GeminiGetting startedLevel 1: Basic prompts

How Do You Write Your First Gemini Prompt?

Say what you want, give the background, and name the format. Google's help pages describe four areas to cover, persona, task, context and format, and say you don't need all of them. If your question is about a document or a picture, attach it with Add files instead of describing it.

Prompt to copy
You are [a role, e.g. a bookkeeper who explains things to small-business owners].

Task: [one specific thing to produce, e.g. a reply to the customer review below].

Context: [who it's for, what happened, what you've already tried]. [Paste the text here, or attach the file.]

Format: [e.g. 4 sentences, plain language, no bullet points].

Avoid: [anything you don't want]. If something essential is missing, ask me one question before you answer.
  • Personathe role Gemini should play
  • Taskwhat you want made or done
  • Contextbackground it can't guess
  • Formatstructure and length of the answer
The four areas Google suggests for Gem instructions, which it says follow good prompt writing. You rarely need all four.

Try it now

  1. 1Go to gemini.google.com, sign in, and click New chat.
  2. 2If your task involves a file, click Add files, then Upload Files. Paste the prompt above with the brackets filled in, and click Submit.
  3. 3If the answer misses, click Edit text next to your prompt, fix it, and click Update. Gemini regenerates the response.
  4. 4You'll know it worked when you can use the answer with only small edits.

Use it when

Any time the answer has to be used: an email you'll send, a plan you'll follow, a summary someone else will read. It costs one extra sentence or two, and it saves rounds of correction.

Skip it when

A simple fact, such as a time zone or a word's meaning. Skip the structure too when you don't yet know what you want and need ideas first. Ask loosely, then write a proper prompt once you can see the direction you prefer.

Why it works

Google's prompting docs tell developers to state the goal and constraints directly, define ambiguous terms and name the output format. They also say Gemini 3 models give direct answers by default, so you have to ask if you want more detail or a conversational tone. Those docs are written for the API, but the same habit fits the app. It matters more on a free account, because Google says auto model selection sends most prompts to Flash-Lite, and the default thinking level is Low. A vague prompt there gets a quick, light answer.

Example

Say a customer left your cafe a one-star review about a cold sandwich. "Reply to this review" leaves Gemini to guess the tone, the length and what you can offer. Instead, tell it that you're the owner, paste the review, and ask for 4 warm sentences that don't sound defensive, offer a free replacement on their next visit, and don't mention the staff member's name. You've now made every decision that Gemini would otherwise have guessed at.

Going further

You can ask for the finished file in the prompt itself. Google's help says Gemini can create formatted files directly from a chat, including Docs and Sheets, .pdf, .docx, .xlsx, .csv, LaTeX, TXT, RTF and Markdown. So "put this in a spreadsheet with one row per student" gets you a file you can download or export to Drive, not a table you have to copy out.

Common mistakes

  • Describing a screenshot or a report in words when you could attach it. Gemini reads images and most file types directly.
  • Correcting a bad first prompt with five follow-ups. Edit text fixes the original, and Google says long histories use more of your limit.
  • Expecting an essay from one line. Google's docs say Gemini 3 models answer directly by default, so ask for depth if you want it.
  • Pasting a client's private details before checking what Keep Activity does with them. Some chats are read by human reviewers.

Sources

Checked 6 November 2026. Products change often, so confirm details with the vendor.

Related

Beyond the course