How to Make ChatGPT Work More Like Claude

To make ChatGPT work more like Claude, describe the response habits you prefer and give ChatGPT examples, task context, and explicit output requirements. You can guide its tone and working approach, but a prompt does not turn ChatGPT into Claude or give it Claude’s features.

The practical goal is to carry a useful way of working between assistants. fenbs can keep the approved style brief and project record outside either app, so both connected assistants can read the same instructions and current next step.

Which part of the Claude experience do you want to reproduce?

1. Choose a Claude response that worked well for a real task. Remove any private information before using it as a reference.
2. Name the observable qualities you liked: a direct opening, connected paragraphs, useful questions, or a clear distinction between facts and suggestions.
3. Give ChatGPT the same source material, audience, goal, and constraints that were available in the successful task.
4. Test the instructions on a small new task. Compare usefulness and accuracy, then revise the rules you can actually evaluate.

“Be like Claude” leaves too much unstated. You might prefer one particular answer because its structure suited your task, not because every response from that assistant has that quality. Write down the behavior you want without treating either product as having a single permanent personality.

For differences between the products themselves, read Claude versus ChatGPT. This page focuses on adapting response instructions rather than selecting a winner or reproducing an app interface.

How do you translate a preference into a useful instruction?

Describe what should appear in the answer, when it should appear, and what the reader needs it for. A good instruction gives ChatGPT a decision it can apply to the next task. An instruction that only names a mood is harder to review and easier to interpret differently.

OpenAI’s personality guide↗ (opens in a new tab) says personality changes guide communication rather than capabilities. Treat any available style setting as a starting point. Your particular deliverable still needs instructions about its audience and purpose.

What prompt should you try first?

Use a compact working brief rather than a role-play instruction. Keep your preferred response habits separate from the project facts so you can reuse the habits without carrying unrelated assumptions into a new assignment.

Help me complete [task] for [audience]. The desired result is [deliverable].

Response preferences:
- Answer the main question first.
- Use plain language and connected explanations.
- State important assumptions and uncertainties.
- Ask a focused question only when missing information changes the outcome.
- Give concrete next steps when action is needed.
- Challenge a weak premise with a reason; do not add praise by default.

Approved facts and constraints: [paste current context]
Example of the structure I like: [paste a permitted short example]
Use the example for structure and tone only. Do not copy its facts or wording.

Return [format and length].

Try the prompt on an ordinary assignment, such as explaining a delayed website launch to a client. Check whether ChatGPT states the new plan clearly, preserves the approved date, and identifies any missing decision. A response that resembles your example but invents a commitment has failed the more important test.

OpenAI’s prompting guidance↗ (opens in a new tab) provides a reference for structuring requests. For the separate problem of making the actual prose sound more natural, see writing like a human with ChatGPT.

Where should recurring instructions and project facts go?

Keep stable communication preferences in a reusable place, and supply current task facts when you ask for work. A rule to explain technical terms may apply broadly. A particular client’s approved terminology or an unfinished review belongs with that project.

ChatGPT offers custom instructions↗ (opens in a new tab) for response preferences. Review what you store there so a one-time request does not affect unrelated work. Keep the task-specific details short enough to inspect and update when the work changes.

If you use both assistants, keeping two separate copies of the whole project creates an upkeep problem. A changed decision may reach one chat while the other still works from an earlier answer. A shared work record makes it easier to identify which instruction is current, provided each assistant reads it before working.

The fenbs integration guides explain how supported assistants connect to the same record. For moving an existing body of work between the apps, use the separate ChatGPT-to-Claude guide.

What should you do when the answers still differ?

Expect some variation. The assistants can make different choices about wording, structure, and which detail deserves attention. Compare them against a short acceptance checklist rather than asking one to imitate every sentence from the other.

Use the same task and source material in both. Check three things first: whether the result answers the question, whether the claims are supported, and whether the next action is clear. Then adjust one instruction at a time. Changing five preferences at once makes it hard to tell which change helped.

An instruction cannot add another app’s tools, permissions, or interface. If you need a particular product feature, evaluate that separately rather than writing a longer persona prompt. The AI prompts hub covers more focused instruction problems.

How to do this with fenbs?

1. Create a free account and add the project you work on across ChatGPT and Claude.
2. Save the approved response brief in Skills and prompts. Include a short example and a checklist describing a useful answer.
3. Add the current tasks and record the choices you have made in Decisions and rules. Keep unresolved choices open for a person to decide.
4. Connect the supported apps and ask each to read the project context before working. Before switching, save progress with what was completed, what was checked, and the next step.

fenbs does not change either model or copy private chat history between them. It is not necessary if you use one assistant for isolated questions. It is useful when a reusable brief and an accurate record of ongoing work save you from repeating the same setup.