Gemini prompting for marketers: models, modes and context

#01What does a strong marketing prompt look like?

Google's Workspace prompting guide recommends four building blocks. Here's how they map to marketing work.

Two extra parts make prompts reliable for marketing:

• Source rules. "Use only the attached files. If a fact isn't in them, write 'not in sources'."
• A self-check. "After drafting, list any claim that needs a citation, and any number you didn't take from the files."

Here's a full template you can paste:

You are [role] at [company type], writing for [audience].
Task: [deliverable], [quantity], for [channel].
Context: [product facts, offer, dates]. Use only [files or @Drive doc].
Constraints: [banned words, legal lines, reading level].
Format: [structure, length, table or bullets].
Before you answer, ask me up to 3 questions if anything is unclear.
After the draft, list claims that need a source and any assumptions you made.

#02Which model should you choose?

On AI Pro and Ultra the app has a model picker. On the free plan you use a Flash-Lite model.

Our view: the model matters less than the context. A Flash prompt with your real customer quotes beats an Argon prompt with no context.

#03How do you give Gemini your real material?

1. Upload files. Click the + in the prompt bar and add PDFs, docs, slides, spreadsheets, images or audio. Google says the app handles documents of up to 1,500 pages.
2. Mention Workspace apps. Type @ and pick Drive, Gmail, Docs or Calendar, then refer to the file or thread. For example: "@Drive summarise the Q3 win-loss notes doc."
3. Connect apps. Under Customize, Google now lists more than 40 connected apps, including Canva, Adobe and others. Many are US English only at first.
4. Paste a URL. Gemini can read public pages, but it may summarise rather than quote. For exact wording, paste the text.

#04When should you use Canvas, Deep Research or a normal chat?

In Canvas you can highlight a paragraph and ask for a change to just that part. That's much faster than asking for a whole new draft.

#05What prompt patterns work best for marketers?

#06How do you check Gemini's work?

• Ask for sources. In grounded answers, open at least the two sources behind any claim you'll publish.
• Use "double-check". Some Gemini responses offer a Google-it or double-check option that highlights statements and finds supporting or conflicting search results. Labels vary by surface.
• Check numbers yourself. Treat every statistic as unverified until you've opened the source.
• Watch the dates. Ask "What date is each source from?" for anything about prices or features.

Our view: one source check per published claim is the minimum. It takes two minutes, and it's the difference between a useful assistant and an embarrassing correction.

#07What about API prompting?

If a developer is building with the Gemini API (lesson 10), the same structure applies, with two changes:

• Put persona, rules and format in the system instruction, and the variable input in the user message.
• Keep the defaults that Google's prompting strategies guide recommends for current models, and change one thing at a time when you test.

#08Worked example: Ledgerline

Ledgerline is a fictional accounts-payable software company. Details are illustrative.

Ledgerline's content lead needed a launch email for a new approval feature. The first prompt was "Write an email about our new approval feature."

Walkthrough

1. The lead wrote a context pack doc in Drive with ICP, proof points and banned words.
2. She used the template above with "@Drive Ledgerline context pack" and the release notes attached.
3. She asked Gemini to ask up to three questions first. It asked about the CTA and the send segment.
4. She ran the critic pass as "a sceptical finance controller at a 500-person company".
5. She checked the customer quote against the original interview transcript before sending to review.
6. She saved the final prompt as a draft skill for lesson 5.

#09Common mistakes

• One-line prompts. No persona, context or format means generic output.
• No source rules. Gemini fills gaps with plausible but unverified claims.
• Using Argon for everything. It's slower, and quality gains on short copy are small.
• Pasting a URL and expecting exact quotes.
• Rewriting a whole draft to fix one paragraph. Use Canvas edits instead.

#10Try it yourself

Take a real task from this week. Write it as a one-line prompt, then as a full template prompt with a context pack. Run both and compare them side by side. Note which elements made the biggest difference, then save the better prompt in a doc for lesson 5.

#11Quick quiz

1. What four parts does Google's prompting guide recommend?

Answer: Persona, task, context and format.

2. How do you refer to a Drive file in a Gemini app prompt?

Answer: Type @ and choose Drive, then name the file.

3. When should you use Canvas instead of a normal chat?

Answer: When you're editing a document over several turns and want targeted changes.

#12What should you check before the next lesson?

• You have a context pack doc in Drive.
• You have one tested prompt template for a recurring task.
• You know which model you'll use for drafts and which for analysis.

Previous: Gemini plans and pricing for marketing teams. Next: Gemini Deep Research for competitor and market research. Or return to the Gemini for Marketers track. Related: Perplexity prompting and source control, Claude Projects for brand voice, Gemini skills.

#13Common questions

What is the best prompt structure for Gemini? Google's own guide suggests four parts: persona, task, context and format. Add your source rules and a check step, and you have a prompt that holds up for marketing work.

Can Gemini read my Gmail and Drive? Yes, if you allow it. Mention the app in your prompt, for example @Drive or @Gmail, or connect apps under Customize. Workspace admins can switch these off.

How long can my files be? Google says the Gemini app accepts documents of up to 1,500 pages. Long files work best with a Pro-tier model and a clear question.