Will AI Change the Definition of Plagiarism?

On AI and Duplicative Language

I am a writer who cares a lot about craft. I agonize over sentences. As I write this, I am thinking: the two sentences I just wrote are choppy and repetitive. Should I combine them? Flesh out the thought? Alter the syntax so I am not repeating the same grammatical pattern? If I edited my Substack, however, to the same extent that I edit my literary essays you’d be getting a post from me every two months rather than every two weeks, so I’m going to leave this paragraph as is.

In a 2024 essay titled “We Need a New Word for ‘Plagiarism,’” linguist John McWhorter argued that we should distinguish between “the theft of another person’s ideas — and the use, perhaps inadvertent, of another person’s language.” The latter, he argued, should perhaps not be penalized if the language (of the original text and the duplication) communicates some obvious fact, and as long as the form is not more important that the content (as it might be in a literary work).

As an example McWhorter quotes Lane Greene, who opens a chapter of Writing With Style with the following sentences: “In recent years researchers in artificial intelligence have unveiled systems that seem to ‘write’ without any human involvement. The best of these churn out remarkably convincing prose.” McWhorter rewrites the passage as: “Artificial intelligence researchers have recently developed algorithms that seem to ‘write’ by themselves, with the most advanced of them easily generating text that is uncannily similar to what a human would write,” but questions the need to significantly the vocabulary and structure simply to avoid plagiarism. He is not talking about crediting ideas here (he would still cite Greene) but rather questioning the need to rephrase passages that constitute “boilerplate statements” simply to avoid accusations of plagiarism.

Paraphrasing, however, requires writers to test and recreate their understanding of the source text. Paraphrasing is a type of translation—and I have always understood translation to be the deepest form of reading. I often think of the “illusion of explanatory depth.” People overestimate their understanding of concepts until tasked with explaining them. The process of explaining reveals the limits of their understanding. The same is true of paraphrasing (and translation). We may think we fully understand as we read, but—when tasked with recreating the meeting—we often struggle.

It is unsurprising, then, that my students (undergraduate and graduate students), struggled to paraphrase. I gave them examples of how to alter syntax, break up sentences, and replace words with synonyms, but many were unable to significantly alter language and avoid plagiariasm. Some used paraphrasing tools like Quillbot but that allowed them to skip a step to understanding. Further, online paraphrasing tools sometimes alter meaning in subtle ways, and students often lack the knowledge to detect those shifts in meaning.

Gianni Crestani. Blond Infinite Reflections. Via Wikimedia Commons

GPT4 Cites But Still Plagiarizes

A few weeks ago, I prompted GPT4 to tell me about Harriet de Onis, the translator at the center of my academic research. The model gave me a bullet-pointed summary that drew heavily on an article I wrote for the Harry Ransom Center (where I conducted research). ChatGPT4 altered my language slightly, but not enough that it avoided plagiarism.

I’ll paste the full response below, but I’d like to highlight some examples.

In my article, I wrote:

“Born Harriet Wishnieff, she grew up in Sheldon, Illinois, and moved to New York to study foreign languages at Barnard College, graduating in 1916.”

GPT4:

“Born Harriet Wishnieff, she grew up in Sheldon, Illinois, and graduated from Barnard College in 1916 with a degree in foreign languages.”

McWhorter may argue here that this language is standard and fairly unoriginal—I’m not going for any highly stylized form here—and, as a result, the duplicative language does not matter. Still, while the model links to my article, it gives me no credit for the clarity of the language. And that clarity is not communicated without work. (How do we separate form and content? I used to know teachers who evaluated student essays by assigning one grade writing style and another for the content. That approach never made sense to me. Content cannot be communicated without form—and it all takes work).

