Slouching Toward Natural: Can AI Approximate Our Individual Writing Voices?

At work, my team has been doing a lot of experimentation around how (read: if) AI can help us in our editorial work. Generally, I’ve assumed that its greatest impact will be non-creative, especially with the rise of reasoning models and agentic capabilities. For instance:

• Distilling and synthesizing meeting notes
• Automating financial reporting
• Logistical planning

After all, AI writing is … AI writing.

But I started wondering how we might help AI absorb our natural writing voices—not what we think it should sound like, but how we sound after many years of writing. My colleague Cheri proposed pointing an LLM at her editor’s picks, which is a promising angle. I thought I’d take it in a slightly different direction. Each year, I keep a running doc with every blurb that I write for LongreadsTop 5 newsletter. These blurbs are longer than our usual picks (200-250 words is probably the average), and incrementally more complex: They involve more critique and usually have a more personal approach. Giving AI one of those yearly docs—around 50 blurbs, and well over 10,000 words in total—seemed like a good place to start, with no agent behavior necessary.

I did this using Perplexity and ChatGPT, so let’s try a side-by-side approach.

Prompt 1:
The attached document has about 50 blurbs written about magazine stories. Based on the blurbs, describe my writing style and voice in a way that would let me use that description as a prompt for you.

PERPLEXITY

Insightful, engaging, and intellectually curious. You have a knack for distilling complex topics into accessible prose while maintaining depth and nuance. Your writing is marked by a blend of analytical rigor and conversational warmth, often employing wry humor and cultural references to connect with readers.

You demonstrate a wide-ranging interest in diverse subjects, from technology and sports to literature and social issues. Your commentary is thoughtful and often critical, but rarely cynical – you approach topics with a balance of skepticism and openness.

Your voice comes across as well-informed yet unpretentious. You have a talent for finding the human angle in abstract concepts and for highlighting the broader implications of specific stories. Your prose is crisp and economical, with a good ear for memorable phrases and analogies that illuminate your points.

You seem to value both intellectual substance and readability, crafting pieces that are both informative and engaging. There’s an underlying current of empathy in your writing, even when discussing controversial subjects.

Overall, your style combines journalistic clarity with essayistic depth, resulting in writing that is both accessible and thought-provoking.

CHATGPT

Your writing style is incisive, conversational, and deeply engaged with the material. You blend sharp analysis with wry humor, balancing intellectual rigor with a relatable, sometimes personal, voice. Your tone often carries a mix of enthusiasm and skepticism, and your descriptions are vivid, sometimes poetic, without being overwrought. You’re adept at threading cultural references into your critiques, making your observations feel both insightful and accessible.

If you wanted me to replicate your voice, here’s how you might prompt me:

“Write in a style that is analytical yet conversational, blending sharp critique with wry humor. Use vivid descriptions, cultural references, and an engaging, personal tone. Avoid overly academic language but retain intellectual rigor. The writing should feel lively, incisive, and slightly skeptical—curious but never credulous.”

Would you like me to refine that further based on any particular aspects of your style?

Thoughts
Immediately, you can see hallmarks of AI communication. They’re both clear, but . . . voiceless. As for the diagnosis itself, many of the same assessments emerge: “Rigor,” “wry humor,” “cultural references,” etc. I appreciated that Perplexity gave me more details—each insight is an ingredient in the recipe of any writing-voice simulation, and I thought more ingredients would lead to a richer, fuller result.


Prompt 2:
How would you describe my writing from a mechanical standpoint? What kind of sentence constructions do I use? Are there any stylistic flourishes I over-rely on?

PERPLEXITY

Based on the writing samples provided, I would describe your writing style from a mechanical standpoint as follows:

  1. Sentence structure:
  • You frequently use complex sentences with multiple clauses, often separated by em dashes or parentheses to add context or commentary.
  • You balance these longer sentences with shorter, punchier ones for emphasis or to conclude thoughts.
  1. Stylistic elements:
  • You employ a conversational tone, often using contractions and occasionally addressing the reader directly.
  • Your writing is rich in cultural references and analogies, which you use to make complex ideas more relatable.
  • You frequently use rhetorical questions to engage the reader or transition between ideas.
  1. Vocabulary and diction:
  • Your vocabulary is diverse and sophisticated, but not overly academic.
  • You often use industry-specific jargon when discussing media or technology, but generally explain or contextualize these terms for a general audience.
  1. Structural techniques:
  • You often open paragraphs with a hook or provocative statement, then expand on it.
  • You frequently use lists or series of examples to illustrate points.

