Skip to content

Claude vs ChatGPT for Writing - Which Is Better?

I ran the same three writing jobs through Claude and ChatGPT, including one in my own voice, to see which I would actually keep. Here is what I found on voice, editing and how human the writing felt, with current prices.

Eddie Ochieng

Eddie Ochieng

February 6, 2026· Updated Aug 26, 2026

7 min read
Main Image: Image of a smart phone screen containing icons of AI platforms
Photo by Solen Feyissa on Unsplash

There is a lazy way to answer "Claude or ChatGPT for writing", which is to quote a benchmark, name the newest model, and move on. I wanted the honest version, so I gave both the same three writing jobs on their free tiers, including one written in my own voice, and kept the ones I would actually use. The result was more split than the internet lets on.

This was August 2026, on the free tiers of both. Claude served me Sonnet 5 on Medium and showed a little "thought for 9 seconds" before answering; ChatGPT did not show its model. Same prompts, pasted verbatim, a fresh chat each.

How I tested

Three writing jobs, two tools, free tiers. A paragraph written in my own voice from a sample, a 500-word first-person piece from a blank page, and a bloated paragraph to tighten. I judged them on voice, on how human the writing felt, and on how much I would have to fix afterwards.

The honest headline

They are far closer than the arguments online suggest, and my results genuinely split. Claude was the better mimic of my own voice. ChatGPT wrote the more human piece from scratch and gave the edit I would actually keep. For my own work I now open ChatGPT first, but I did not delete Claude, and the rest of this explains why.

Writing in my own voice

First I pasted in about 150 words of my real writing, a plain account of how I got into building software, and asked both to announce a weekly deals roundup to my newsletter in that voice. This is the one place where the familiar wisdom, that Claude protects your voice, actually held up for me. Claude sounded more like me. It picked up the dry, understated register of the sample and ran with it. Here is how it opened.

Claude · actual output

I didn’t set out to write a newsletter about deals. What happened was simpler, I kept finding AI tools worth using, then losing track of which ones were actually worth paying for versus which had a good landing page and nothing else.

ChatGPT’s was perfectly good and practical, but a step more generic, cleaner and a little less me.

ChatGPT · actual output

I’ve never been particularly interested in collecting AI tools just for the sake of it. What I do like is finding something useful, seeing what it can actually do, and figuring out whether it’s worth the money.

Both are competent. Only one of them read like something I would have written, and if you need writing to sound like a specific person, that difference is the whole game.

Writing from a blank page

Then I flipped it and asked for a 500-word first-person piece from nothing, on quitting a dozen productivity apps for a paper notebook. Here the winner flipped too. ChatGPT's read more human to me, with a more engaging, opinionated flow, while Claude's was clean but a touch more even in tone. That is a useful distinction to hold on to: Claude is the stronger at matching a voice you hand it, and ChatGPT is the stronger at conjuring an engaging voice out of thin air.

Editing your own words

The last job is the one that decides most days, because most writing is rewriting. I gave both a deliberately bloated sentence and asked them to cut it by about 30 percent without losing the meaning or my voice. Both did it well. Here is the same sentence after each of them, verbatim.

ChatGPT tightened it to this.

ChatGPT · actual output

In today’s fast-paced, ever-changing digital world, effectively managing your time can be a key differentiator between those who succeed and those who don’t, especially in the workplace, personal life, and other endeavours.

Claude kept a little more, then listed what it had removed, dropping "it is very important to note that" and "having the ability to" as the dead weight.

Claude · actual output

In today’s fast-paced, ever-changing digital world, effectively managing your time can, in many cases, be the key differentiator between those who succeed and those who unfortunately do not, especially in the workplace and in one’s personal life and various endeavors.

ChatGPT gave the leaner cut, the one I would paste straight back into a document. Claude trimmed a touch less but showed its working, which is genuinely useful if you are trying to learn to self-edit. So ChatGPT for the result, Claude to see why.

The question was never which model is smarter. It is which one leaves me less to fix, and that turned out to depend on the job.

Which one I actually keep

For my own writing I reach for ChatGPT first. The from-scratch writing felt more human and its edits were tighter, which is most of what I need day to day. But Claude stays one click away. When something has to sound like me, or when I want an assistant that explains its edits rather than just performing them, Claude is the one I switch to. The tidy line that "Claude wins writing, full stop" simply did not survive contact with three real jobs.

ToolPriceBest for
ChatGPT (Plus)Free. Go $5/mo, Plus $20/mo.From-scratch writing and tighter edits
Claude (Pro)Free. Pro from $17/mo (annual).Matching your voice and explaining edits
Free tiers$0, both are genuinely capableMost everyday writing
ChatGPT (Plus)

+ Pros

  • + More human writing from a blank page
  • + Tighter, cleaner edits
  • + Editable canvas and no login to start

– Cons

  • More generic when matching a set voice
  • Hides which model you are on
Claude (Pro)

+ Pros

  • + Best at matching a voice you give it
  • + Explains what it cut and why
  • + Shows its thinking before answering

– Cons

  • From-scratch writing felt more even-toned
  • Requires an account to begin

A few things I keep relearning

  • The first draft is a fast intern’s draft, not a final. There is always another pass.
  • Give a voice sample. One paragraph of your real writing changes the output more than any clever prompt.
  • Verify anything checkable. Both invent facts and quotes with total confidence.
  • A clear two-line brief usually beats a 300-word mega-prompt.
  • Switch by task instead of using one tool out of habit.

Prices here were verified in August 2026. See the live pricing comparison and the pricing changelog, or match a tool to your task with the recommender.

For all three big assistants side by side rather than just these two, see ChatGPT vs Gemini vs Claude.

FAQ

Is Claude or ChatGPT better for writing?+

Closer than the internet pretends. In my test ChatGPT wrote the more human piece from scratch and gave the tighter edit, while Claude matched my own voice more faithfully and explained its edits. I open ChatGPT first and keep Claude for those jobs.

Are the free tiers good enough for writing?+

For most writing, yes. I ran this whole test on the free tiers of both and never hit a wall. Paid plans mainly add higher limits and smarter modes.

Which is better at sounding like me?+

Claude, in my test. Given a sample of my own writing it mirrored the tone more faithfully, while ChatGPT drifted a little toward a generic newsletter voice.

Which is better for editing?+

ChatGPT gave the cut I would actually keep, leaner and cleaner. Claude’s edit was good too, and it told me what it removed and why, which helps if you want to learn to self-edit.

Do I really need both?+

Not strictly, but they are cheap enough that many writers keep one paid and use the other’s free tier for its speciality.

Eddie Ochieng

Eddie Ochieng

Eddie Ezekiel Ochieng is a software developer and the editor of The Test Card. He has been writing code since 2016 and has spent the last six years building production web applications for clients, work that runs from a non-profit’s platform to a community dictionary and a personal-safety service. He builds mostly in TypeScript, React and Next.js, with Node, Python and PostgreSQL behind them.

He started The Test Card out of mild irritation. Most AI tool coverage is written by people who never ship anything and never have to live with a bad tool choice. He does. The question that interests him is the practical one, which of these tools survive contact with real work, and which are just a subscription you forget to cancel.

He is not an AI researcher and does not pretend to be. What he brings instead is a builder’s scepticism, a habit of actually reading the documentation and the pricing page, and a bias toward simplicity over novelty. Good software, as he puts it, should feel as good as it works.

eddie-ezekiel.com

Leave a Comment

Share your thoughts about this article