How to Make Gemini Work More Like ChatGPT

To make Gemini work more like ChatGPT, describe the response habits you want, provide an example, and supply the task context Gemini needs. You can make the experience more familiar, but instructions do not turn Gemini into ChatGPT or guarantee identical answers.

A shared prompt and work record in fenbs helps you reuse those instructions across assistants instead of rebuilding them for every switch. Start with what made the earlier answer useful: its structure, depth, vocabulary, or way of handling uncertainty.

What should you transfer first?

1. Choose one real task you often do in ChatGPT, such as revising a client email or reviewing meeting notes.
2. List three observable preferences. For example: answer first, keep paragraphs short, and identify missing information before making a recommendation.
3. Give Gemini those preferences and the task’s current facts. Include a short, permitted example of the format you want.
4. Review the result against your criteria, then change one instruction at a time. Keep the version that produces work you can use.

Avoid starting with “be exactly like ChatGPT.” That gives Gemini no clear test to pass. “Give the recommendation in the first sentence, then show the evidence and unresolved questions” tells you and the assistant what success looks like.

If you are deciding which app suits the work, Gemini vs ChatGPT covers that choice. This guide assumes you want a more familiar working style inside Gemini.

Where should you put reusable preferences?

For an initial test, put the preferences directly in the prompt. Once they work, use the persistent instruction controls available to your account. Google’s instructions guide↗ (opens in a new tab) explains where to add, edit, disable, or remove them and which account requirements apply.

Check the current labels in Settings & help rather than relying on an old screenshot. Keep lasting preferences separate from a temporary assignment. “Use US English” may apply to many chats; “review the attached October proposal” belongs with the project task.

OpenAI’s personalization guide↗ (opens in a new tab) similarly distinguishes response preferences from capabilities. Review your ChatGPT preferences for the ideas you want to carry over, then rewrite them in plain language instead of assuming settings transfer automatically.

Store the tested instruction text in fenbs’s Skills and prompts, so you can read and revise one clear version when using different assistants.

How do you turn a vague preference into a useful instruction?

Describe behavior you can recognize in the answer. “Be smarter” is hard to evaluate. “Separate confirmed facts from assumptions, and tell me what evidence would change your recommendation” creates a concrete review step. Keep the list short enough that you can notice when a rule is missed.

Add one example of a successful result when wording alone is insufficient. Use your own text or material you are permitted to share. Tell Gemini which properties to follow, such as the order of sections or amount of detail, so it does not simply imitate incidental phrasing.

What is a good starting prompt?

Use the following structure for a familiar, practical response. Replace the task details, then paste your approved sources or attach the files you are allowed to use. The point is to transfer a working method, not to request another company’s identity.

Help me revise a client update for a US design studio.
Start with the revised update, then give a short list of material changes.
Use plain language and short paragraphs. Keep approved project names unchanged.
Do not invent progress, dates, approvals, or customer statements.
If essential information is missing, name the gap before drafting that part.

Current goal: explain the completed design review and the next approval needed.
Approved facts: [add reviewed facts].
Source text: [paste the draft].
Keep this structure: completed work, open question, next action.

When the result misses, point to a specific failure: “You presented a suggested date as approved; remove it and ask for confirmation.” Save the corrected prompt after testing it on another example. A prompt that works once may still depend on information you forgot to include.

How do you carry the task context across?

Give Gemini the facts needed for the next action: the goal, audience, approved choices, current draft, relevant sources, and unresolved questions. Remove abandoned ideas from the main summary, or label them as rejected. Otherwise the assistant may treat an old option as the current direction.

If a task needs files, Google’s upload guidance↗ (opens in a new tab) explains the available upload paths and current restrictions. Provide the permitted source separately; a reference to a ChatGPT attachment is not the attachment itself. Ask Gemini to confirm which source it can read before relying on the analysis.

Say a studio is drafting a launch announcement. The reusable preference might be “answer first and avoid promotional filler.” The project facts are the approved product description, intended audience, and publication decision. Keep those two kinds of information distinct so a writing preference does not become a false project fact.

For personalization behavior specifically, see whether Gemini has memory like ChatGPT. For reusable configurations, the Gemini Gems guide covers that neighboring topic.

How can you tell whether the setup is working?

Test with a small set of tasks whose quality you can judge. Check whether the answer follows the requested structure, preserves approved facts, uses the provided evidence, and admits gaps. Compare time spent correcting the result rather than judging only how familiar the wording feels.

Use Google’s guidance for Gemini Apps↗ (opens in a new tab) for the available interface and remember to verify responses. A preferred tone is not evidence that the answer is correct. Review names, figures, quotations, and source claims before using the output.

How to keep a reusable setup with fenbs

1. Create a free account, then add the project you use across assistants.
2. Save the prompt you have tested in Skills and prompts, with a clear description of when to use it.
3. Record the project’s approved choices and create tasks for the remaining work. Save the exact next step in Where we left off.
4. Use the integration guides for supported connections. Have a connected assistant read the project context before using the prompt, and save reviewed progress afterward.

fenbs is free for one person, no credit card. It is useful for repeatable work across sessions; a single temporary preference may only need one prompt. More guides in the switching models hub explain how to keep the project consistent while accepting that different assistants will answer differently.