This Space holds a small comparison kit for writers. It gives you five writing tasks, one prompt for each and a short scorecard. You run the same prompt in Claude and in ChatGPT, paste both answers next to each other and score them.
There's no winner announced here. Both companies update their models every few months, so any verdict goes stale. The useful question is which assistant does better on the kind of writing you do, and only your own drafts can answer that.
Write a short email to a client explaining that their March invoice will go out one week late because we're replacing our billing system. Keep it polite and under 120 words.
Look at the first two lines. Does the delay appear there, or is it buried under pleasantries? Is there a new date? Does the email apologize once, or three times?
Write one paragraph for a college essay arguing that public libraries should stay open until 10 PM. Use one concrete example.
A strong answer makes a clear claim in the opening sentence and backs it with one specific scene, such as a night shift nurse who can only study after work. A weak answer opens with a broad line about modern life and never lands on a person.
Write a 200 word scene where two siblings argue about selling their late father's fishing boat. Show the tension through dialogue only.
Read the dialogue out loud. Do the siblings sound like two different people? Does the scene leave something unsaid, or does a character announce exactly how they feel? Style shows up most in fiction, so give this card extra attention.
Summarize the article below in five bullet points for a manager who won't read the original.
Paste in a real article of about 600 words. Then check every bullet against the source. Mark anything that's missing, reordered in a misleading way or simply made up. A summary with one invented fact should lose points even if it reads well.
Rewrite this paragraph so a general reader can follow it. Keep every fact.
It has been determined that, in light of the aforementioned budgetary constraints, the facility will be operating on a reduced schedule commencing on the first of next month, with weekend hours being discontinued until further notice.
Count the facts before you score. There are four. Money is tight, hours are being cut, the change starts next month and weekends are closed for now. A good rewrite keeps all four and sounds like a sign a person would tape to the door.
Give each answer a score from 1 to 5 on these questions.
That gives each answer a total out of 25 for each task.
Compare totals task by task, not as one grand score. It's common for one assistant to take the email and the summary while the other takes the fiction. That isn't a contradiction. It just means you might use both.
Also look for habits across tasks. Some answers lean on bullet lists when you asked for prose. Some hedge every claim. Some reuse the same openers in every draft. A habit that shows up in all five cards will show up in your real work too.
If the totals are close, pick the one that needs less editing. The editing load question is usually the one that matters on a busy Tuesday.
Task 5 is different from the other four. A rewrite has a fixed target. The goal is the original meaning in clearer words, with no new ideas slipped in. General chat tools are built to be helpful, so they sometimes add a sentence or drop a detail they decided wasn't important. If most of your work is reworking text you already have, read how a dedicated rewriter compares with ChatGPT before you commit to either assistant.
Card five can hold a third contender too, so hand the facility notice to Rephrase AI with Professional selected and mark its answer as version C on that same 25 point scale.
Post your filled scorecard in the Community tab of this Space. Include the model names, the date and which card surprised you. Other writers can then see how results shift from one version to the next.