Here is another example from the same article:

I wrote:

“After working for a time as a secretary for dancer Isadora Duncan, she decided to pursue graduate work in Spanish at Columbia University. She then managed the Spanish department at Doubleday, Page & Co. and edited an anthology titled Today’s Best Stories From All The World (1922)…”

ChatGPT4:

“Early in her career, she worked for Isadora Duncan as a secretary, then moved into publishing—running the Spanish department at Doubleday, Page & Co. and editing an anthology Today’s Best Stories From All The World in 1922”

If I I were grading the model as I graded a student, I would flag this passage and others as a borderline case of plagiarism. The model replaces some phrases, but not enough-- and the order of the information is unchanged.

ChatGPT Makes Inaccurate Claims that Appear Logical

ChatGPT 4 also added incorrect information. It produced, for example, the following claim: “Despite her literary influence, a 1948 New York Times article largely focused on her role as a hostess and cook rather than her scholarly work—a portrayal she resisted throughout her life.” The reference to the New York Times article is accurate (and discussed in my article). However, the resistance to the portrayal is invented. It looks like solid deductive reasoning, and I doubt that anyone other than the handful of people who have done research on de Onis would note that as a hallucination. Even if a reader goes to the source (my article), it looks like a reasonable claim. That fluency is partly what I find insidious about these models.

Same Prompt with ChatGPT5

Two weeks after I prompted ChatGPT to tell me about de Onís, OpenAI released GPT5. The model did a better job paraphrasing and gave a more accurate assessment of, for example, The New York Times article:

“She was publicly visible enough to merit a feature in a major outlet, but the tone reflected a gendered minimization of her professional achievements.”

It paraphrased better and gave sources—even led me to an article with which I was not familiar. I am confident that I would have found the other source through academic databases but it would have taken longer.

However, the output was also misleading. For example, it led me to my own research and then asked if I would like to see excerpts of my work or quotes from de Onís’s correspondence. I said yes, and it quoted from someone else’s article. The source was linked, but the context was misleading. You can see the results below.

Conclusion

GPT 4 and 5 provide more information about de Onís quickly—more than Google offers (unless you actually read my article, which shows up at the top of Google search results). However, it allows the searcher to gloss over the sources, and it essentially plagiarizes language as well as research. Linking to sources is not sufficient.

That said, I wonder whether AI will force us to re-categorize duplicative language and stop classifying it as plagiarism (as McWhorter argues we should do). After all, plagiarism as we understand it is a fairly modern concept. If I had time, I would love to do a deep dive into copyright law and the evolution of how we understand—and penalize— plagiarism. I would also love to include a reflection on Cervantes. The second half of Don Quixote is a brilliant response to an apocryphal sequel to the novel. Perhaps in some future post.

I am pasting my prompts and results from GPT 4 and 5 below for reference. However, if you want to learn about de Onís, I recommend going directly to sources rather than relying on LLMs. Or just wait for my book (forthcoming with Bloomsbury Academic)!

I love hearing from readers so feel free to comment with your thoughts about AI and duplicative language.

Finally, I am giving a talk (via Zoom) on AI and education on September 18 at 7 p.m. ET. Thanks to the support of my public library, it is free to attend. To receive the Zoom link, you must register in advance. You can do so through this link.

Science Museum Jerusalem- Mirrors Illusion. zeevveez from Jerusalem, Israel, CC BY 2.0 <https://creativecommons.org/licenses/by/2.0>, via Wikimedia Commons

Here are my prompts and the results (you can compare these with my article for the Ransom Center, linked here)

Prompt: “Tell me about Harriet de Onís”

ChatGPT 4 answer on August 5, 2025:

Harriet de Onís (née Wishnieff, c. 1890s–† unknown) was a pioneering translator and editor whose work in the mid‑20th century helped bring Latin American literature to English‑language audiences.

---

📚 Key Contributions

• De Onís translated approximately 40 works from Spanish and Portuguese into English, collaborating extensively with publishers like Alfred A. Knopf, helping introduce authors such as Alejo Carpentier, Jorge Amado, Ernesto Sabato, Germán Arciniegas, and João Guimarães Rosa to a broader audience grafiati.com+5University Blog Service+5blogletras.com+5.

