We Built a Writing Profile for Every AI Model — Here's What We Found
We profiled 5 AI models like human writers. Claude, GPT, and Gemini each have a measurable writing personality — and none of them match yours.
We spend most of our time building writing profiles for humans — analyzing your emails, reports, and messages to capture what makes your communication uniquely yours.
But we got curious: what happens when you point the same analysis at the AI models themselves?
We took our stylometric framework — the same six-axis measurement system behind every Writing DNA Snapshot — and applied it to 320 samples generated by five major AI models. Then we did something we normally do for human writers: we gave each model a personality label based on its data.
The results are revealing. And shareable.
The Methodology
Same setup as our broader benchmark study: five models, eight prompt types, four languages, two variants per combination, 320 total samples. The six axes: sentence complexity, vocabulary richness, expressiveness, formality, consistency, and conciseness. (For the technical details of how we measure these, see how style extraction works.)
But instead of averaging all five models into a single "Average AI" baseline, we kept each model's scores separate and looked at the profile holistically. What kind of writer does each model resemble? What are its habits, its quirks, its blind spots?
Here's what emerged.
Claude Opus 4.6: "The Diplomatic Essayist"
Defining traits: High complexity, high formality, high vocabulary richness, low conciseness.
Opus writes the way a senior partner at a consulting firm talks: thoroughly, precisely, and at length. Its sentences are architecturally complex — nested clauses, carefully placed qualifiers, subordinate structures that unfold over 25-30 words. It hedges with intention, deploying phrases like "it's worth noting that" and "one might reasonably argue" not as filler but as epistemic markers.
Vocabulary richness is where Opus distinguishes itself from every other model. It reaches for the specific word over the common one. Where GPT might say "use," Opus says "deploy." Where Gemini might say "look at," Opus says "examine" or "scrutinize," depending on context.
The tradeoff is conciseness. Opus doesn't do brevity. A request for a "brief summary" will produce four substantial paragraphs. A "short email" will run 200-300 words. Opus treats every communication as an opportunity for thoroughness.
Who writes like Opus: Senior executives, academic researchers, policy advisors, management consultants. People whose professional context rewards precision and completeness over speed.
The Opus gap: If you're naturally concise — the kind of writer who sends three-sentence emails and considers a five-bullet Slack message "comprehensive" — Opus's output will feel bloated. Your style profile would need to push hard on conciseness (target: 65+, Opus default: ~35) while preserving its natural formality.
Claude Sonnet 4.5: "The Balanced Professional"
Defining traits: Moderate across all axes. Highest consistency. Solid vocabulary. Adaptable register.
Sonnet is the generalist. Not in a pejorative sense — in the way a skilled project manager is a generalist. It reads the room and adjusts. Its scores cluster near the center of every axis, but its consistency score edges above all other models, giving its output a steady, predictable rhythm.
This balance makes Sonnet the hardest model to caricature and the easiest to work with across diverse contexts. A formal email from Sonnet sounds appropriately formal. A casual Slack message sounds appropriately casual. It shifts register more smoothly than Opus (which stays formal regardless) and GPT-5.2 (which stays enthusiastic regardless).
Who writes like Sonnet: Operations leaders, project managers, business development professionals, team leads. People who communicate across organizational levels and need writing that works for board presentations and team standups alike.
The Sonnet gap: Sonnet's balance is also its limitation. If your writing has a strong personality — very formal, very casual, very concise, very expressive — Sonnet's moderate defaults won't capture it. A style profile is especially important for distinctive writers who use Sonnet, because the starting point is "average" and the target is "you."
Claude Haiku 4.5: "The Rapid Responder"
Defining traits: Highest conciseness in the Claude family. Strong expressiveness despite short sentences. Lower vocabulary richness.
Haiku breaks the assumption that shorter means duller. Its sentences are compact — averaging fewer words than any other Claude model — but it packs expressive markers into those short structures. Questions, exclamations, direct address. Haiku writes the way a skilled Slack communicator writes: tight, punchy, engaged.
The tradeoff is vocabulary richness. Haiku reuses words more frequently, reaching for common terms over precise ones. This isn't a flaw in casual contexts — most Slack messages and quick emails don't need lexical diversity. But in professional documents that require precision, Haiku's vocabulary can feel limiting.
Who writes like Haiku: Startup founders in early-stage companies, social media managers, sales development reps, anyone whose professional communication happens primarily in chat. People who value speed and energy over elaboration.
The Haiku gap: If you write both chat and formal documents, Haiku's profile matches your chat persona but misses your formal one. This is where context-aware style profiles matter: you need different calibration for different communication types, even within the same model.
GPT-5.2: "The Enthusiastic Communicator"
Defining traits: Highest expressiveness across all models. Lowest consistency. Conversational formality. Dynamic rhythm.
