Prompt for Gemini AI

Prompt Engineering Updated August 24, 2026 by GenPrompto
Prompt
Act as a [role, described specifically -- e.g. "senior copywriter with 10 years in B2B SaaS" rather than just "writer"]. Task: [describe exactly what you need, including any constraints on length, tone, or approach]. Context: [any background information Gemini needs to do this well -- audience, prior attempts, what hasn’t worked]. Format the response as [desired format, e.g. "a numbered list of 5 items, each with a one-sentence explanation" or "a single paragraph under 100 words"]. If anything in my request is ambiguous, ask a clarifying question before answering.

What this does

A flexible Gemini prompt template -- role, task, context, format -- that adapts to research, writing, planning, or coding tasks.

Gemini responds best to prompts that separate four things clearly: who it should act as, what exactly you need, any context it needs to know, and how you want the answer formatted. This one template adapts to research, writing, coding, or planning tasks just by changing what’s inside the brackets.

What this produces

A response shaped by the role, task, and format you specify — rather than Gemini guessing at tone, depth, or structure from a bare question.

Variations

Two ready-to-copy versions for the most common starting points.

Research Question Version
Act as a research assistant with a habit of citing sources and flagging uncertainty. Answer this question with current, accurate information: [your question, stated specifically]. For each claim, note your confidence level and, where possible, the type of source it likely comes from (official documentation, news reporting, general consensus). Explicitly flag anything that might be outdated or that you’re not fully certain about, rather than presenting it with the same confidence as verified facts.
Step-by-Step Task Version
Act as a [role, e.g. “project planner with experience running small team launches”]. Break down [task or goal, described specifically] into a numbered list of concrete, sequential steps. For each step, include: what needs to happen, roughly how long it should take, and a one-sentence note on why that step matters or what commonly goes wrong there. Flag any steps that depend on external factors outside direct control.

Who this is for

One prompt structure, four slots to fill — built for research, planning, or writing tasks where you don’t want to reinvent the format each time.

Example Output

Filling in a specific role and format (e.g., “act as a copywriter, write 3 headline options, format as a numbered list”) reliably returns exactly 3 headlines in list form — leaving format unspecified, Gemini might return a paragraph, a table, or a different number of options depending on how it interprets the request.

Tips for better results

  • Name a role even for simple tasks — “act as an editor” produces noticeably different output than no role at all.
  • State the output format explicitly (list, table, paragraph, code block) — this is the detail most requests skip and most often causes a mismatched response.
  • For research or fact-based questions, ask Gemini to flag uncertainty rather than state everything with equal confidence.
  • Keep context brief and relevant — padding the prompt with unrelated background can dilute focus on the actual task.

What didn’t work as well

A bare question forces Gemini to guess at tone, depth, and format all at once. Naming even one of the three — role, context, or format — narrows that guesswork substantially.

Related: try Best Prompts for Claude or Effective Research Agent Prompts for Information Retrieval for a different angle on this.

FAQ

Does this work for image generation too?

This template is built for text-based tasks — research, writing, planning, coding. For image generation specifically, use the dedicated Best Gemini Prompts for Images template instead.

Do I need to restate the role in every follow-up message?

No — Gemini generally keeps the role and context from earlier in the conversation. Restate it only if you’re switching to a different type of task.

What if Gemini’s response is too long or too short?

State a length explicitly next time (“in 3 sentences” or “in detail, covering X and Y”) — Gemini follows explicit length instructions much more reliably than vague ones like “briefly.”

Related