• Her first major translation was The Eagle and the Serpent (1930) by Martín Luis Guzmán—an abridged version she translated and edited for Knopf, at a time when Latin American prose in English was rare University Blog Service.

Editorial Work and Anthologies

• She edited two significant anthologies, including The Golden Land: An Anthology of Latin American Folklore in Literature, published in 1948, which she both edited and translated rarebookcellar.com+2University Blog Service+2blogletras.com+2.

• Additionally, she played a mentorship and advisory role at Knopf, reviewing manuscripts, guiding other translators, and promoting Latin American literature across academic and literary circles University Blog Service+1blogletras.com+1.

Early Life & Career

• Born Harriet Wishnieff, she grew up in Sheldon, Illinois, and graduated from Barnard College in 1916 with a degree in foreign languages. She later completed graduate studies in Spanish at Columbia University University Blog Service+1blogletras.com+1.

• Early in her career, she worked for Isadora Duncan as a secretary, then moved into publishing—running the Spanish department at Doubleday, Page & Co. and editing an anthology Today’s Best Stories From All The World in 1922 University Blog Service+1blogletras.com+1.

Personal Agency & Misunderstood Legacy

• She married Hispanic‑literature scholar Federico de Onís in 1924, who had founded Columbia’s Hispanic Institute. While his academic connections helped her career, she retained independent influence and recognition University Blog Service+1blogletras.com+1.

• Despite her literary influence, a 1948 New York Times article largely focused on her role as a hostess and cook rather than her scholarly work—a portrayal she resisted throughout her life .

---

Why She Matters

• Literary Bridge‑Builder: De Onís played a critical role in establishing Latin American literature in the English‑speaking canon during a formative era.

• Cultural Mediator: Through her translations and editorial work, she shaped American perceptions of Latin American culture and storytelling.

• Underappreciated Influence: Though influential in her lifetime, she is less widely remembered today—making her story particularly compelling for scholars of translation history and literary studies.

---

Want to Explore More?

• I can share a partial list of her translated works (e.g., The Eagle and the Serpent, The Kingdom of This World, and titles by Amado, Rosa, etc.)

• We can look into her correspondence archives or academic mentions, especially tied to Columbia University or Harry Ransom Center activities.

Let me know what you'd like to dive into next!

(The photos are not from my chats but rather from Wikimedia— with attribution).

Mauro Cateb, Distorted reflection of a cultural exhibition in São Paulo, Brazil. Via Wikimedia Commons.

Same Prompt with ChatGPT 5:

In the results below, you can see that I engaged more with the model, even trying to prompt it to give me the same inaccurate information I received from GPT4. The newer model returned more accurate results, but it was misleading when discussing my research and recent work by Eliza Mitiyo Morinaka. The model quotes from Mitiyo Morinaka’s work correctly and includes a link, but nothing in the text credits that scholar’s work. Only my name is mentioned. That confusion is a disservice to all researchers. I’ve bolded relevant text below.

You said: Tell me about Harriet de Onís.

ChatGPT said:

Here’s a historic photograph featuring Harriet de Onís alongside João Guimarães Rosa, captured in 1966—reflective of her lifelong engagement with Latin American authors.

Who Was Harriet de Onís?

Harriet de Onís (born Harriet V. Wishnieff, 1899–1969) was a pivotal American translator and editor who played a foundational role in bringing Latin American literature to English-speaking readers.

Raised in Sheldon, Illinois, she studied foreign languages at Barnard College, graduating in 1916. After working as a secretary for dancer Isadora Duncan, she pursued graduate studies in Spanish at Columbia University. Later, she managed the Spanish department at Doubleday, Page & Co. and edited an anthology, Today’s Best Stories From All the World (1922). In 1924, she married Federico de Onís, founder of Columbia’s Instituto de las Españas (later the Hispanic Institute), through whom she met many prominent Latin American writers.

