How to Make Gemini Write Like Claude: Transfer Your Brief

To make Gemini write more like the Claude drafts you prefer, give it a specific writing brief, examples you approve, and a checklist for editing the result. Asking it to imitate Claude by name leaves the actual requirements unstated and cannot make the models identical.

Treat your favorite draft as evidence of your taste: perhaps it leads with the answer, uses concrete examples, and removes filler. Tools like fenbs keep that approved prompt outside either AI app, so you can reuse the same standard when you change models.

How do you transfer a writing style in five steps?

1. Choose a real assignment and one or two examples you have reviewed. Use text you are allowed to share, with private details removed.
2. Describe the reader, purpose, desired length, tone, and structure. Replace vague praise such as thoughtful with observable requirements.
3. Label what the examples demonstrate. Explain whether you want their pacing, opening, use of evidence, or level of detail.
4. Give Gemini a fresh assignment with the same requirements. Keep the source facts separate from the sample prose.
5. Review the draft against a short checklist, revise the weakest part, and save the instructions that improved the result.

For example, a design studio could test a client update explaining a delayed illustration. Success means the client understands what changed, what happens next, and whether they need to respond. Sounding pleasant without delivering those facts is a failed draft, even if its wording resembles an earlier answer.

What should you extract from a Claude draft?

Separate the qualities of the prose from the facts of the old assignment. A useful opening might state the answer immediately; it does not follow that every future piece needs the same opening sentence. Describe the pattern so Gemini can apply it to different material.

Avoid treating a brand name as a universal style guide. The broader Claude versus Gemini comparison helps with choosing an assistant; this exercise establishes your own editorial standard. A draft that meets it can be useful even when its rhythm differs from the sample.

What does a reusable writing prompt look like?

Start small enough that you can tell which requirement changed the output. Google’s prompt design guidance↗ (opens in a new tab) describes using examples to guide phrasing and format. Anthropic’s prompting guidance↗ (opens in a new tab) also recommends explicit instructions about the desired output.

Write a client update for a busy design client.
Goal: explain the delay and the next agreed step.
Length: 180-230 words.
Lead with what changed. Use short paragraphs and familiar words.
Be calm and direct. Do not add enthusiasm or apologies we did not request.
Use only the facts below. Flag missing dates instead of inventing them.
Sample: [paste an approved update]
Copy its direct opening and practical ending, not its facts or phrases.
Facts for this assignment: [add verified facts]
Before returning the draft, check accuracy, purpose, tone, and length.

Add one sample that shows the desired effect and explain why you selected it. If the draft still uses overly formal language, give a short before-and-after example from your own editing. A contrast such as utilize versus use is easier to apply than a request to sound more human.

Keep the approved prompt in fenbs Skills and prompts, where connected assistants can list and read it instead of relying on a remembered chat.

How do you edit without chasing an exact imitation?

Read for meaning before polishing sentences. Check that each date, name, promise, and source matches the material you provided. Then check whether the opening answers the reader’s question and the ending gives them an appropriate next step.

• Accuracy: can you trace every factual claim to supplied material?
• Usefulness: does the reader know what to understand or do?
• Voice: does the language fit your audience and approved examples?
• Economy: can you remove a sentence without losing anything the reader needs?

Ask for a targeted revision, such as shortening the second paragraph while preserving the delivery date. Repeatedly requesting another Claude-like version gives you many alternatives without a clear reason to prefer one. Record the correction that worked so your next draft starts closer to the standard.

Where should the prompt and examples live?

Keep one approved version with a clear purpose, such as client updates or product explanations. Put assignment-specific facts in the current request. This prevents a deadline or client name from becoming part of your permanent writing instructions.

If you want to save instructions inside an app, check the current setup described in Claude Projects versus Gemini Gems. If your real problem is conversation style and starting context, use the separate guide to making Gemini act like Claude.

How to do this with fenbs

1. Create a free account and add the project whose writing you want to make consistent.
2. Save the approved prompt, a permitted example, and the editing checklist in Skills and prompts. Give the entry a name that describes its job.
3. Follow Integrations to connect a supported assistant, such as Gemini CLI or Claude. MCP is a standard way for an AI app to use an outside tool; the connection uses browser sign-in.
4. Ask the assistant to read the saved prompt, then provide the current assignment and verified source facts. Review the result and save any approved improvement to the prompt.

fenbs does not train Gemini or copy Claude’s writing ability. It is unnecessary if you only need one isolated draft and already have its instructions at hand. It helps when the approved standard needs to survive many assignments or a change of assistant.

How can you tell the transfer worked?

Try the prompt on a different assignment before calling it finished. A second client update should preserve the same clarity without recycling the old facts. Keep the stronger prompt when it reduces your editing and produces an accurate, useful piece, regardless of which model supplied the first example.

For background, what Claude is explains the product; the switching models hub covers other transfers. Your immediate next step is simpler: save one approved example and the four checks that made it worth keeping.