Is Claude or ChatGPT better at writing in my style?

What each company says about its own default style

OpenAI describes ChatGPT's Default personality as clear and neutral and offers six others. Its help page also draws a line that matters to anyone writing: the personality shapes how ChatGPT talks to you, while a draft you ask for, such as an email, a post or a résumé, follows your instructions and the context instead.

Anthropic's prompting guide says its latest models write in a more concise, more conversational way than earlier Claude models and may skip summaries unless asked. That compares Claude with older Claude, not with ChatGPT.

Neither company publishes a test of which model sounds more like a given person, and we haven't run one. Verdicts online describe a moment: Anthropic's release notes alone list three new models between 22 September and 7 October 2026. Check the date on any comparison, this one included.

Setting up ChatGPT for your style

ChatGPT gives you three layers. Custom Instructions hold rules for every chat, inside the 1,500 or 5,000 character limit. A project holds your samples as files plus its own instructions, which replace the global ones while you work in it. The personality and Characteristics settings adjust warmth, enthusiasm, formatting and emoji, which shapes chat replies far more than a cover letter.

The layer many older guides recommend, the custom GPT, is closing. OpenAI says personal accounts can no longer create new GPTs, that custom GPTs are scheduled to retire on 11 December 2026 and that plugins are the replacement. Its plugins page says personal skills are generally available on Business, Enterprise, Healthcare and Edu, so on a personal plan today a project is the practical home for a writing style.

Setting up Claude for your style, including a writing-style skill

Claude's personalization page lists three tools. Instructions for Claude, under Settings, apply to every conversation. Project instructions apply inside one project, next to a knowledge base where your samples can sit. The projects page offers projects on every plan, up to five on Free, though the personalization page lists project instructions under paid plans. Skills are the third, and the personalization page says they can apply communication patterns based on your own writing.

A skill is a folder with a skill.md file that starts with a name (up to 64 characters) and a description (up to 200). Claude reads the description to decide when to load the skill, so write it as a trigger. This one is 137 characters: “Use when I ask for an email, post or letter that goes out under my name. Applies my sentence length, word choice, openings and sign-offs.” Under it, paste your style recipe and two or three short excerpts of your own writing.

To add it, turn on Code execution and file creation under Settings, then Capabilities. Open Customize, then Skills, choose + and Create skill, and upload the folder as a ZIP file. Anthropic lists skills on the Free, Pro, Max, Team and Enterprise plans.

Older guides mention a Claude feature called Styles. On 10 October 2026 we couldn't find a current Anthropic help page for it, and the personalization page lists the three tools above.

A side-by-side test you can run in 15 minutes

Use the kind of writing you do most. You need three samples, one style recipe and the bare facts of a piece you have already written yourself.

Set both tools up the same way: in ChatGPT, a project with the samples as files and the recipe as its instructions; in Claude, a project or a skill holding the same material. Then paste the same facts and the same one-line request into a new chat in each.

Judge the two drafts against your own piece, in this order. How many sentences would you rewrite before sending it? How many facts appear that you never gave? How many words would you never use? Does it open and close the way you do? The draft with fewer rewrites wins, whatever the rest says.

Do it for three requests, not one, because any single draft can be lucky. Repeat the test when either company releases a new model.

Which one to use, by the job

A few decision rules, drawn from the documentation above rather than from a taste test.

You already pay for ChatGPT and mostly write short pieces such as emails and posts: put your rules in Custom Instructions and keep one project per kind of writing.

You want one style procedure that loads by itself whenever you ask for a certain kind of text: that is what Claude's skills are for, and they work on the Free plan once code execution is on.

Your samples are long, such as chapters of a book: both tools take files in projects. Claude's paid plans stretch project knowledge with retrieval, and ChatGPT's file limits per project are in the table.

You would rather not maintain samples and recipes in either: a dedicated writing tool is a third route. HidenGPT is one; it keeps a voice profile built from your own samples, drafts in it and runs a red pen over stock phrases, on a paid plan. Whichever you choose, the facts and the final read have to come from you.