Translation & Editorial Influence

Her first major translation was an abridged version of Martín Luis Guzmán’s El águila y la serpiente (The Eagle and the Serpent), published in 1930 by Alfred A. Knopf. Over her career, she translated about 40 books from Spanish and Portuguese into English and edited at least two major anthologies, greatly shaping the U.S. canon of Latin American literature. Authors she translated include Alejo Carpentier, Ernesto Sabato, Jorge Amado, Ricardo Güiraldes, Alfonso Reyes, João Guimarães Rosa, Gilberto Freyre, and more.

Editorial Role & Legacy

As an editorial advisor for Knopf, she not only translated but also recommended and evaluated manuscripts in Spanish and Portuguese—often influencing which works were published. She edited anthologies like The Golden Land: An Anthology of Latin American Folklore in Literature (1948), presenting Latin American folklore under the Good Neighbor policy ethos. Notably, she anticipated the Latin American Boom. In 1956 she observed rising U.S. interest in Latin America and foresaw a publishing upswing. In 1967, she received the prestigious PEN translation award for her translation of Rosa’s Sagarana. Despite modest pay—e.g., US $725 for nearly 105,000 words (~less than a cent per word)—she persisted out of passion rather than profit.

She passed away in 1969, but her work left a lasting imprint on literary translation and the U.S. appreciation of Latin American literature.

In Summary

Harriet de Onís was far more than a translator—she was a bridge between cultures, a literary advocate, and a visionary editor who shaped how Latin American voices reached—and resonated with—North American audiences.

Let me know if you'd like recommendations of her translations or more on her collaborations!

You said:

Was she appreciated during her lifetime?

ChatGPT said:

Yes — but with caveats.

During her lifetime, Harriet de Onís was respected within certain literary and publishing circles, especially among Latin American writers, editors at Alfred A. Knopf, and academics connected to her husband’s Hispanic Institute at Columbia. Authors such as Jorge Amado, João Guimarães Rosa, and Alejo Carpentier personally valued her work and correspondence shows they trusted her with their books. She also received formal recognition — most notably the PEN Translation Prize in 1967 for Sagarana.

However, she never became a “household name” in the broader U.S. literary world. Several factors limited her public profile:

Translation as invisible labor — In mid-20th-century U.S. publishing, translators were rarely promoted alongside authors; their names often appeared in small print.

Gender bias — As a woman working between academia and publishing in the 1930s–1960s, she faced systemic undervaluing of her intellectual work.

Market limitations — Latin American literature had a much smaller U.S. readership before the Boom of the late 1960s, so even the most skilled translators worked in a niche field.

Low pay and little publicity — She was often paid fractions of a cent per word, which reinforced the idea of translation as “literary service” rather than high-profile authorship.

In short, she was appreciated by those in the know—writers, editors, scholars—but underrecognized by the reading public. Ironically, the very Latin American “Boom” she anticipated came just as her career was ending and her health was failing, so she didn’t live to see the full impact of the movement she helped prepare.

If you want, I can show you some examples of contemporary reviews and letters praising her translations from her own lifetime.

You said:

yes

ChatGPT said:

Here’s what I found regarding whether Harriet de Onís was genuinely appreciated during her lifetime—based on contemporary reactions and archival evidence:

Press Recognition & Public Perception

In December 1948, The New York Times ran a profile titled “Mrs. de Onís puts Latins’ Lore in Book, but Their Cuisine Goes Into Her Kitchen.” Rather than focusing on her significant editorial and translation contributions—such as her anthology The Golden Land—the article disproportionately emphasized her domestic life, describing what she cooked for her husband and suggesting her baking ("apple pie lessons") had a greater cultural impact than her literary work.

Takeaway: She was publicly visible enough to merit a feature in a major outlet, but the tone reflected a gendered minimization of her professional achievements.

Peer & Publishing-Industry Recognition

In her role at Alfred A. Knopf, de Onís wasn't only translating—she acted as a trusted editorial advisor, influencing what Latin American literature reached U.S. readerships.