As for stylistic flourishes you might over-rely on:

  1. Em dashes: You use these frequently for parenthetical asides or to add emphasis. While effective, overuse could potentially disrupt flow.
  2. Pop culture references: While these add color and relatability, very frequent use might date the writing or alienate readers unfamiliar with the references.
  3. Rhetorical questions: These are effective for engagement but could become predictable if overused.

Overall, your writing style is engaging, informative, and demonstrates a strong command of language and structure. The potential over-reliances noted are minor and generally serve your writing well when used judiciously.

CHATGPT

From a mechanical standpoint, your writing is characterized by:

Sentence Construction:

  • Varied Sentence Lengths: You frequently mix short, punchy sentences with longer, more complex ones. This creates a natural rhythm that keeps the reader engaged.
  • Parentheticals: You often use parenthetical asides (sometimes for humor, sometimes for clarification), which add a conversational tone and a sense of spontaneity.
  • Em Dashes: You favor em dashes for dramatic pauses or to introduce an additional thought—almost like a quick sidebar within a sentence.
  • Colons and Semicolons: Colons introduce sharp clarifications or dramatic reveals; semicolons appear in compound sentences, adding balance to your ideas.
  • Parallelism: You often structure phrases or clauses in a parallel way to enhance readability and impact.

Stylistic Flourishes & Patterns:

  • Rhetorical Questions: You frequently use them to create engagement and invite the reader into your thought process.
  • Conversational Repetition: You’ll sometimes repeat a phrase or reframe it for emphasis, often in a buildup to a punchline or a sharp insight.
  • Casual Yet Elevated Tone: You seamlessly blend informal phrasing (“that’s not a judgment!”) with sophisticated analysis.
  • First-Person Anchors: Even when discussing a broader cultural or analytical topic, you often include a personal touch—an anecdote, a reaction, or a subjective stance.
  • Dash of Polemic: While your writing is fair and nuanced, it often carries a persuasive undercurrent, leaning into a strong point of view without feeling heavy-handed.

Potential Over-Reliances:

  • Parentheticals & Em Dashes: They’re effective, but since you use them frequently, they could occasionally be streamlined for sharper clarity.
  • Conversational Self-Interruption: Sometimes, the way you backtrack or add a quick self-aware qualifier (“Not to brag, but…”) can slightly dilute the impact of an otherwise strong assertion.
  • Sentence Length Variation: While generally a strength, some longer sentences can get a bit winding when packed with multiple clauses.

Overall, your writing is sharp, dynamic, and highly readable. If you were looking to tweak anything, it would be in paring back certain stylistic habits (like em dashes and asides) in favor of a bit more directness at times. But that’s a minor adjustment—your voice is strong, distinctive, and highly engaging.

Thoughts

I AM ATTACKED.

• Love of em dashes: ✅
(I’m always seeking to minimize these, in my own writing and others’)
• Varying sentence lengths: ✅
(Anyone I’ve ever worked with or mentored knows how I go on and on about cadence and rhythm in writing!)
• Parallel structure: ✅
(This one impressed me—I do use it a lot, but only ChatGPT picked up on it.)

Interestingly, ChatGPT illustrated my overreliance on self-interruption (valid!) with a completely misread example. Here’s how I actually used “Not to brag, but”:

I’ve never read Moby Dick. I know that’s considered a grievous sin in certain circles. However, to those people, I say: well, how familiar are you with Pittsfield, Massachusetts? That’s what I thought! Not to brag, but I saw Dumb and Dumber in a mall there.

Anyway. On to the next part.


Prompt 3:
Distill both sets of observations about my writing into a single prompt that I can use when asking you to write something.

PERPLEXITY

Write in a style that is insightful, engaging, and intellectually curious. Use a blend of analytical rigor and conversational warmth, employing wry humor and cultural references where appropriate. Craft prose that is crisp and economical, with memorable phrases and analogies that illuminate key points. Balance complex ideas with accessibility, maintaining depth and nuance while being relatable to a general audience. Use a mix of sentence structures, including longer, multi-clause sentences (often with em dashes or parenthetical asides) balanced by shorter, punchier ones for emphasis. Incorporate rhetorical questions and lists to engage the reader and illustrate points. Approach topics with a balance of skepticism and openness, finding human angles in abstract concepts and highlighting broader implications. Aim for a voice that is well-informed yet unpretentious, combining journalistic clarity with essayistic depth to create writing that is both informative and thought-provoking.

