couples with “tango” to produce a delightful rhyme; and the passive constructions in “The melody’s lost / before the song’s begun” lead verse smoothly into chorus. That the invitation for love is delayed until after the party’s end gives the song a simple intimacy, as song, dance, and noise fall into one pocket, and their sweet effect slots into another.
Compare this to an AI-generated equivalent, created with the following prompt: “A love song in the style of a 70s folk ballad, in which the singer invites their lover to a courtship based on music.” The machine took its instructions well enough, spitting out lyrics like, “No grand fanfare, no hurried pace / just our breaths tangled in the space / between a chord and a whispered name.” Unlike Ian’s songwriting, this passage exhibits a number of technical successes: it employs interesting imagery, figurative language, and satisfying meter towards a cohesive depiction of intimacy. Its metaphor is layered, making sense in the context of romance. By basic standards, this is a perfectly good piece of writing.
Still, there is something claustrophobic about its lyrics. Each word seems carefully calculated to maximize profundity and dramatic effect, rather than flowing from the one before it. As a result, every beat of the sentence cries out simultaneously for our attention, so that the various parts—“grand fanfare,” “breaths,” “chord,” “whispered name”—crowd one another out before they can accumulate into a beautiful whole. In short, the writing feels forced. Emotion does not “spontaneously embod[y] itself” here; rather, these “thoughts and words” reverse engineer their way to a feeling.
What makes Ian’s writing moving, while the AI’s falls flat, is its counterintuitiveness. We do not expect the repetition of “song and dance,” nor do we anticipate the admission, outright, that the song’s purpose is to bring a lover back, yet these details help our hearts understand. They make visible the mind that created them, the inelegant tremors of Ian’s “human sensibility.”
The Romantic poets seemed less concerned with the final product of their writing than with the process of producing it; effort, self-reflection, and creativity had to be mustered in the act of writing, as well as the ever-present requirement of authenticity. They wished not for cohesion and comprehensibility, but for “spontaneous overflow.” The emphasis on originality reflected a desire to witness the struggle with one’s emotions in new ways, as opposed to reaching an ultimate explanation—love, lest we forget, is extremely difficult to articulate. Bad or counterintuitive writing, therefore, is better suited to love than technically good writing, because it symptomatizes, rather than surmounts, the depth and complexity of our feelings.
For another example of human romantic expression, we can turn to Wilco’s 2001 hit, “I Am Trying to Break Your Heart.” Like Ian’s, this song contains a brusque thesis—only this time, it seems to deny love’s existence altogether, as the singer declares over and over his intention to break his lover’s heart. Still, an undeniable romance permeates the song, and the final lines give up the act: “Loves you / I’m the man who loves you.”
Wilco’s trick of negation continues a long tradition in the romantic genre, spanning from Shakespeare’s “My mistress’ eyes are nothing like the sun” to Elizabeth Bishop’s “The art of losing isn’t hard to master.” In both poems, a narrator spends the majority of their time pretending dry heartlessness, before admitting their love in the last lines. The archetype reaches its platonic form in a 1975 song by 10cc titled “I’m Not in Love,” which is made up of thinly veiled denials like “It’s just a silly phase I’m going through.” In a 2005 interview, songwriter Eric Stewart explained that the rationale behind the lyrics was to find a new way to tell his wife he loved her, so that he would not have to keep repeating, “I love you, darling, I love you, blah, blah, blah.” The contradiction of the song, its obvious lie, carries its nuance, as it does for Wilco’s “I Am Trying to Break Your Heart.”
Like bad writing, negation satisfies the Romantic poets’ desire to demonstrate, authentically, the complicated intensity of emotions. In the process of denying love, these singers reveal the extent to which they have tried to grapple with it—and, thereby, the magnitude of their feelings. By the time these songs reach their final lines, love has already started to spill over the sides.
This time, I decided to let the AI generator in on the trick, prompting, “A love song in the style of 2000s indie rock, in which the singer conveys their love by pretending not to be in love.” Despite the advantage of knowing the central nuance, the lyrics nonetheless strip it of its power. The song begins with a coffee shop scene, where the narrator’s affection is both too obvious and too hollow: “You ordered oat milk, I remembered that / from three months ago, imagine that.” Verses cycle through various situations in which the singer acts cold or disinterested and the last lines—“Just coincidence, my dear / just coincidence, I fear”—lack that vital earnestness present in Wilco, 10cc, Bishop, and Shakespeare, professing more doubt than ardor. Insofar as lies can be sincere by uncovering a deeper truth, these lyrics altogether fail to qualify.
Aside from its central conceit, Wilco’s “I Am Trying to Break Your Heart” also subverts expectations with its nonsensical lyrics. What is an “American aquarium drinker,” for instance, and how does he “assassin down the avenue”? When is the “Bible-black predawn”? How do “cross-eyed strangers” undress? These strange constructions reveal only the rough outline of people and their desires, never deciding on an exact meaning to convey. Almost synesthetically, they appear as memories and associations rather than complete thoughts, and compel us for that reason.
Other famous love songs employ the same technique. Mazzy Star’s hazy classic, “Fade Into You,” is filled with mysterious phrases like “I wanna hold the hand inside you” and “Its smiles cover your heart.” Playing upon the synesthetic effect, the Goo Goo Dolls sang in “Iris,” “All I can taste is this moment / and all I can breathe is your life.” Even The Cranberries, whose lyrics tend to be straightforward, turn toward ambiguity at a crucial moment—do we know what exactly is being made to “linger,” and was that really the rhyme we expected for “finger”?
Vague and impressionistic lyrics appeal to the Romantic taste for spontaneity, wherein each new word cannot be predicted by its predecessors. They are helpful for writing about love because the world of nonsense is much wider than the one typically available to us; one may reach out for any word that resonates with their romantic feelings, striking universality through particulars. I don’t have to generate another song to know that large language machines, with their statistically modeled text, do poorly at this task. Capable only of rote recombination, their lyricism cannot surprise us and, to our kaleidoscopic hearts, can have no appeal.
The Romanticists’ ideal of originality makes the amazing assumption that there are untapped areas of the human experience waiting to be put in words. Centuries later, their challenge still holds. The ability to articulate a single droplet from that deep and troubling well is a profound success; it is what separates good romantic writing from bad, and it provides a worthy goal for any writer or lover. Those who wish to pursue this objective need not compose elegantly, clearly, or even sensically. They may ignore conventions, lie, contradict themselves, and speak in riddles. There is, however, one requirement that most of us meet, but machines do not: to write about love, one must first have known it.
Image: Sir Edward Burne-Jones, The Love Song, oil on canvas, 1868–77 (Metropolitan Museum of Art, New York)