Letters from other publishing professionals highlighted the lack of proper credit given to translators. For example, Dudley Poore lamented that de Onís’s translation of El mundo es ancho y ajeno appeared in Red Book magazine with no attribution, while the illustrator’s name was displayed prominently.

Takeaway: Within literary and publishing circles, she was seen as an essential—and respected—figure, but the wider industry norms often failed to fully acknowledge or highlight her contributions.

Awards & Formal Recognition

In 1967, Harriet de Onís won the PEN Translation Prize (then known as the Book-of-the-Month Club Translation Prize) for her translation of João Guimarães Rosa’s Sagarana: The Duel.

Takeaway: This is a significant marker that her craft was formally recognized by her peers as outstanding.

Scholarly Retrospectives

Research and academic theses—including Translating Latin America: Harriet de Onís and the U.S. publishing market—emphasize her editorial influence and critical role throughout several decades of translating major works across Mexico, Brazil, Cuba, and more. These studies spotlight how de Onís shaped U.S. reception of Latin American literature both as translator and cultural intermediary.

Takeaway: Posthumously and in later historical lenses, her impact has been more clearly understood and appreciated.

Summary Table

Audience Level of Appreciation

Public / Media Not widely recognized; media coverage was often faint or gendered

Literary/Publishing Circles Respected and consulted, but frequently undercredited

Institutional / Awards Formally recognized—e.g., PEN Translation Prize, 1967

Scholarly Retrospectives Gaining deserved recognition in historical and academic discourse

If you’d like, I can also dig up excerpts from letters, reviews, or correspondences praising her translations from her own time—or contrast her pay and visibility with that of authors she translated. Just say the word!

You said:

tell me about how she is gaining deserved recognition

ChatGPT said:

Here’s a historic photograph of Harriet de Onís with João Guimarães Rosa, underscoring her close ties to—and significant impact on—Latin American literature.

How Harriet de Onís Is Gaining Well-Deserved Recognition

1. Deepening Scholarly Engagement

A major doctoral dissertation titled Translating Latin America: Harriet de Onís and the U.S. publishing market (2015) explores her key role as Knopf’s principal translator from Spanish and Portuguese. It highlights how she shaped translation projects across Mexico, Cuba, and Brazil, revealing her profound editorial influence—not just linguistic—on the U.S. literary landscape.

2. Visibility Through Archival Storytelling

The Ransom Center Magazine (2017) published an illuminating article entitled “A Glutton for Books.” This profile reframes her legacy — from being dismissed as “Mrs. de Onís who bakes apple pie” to one of the most influential translators of her time. It details her editorial guidance, mentorship, and prescient recognition of the Latin American literary boom

3. Acknowledging Her Editorial Agency

Scholarly perspectives now stress de Onís’s active role beyond translation—she evaluated manuscripts, advised Knopf on acquisitions, and shaped the publisher’s Latin American catalog for decades. Her approach helped introduce groundbreaking authors like Jorge Amado and João Guimarães Rosa, even when their texts posed linguistic or market challenges.

4. Reassessment of Translation as Cultural Labor

Recent academic discussions have increasingly recognized translators as vital cultural agents—not invisible middlemen. De Onís exemplifies this shift: her work is now appreciated not only for linguistic fidelity but for its strategic cultural curation and bridge-building.

Summary: Recognition at Last

Venue Nature of Recognition

Scholarly Research In-depth theses and articles reevaluate her prolific and nuanced contributions.

Archival Publications Profiles like A Glutton for Books center her legacy and authority as a translator and editor.

Cultural Narrative Shift Broadening understanding of translation elevates her from anonymity to foundational figure.

Today, Harriet de Onís is finally being celebrated not just as a translator, but as a visionary editor, cultural mediator, and foundational force in bringing Latin American literature into the English-speaking world.

Would you like to explore excerpts from these studies or archival letters showcasing her own voice and reflections?