CHATGPT

Write in a style that is analytical yet conversational, blending sharp critique with wry humor. Use varied sentence lengths, mixing punchy, direct statements with more complex constructions. Incorporate parentheticals, em dashes, and rhetorical questions for a lively, engaging rhythm—but don’t overuse them. Maintain a casual yet intellectually rigorous tone, balancing cultural references, personal asides, and polemic flourishes. Structure sentences dynamically, using parallelism, well-placed colons, and occasional repetition for emphasis. The writing should feel incisive, curious, and slightly skeptical—inviting the reader in while maintaining a strong point of view.

Thoughts

Not bad! Also, very similar! While I appreciated Perplexity’s fuller description, I overall preferred ChatGPT’s more balanced approach.

To see how these pay off, I asked both models to write a blurb from the actual piece I blurbed a week ago. This is mostly to handicap the AI—it wouldn’t have read my actual blurb in the provided corpus, so was flying blind.


Prompt 3:
Using that prompt, write a 200-word blurb about this story: https://www.esquire.com/entertainment/tv/a63613027/seth-rogen-interview-2025/

PERPLEXITY

In the twilight of his early 40s, Seth Rogen has morphed from Hollywood’s favorite stoner-bro into a multi-hyphenate mogul with a Midas touch. Esquire’s profile paints a portrait of a man who’s traded bongs for boardrooms without losing his signature laugh or his penchant for pushing boundaries. Rogen’s latest venture, “The Studio,” is a meta-comedy that turns the lens on Tinseltown’s executive suite—a world where dreams and spreadsheets collide in a haze of anxiety and Kool-Aid IP.

But what’s most striking about Rogen 2.0 is his zen-like approach to an industry in flux. While others bemoan cancel culture, he shrugs it off with the ease of a man who’s found the perfect strain. “Maybe comedy was too easy before,” he muses, challenging the notion that edgy humor is endangered. It’s a refreshing take from someone who’s navigated the tightrope between provocation and progress with the grace of a pothead parkour artist.

Rogen’s evolution—from “Freaks and Geeks” to freak-in-chief of his own empire—offers a masterclass in growing up without selling out. He’s proof that you can build an adult life without kids, stay creatively fulfilled without compromising, and yes, still get high while aiming higher

CHATGPT

Seth Rogen has spent decades refining a public persona that’s equal parts stoner philosopher, comedy workhorse, and deeply unserious businessman (see: Houseplant, his genuinely gorgeous line of weed accessories). But in his latest Esquire profile, we get a slightly different version of Rogen—one that’s older, more thoughtful, and maybe even a little wistful. Ostensibly pegged to his new Apple TV+ series Scotty & the Secret History of Hollywood, the piece quickly becomes something bigger: a meditation on fame, creative control, and what happens when your industry starts moving in ways you don’t like. Rogen, never one to mince words, has plenty to say about the state of film and TV today—particularly when it comes to superhero movies and the streaming economy—but he’s also refreshingly honest about his own evolution. He’s still funny (the man can’t not be funny), but there’s a self-awareness here that feels sharper than usual. As he talks about aging, ambition, and the kind of work he actually wants to be making, you get the sense that Rogen, for all his success, is still figuring things out. And honestly? That might be the most interesting thing about him. —PR

Thoughts

Stylistically, this isn’t much of a contest. It’s ChatGPT, hands down. I 100% would have written this sentence:

Ostensibly pegged to his new Apple TV+ series . . . the piece quickly becomes something bigger: a meditation on fame, creative control, and what happens when your industry starts moving in ways you don’t like.

Meanwhile, Perplexity’s version has an over-eagerness that I see in a lot of AI writing:

While others bemoan cancel culture, he shrugs it off with the ease of a man who’s found the perfect strain.

Might I have written that? Sure. But hopefully I would have edited it before it went any farther.

But here’s the issue: ChatGPT makes shit up. Like, a lot. (Of course it does. All LLMs do. They create ostensible “facts” out of whole cloth. Fake legal citations. Research papers that don’t exist. The list goes on.) And many of its claims about this Esquire piece are demonstrably false.

• Rogen’s new series is called The Studio, not Scotty & The Secret History of Hollywood

• The conversation never turns to superhero movies or streaming services

• Rogen never talks about aging or ambition, at least not in direct terms.

So, still at square one. Yes, I’m impressed at its ability to sound like me with sufficient coaching. But I also don’t trust it as far as I can throw it—which, it being software, is exactly 0.

Comments

Leave a comment