GPT-5.2 is the most emotionally engaged AI writer. It asks rhetorical questions constantly. It deploys attitude markers liberally — "Importantly," "Interestingly," "The truth is," "Here's the thing." It uses exclamation marks and em-dashes at rates that exceed every other model.
This enthusiasm extends to sentence rhythm. GPT-5.2 has the lowest consistency score, meaning it varies sentence lengths the most. Short punchy sentences followed by long explanatory ones. This creates a dynamic, almost conversational feel — but it also means GPT's output has a recognizable cadence that readers increasingly associate with "AI writing that sounds generic."
Formality runs lower than Claude or Gemini. GPT-5.2 defaults to accessible, warm, conversational prose even when the prompt requests formality. It can be formal when explicitly instructed, but its center of gravity pulls toward the casual.
Who writes like GPT: Marketing leaders, community managers, coaches, content creators, HR professionals writing employee communications. People whose professional voice is warm, engaging, and deliberately approachable.
The GPT gap: If you're a restrained writer — someone who considers rhetorical questions unprofessional and exclamation marks juvenile — GPT's expressiveness is your biggest calibration challenge. Your style profile would need to dramatically suppress expressiveness (from ~82 down to perhaps 25-35) while preserving GPT's strengths in readability and vocabulary.
Gemini 3 Pro: "The Methodical Analyst"
Defining traits: Highest formality. High complexity. Lowest expressiveness of the major models. Lowest conciseness.
Gemini writes like a well-organized research paper. Dense, structured, thorough, and almost completely devoid of emotional inflection. Where GPT asks "But what does this really mean?", Gemini states "The implications of this are threefold." Where Claude hedges diplomatically, Gemini hedges systematically.
Formality is Gemini's defining axis. It uses function words at the highest rate, deploys semicolons more frequently than any other model, and structures paragraphs with topic sentences and supporting evidence. This produces output that's immediately credible in institutional contexts — and immediately recognizable as formal in casual ones.
Conciseness is Gemini's weakest dimension. Combined with high complexity, this means Gemini produces the densest text: long sentences packed with information, requiring careful reading. It's thorough to a fault.
Who writes like Gemini: Lawyers, financial analysts, policy researchers, technical writers, compliance officers. People whose professional context demands precision, thoroughness, and institutional credibility — and doesn't penalize length.
The Gemini gap: If you need Gemini's depth but your actual writing is more direct, the conciseness and complexity axes need significant adjustment. A lawyer who writes tight, punchy briefs (they exist) would need Gemini's formality preserved while pushing complexity from ~72 to ~50 and conciseness from ~35 to ~60. See our Gemini writing guide for practical calibration tips.
The Model Personality Map
Here's a summary view of all five profiles:
| Model | Personality | Signature Strength | Signature Weakness |
|---|---|---|---|
| Claude Opus 4.6 | The Diplomatic Essayist | Vocabulary Richness | Conciseness |
| Claude Sonnet 4.5 | The Balanced Professional | Consistency | Distinctiveness |
| Claude Haiku 4.5 | The Rapid Responder | Conciseness | Vocabulary Richness |
| GPT-5.2 | The Enthusiastic Communicator | Expressiveness | Restraint |
| Gemini 3 Pro | The Methodical Analyst | Formality | Conciseness |
AI Model Writing Style Comparison
Comparing Claude Opus 4.6 vs GPT-5.2 in English
What This Tells Us About AI Writing
Three takeaways from profiling the models:
1. Each model has a genuine personality. These aren't random variations. They're stable, measurable stylistic signatures that persist across prompt types and languages. When people say "I prefer Claude's writing" or "ChatGPT's tone works better for me," they're responding to real differences in the data.
2. No personality matches yours. Unless you happen to be a moderate-everything writer who uses above-average expressiveness and below-average conciseness, no model's default profile is your profile. The profiles are interesting — but they're interesting the way knowing your coworker's communication style is interesting. It helps you understand them. It doesn't make them you.
3. The gap between model personality and your personality is measurable. This is the practical payoff. Once you know Gemini defaults to formality-72 and you naturally write at formality-45, you know exactly what calibration is needed. The gap isn't a guess. It's a number. That's the science behind Writing DNA.
For a detailed comparison of the three most popular models, see ChatGPT vs Claude vs Gemini. For how these personalities shift across languages, see Which AI Writes Best in Each Language.
Build Your Own Writing Profile
We gave each AI model a personality based on data. The same system can give you one — but yours is more useful, because it becomes instructions that make any AI model write like you.
Try your free Writing DNA Snapshot — submit a few writing samples and see your six-axis profile compared to Average AI. No credit card required.
Then, if you want to go deeper, a full Style Profile translates that measurement into a master prompt that works across ChatGPT, Claude, and Gemini. Your personality, applied to any model's canvas.