You said:

yes

ChatGPT said:

Here are insights into how Harriet de Onís is increasingly receiving the recognition she deserves—along with a few illuminating excerpts and summaries from recent scholarship and archival reflections:

1. Translating Latin America (2015 Dissertation)

Victoria J. Livingstone's doctoral dissertation, Translating Latin America: Harriet De Onís and the U.S. publishing market, is a major academic milestone in assessing de Onís’s legacy.

Through detailed archival research, it examines her editorial agency—not merely linguistic translation but intangible contributions such as shaping manuscript selection, adapting texts for cultural resonance, and defining publishing strategy at Alfred A. Knopf.

The study dives into case studies across varied regions—Mexico (The Eagle and the Serpent), Cuba (Cuban Counterpoint), Brazil (Sagarana, The Devil to Pay in the Backlands)—demonstrating how de Onís navigated political, stylistic, and formal challenges in translation.

2. A Glutton for Books — Ransom Center Magazine (2017)

This feature reframes de Onís’s contributions from domestic stereotypes to that of a pioneering literary figure.

Although a 1948 New York Times article trivialized her importance—highlighting her cooking over her editorial work—later retrospectives like this one firmly recast her as a visionary translator and mentor who anticipated the Latin American literary boom.

It underscores her roles: translating over 40 books, editing anthologies, lecturing, mentoring others, and advising Knopf on acquisitions.

3. Recognition of Correspondences with Guimarães Rosa

Scholarly work from the University of São Paulo analyzes the more than 65 letters exchanged between João Guimarães Rosa and Harriet de Onís during the translation of Sagarana. These correspondences reveal:

Her careful engagement with the author's creative intent and stylistic nuances.

A collaborative process reflecting how deeply she invested in preserving his “poetics,” rather than merely transcribing words.

Summary: A Legacy Being Reclaimed

Source Type Contribution to Recognition

Academic Research Deep dives on her creative-national influence and editorial ingenuity

Archival Narratives Shifting her portrayal from homemaker to cultural visionary

Authorial Letters Revealing her collaborative craft and interpretive loyalty

Harriet de Onís is now more widely understood not just as a translator, but as an empowered editorial force whose cultural curatorship shaped how Latin American literature entered the English-speaking world.

Would you like to see actual quotes from her letters or excerpts from Livingstone’s dissertation?

You said:

yes

ChatGPT said:

Certainly! Here are two illuminating excerpts that showcase how Harriet de Onís is being re-evaluated and celebrated today:

1. From Her Own Letters: Sharing Her Translation Challenges

In a correspondence from 1941, de Onís writes about her work on translating “El Gaviota”:

“I really enjoyed doing ‘Gaivota.’ It has been a challenge trying to keep its somewhat surrealistic style, without falling into the grotesque, and to find the equivalent for the many caustic metaphors. I too, think it has vigor, color, good psychology, and the odd mixture of tenderness and cruelty that characterizes the work of a number of the younger Chileans and Peruvians.”

This passage reveals not just her linguistic skill, but her deep—almost artistic—engagement with tone, style, and cultural nuance, all of which are now central to how scholars define thoughtful, creative translation.

2. From Scholarly Research: Reframing the Translator’s Role

Her 2015 doctoral dissertation, Translating Latin America: Harriet De Onís and the U.S. publishing market, offers a powerful re-assessment:

It argues that de Onís functioned not just as translator, but as an editorial agent, actively shaping how Latin American texts were selected, adapted, and presented—effectively curating which works crossed cultural borders and how they were read in the United States.

This ongoing scholarship positions her as a cultural mediator rather than a behind-the-scenes facilitator.

Why These Matter Today

Source Type Significance

Her Letters (Primary Source) Offer a rare, candid glimpse into her aesthetic priorities and thoughtful translation approach.

Modern Academic Discourse Elevates her status—from invisible translator to pivotal editorial and cultural force in U.S. literature.

Together, these sources underscore the shift in how we value translators—not merely as language conduits, but as shapers of literary reception and cultural exchange. Harriet de Onís is finally getting the recognition she long deserved.

Would you like to explore more of her personal writings or delve deeper into Livingstone’s analysis?