ChatGPT × Claude
ChatGPT or Claude? Wrong question
I pay for both and use both every day. Here is where each one actually wins, the honest failure they share, and the setup where they work together instead of competing for the same twenty quid.
no sponsor, both paid for
Where ChatGPT wins
The all-rounder
- Images are built in. ChatGPT Images 2.0 lives inside the chat on every plan, and it is genuinely good. Claude has no image generator at all. (Sora, the video app, was retired in early 2026; the still-image tool is what survived and improved.)
- Speed on small thinking. Quick brainstorms, breaking a fuzzy problem into parts, finding the word that will not come. Low ceremony, fast back-and-forth.
- A bit of everything. Voice, browsing, a huge ecosystem of custom GPTs. One app covers the whole surface, shallowly but conveniently.
Where Claude wins
The craftsman
- Writing that sounds like a person. ChatGPT's prose is competent but recognisable at fifty paces. Claude's is closer to how you would actually say it, provided you direct it.
- Your real files. Claude Code and Cowork work inside your actual folders: your site, your product copy, your admin. ChatGPT talks about your work; Claude works in it.
- The long haul. A bigger working memory at the same price, and it holds a brief across a long build instead of drifting halfway through.
The part nobody puts in the table
Both of them flatter you. A 2026 Stanford study measured chatbots agreeing with users roughly half again as often as humans do, even when the user was plainly wrong. ChatGPT wears the reputation (OpenAI had to roll an update back for being too agreeable), but no assistant is immune, and that includes Claude. The fix is the same either way: tell it, in writing, to push back on you. It goes in your brain file, and chapter 6 of the guide covers exactly that.
Claude's honest weakness is usage limits. On a heavy build day you can hit the ceiling and wait, which is properly annoying, and one more reason the two-app setup below works: the quick, chatty work never burns your Claude allowance.
Sketchbook and workshop
The comparison articles all end with a winner. In practice the strongest setup is a division of labour: ChatGPT as the sketchbook, Claude as the workshop. ChatGPT gets the fast, disposable thinking: first brainstorms, untangling something you cannot word, a rough plan, an image study for a moodboard. Claude gets everything that has to survive: the build, the copy, the long document, anything that touches your files.
One line on the imagery: generated images are for prototyping and moodboards, never the finished creative. The finished work is the part only you can do, and handing it to a generator takes the art out. That rule holds whichever app is drawing.
If you only want to pay for one: it depends on what your day is. If your work is mostly asking and deciding, ChatGPT covers it. If your work is making things, pick Claude; it is the one that can sit inside the work with you, and this whole site exists to show you how.
Who gets which job
- ChatGPT: first brainstorms, breaking a problem down, wording and technicalities, quick plans
- ChatGPT: image prototypes and moodboards, never finals
- Claude: the build itself, anything in your files, long documents, the heavy lifting
- Claude: the brain file, the workspace, the work that compounds
sketchbook · workshop
Carry your ChatGPT memory over
Months of context in ChatGPT do not have to stay behind. Claude has a built-in import for exactly this (claude.com/import-memory, on all plans since March 2026). The flow is honest and slightly funny: Claude hands you a prompt, you paste it into ChatGPT, ChatGPT writes out a summary of everything it knows about you, and you paste that back into Claude's memory settings. The same trick works from Gemini or Grok.
Know its limits before you lean on it: it moves your memory and preferences, not your full chat history, and Anthropic still labels it experimental. Read what arrives and correct anything wrong; it is a handover document, not gospel.
What moves, what doesn't
- Moves: what it knows about you, your preferences, your projects
- Moves: custom instructions, as summarised context
- Stays: your full chat history
- Stays: uploaded files and images
experimental · read what arrives, correct it
Put it in a file you own
A model's memory is a black box on someone else's server, whichever model it is. The better home for everything that matters is a CLAUDE.md: a plain text file in your workspace that describes you, your work, your rules and your taste. You can read it, edit it, version it and carry it anywhere, including away from Claude, because it is just text.
The ten-minute move: ask ChatGPT for everything it knows, hand the result to Claude, read it back, cut what is stale, keep what is you. The exact words are in the card. From then on, every conversation starts already knowing your business; chapter 6 of the free guide builds the whole file with you, from scratch or from an import.
The two prompts, verbatim
01 · to ChatGPT"Write out everything you know about me, my work, my preferences and my projects, organised under headings."
02 · to Claude"Help me turn the true parts of this into my CLAUDE.md."
ten minutes · yours forever
01 · Re-explaining
Stop introducing yourself
With a CLAUDE.md in your workspace, Claude already knows who you are and what you make. The paragraph you used to paste into every chat retires today.
02 · Magic prompts
Brief it, don't incant
Claude working in your actual files responds to a plain brief plus your rules, not prompt spells. Say what you want, what good looks like, what it must never do.
03 · Disposable chats
Build, don't chat
Here the work accumulates: files, folders, the brain file. You are building a workspace that compounds, not having conversations that evaporate.
The full method, free
The 16-chapter guide takes you from a fresh machine to a working creative workspace, teaching you to teach yourself as you go.