Gemini is Google's AI assistant, and you use it in the Gemini app on the web and on your phone, or inside Google Workspace apps such as Gmail and Docs. The way to get useful output is the same everywhere: give it the job, the context, the format you want back and one example, then ask it to ask you questions before it writes. This page covers where it lives, what it does well, five prompt patterns you can paste tonight and an honest comparison with ChatGPT; the free guide goes deeper with a mini-course.
Where Gemini lives
There are two front doors. The first is the Gemini app, in a browser and on a phone, where you chat, upload files and keep conversation history. The second is inside Workspace: a side panel in Gmail, Docs, Sheets, Slides and Drive that can read the document you have open and write into it.
The app is where I do thinking work: drafts, research, long files, planning. The Workspace panel is where I do finishing work: a reply in Gmail, a summary at the top of a Doc, a formula in Sheets, a title slide. Knowing which door to use saves more time than any prompt trick.
What Gemini is good at
Four jobs, in the order I reach for them.
Long documents. Paste or upload a contract, a transcript, a report or a set of meeting notes and ask questions against it. Gemini holds a lot of text in one conversation, and that one feature changed how I use it.
Anything already in Google. If your files live in Drive and your mail lives in Gmail, Gemini can reach them from inside the same account. "Find the proposal I sent in March and summarise the pricing section" is the kind of prompt that only works where the files already are.
Research with sources. Ask for a summary of a topic and Gemini can ground the answer in live search results and show where each claim came from. Check the links anyway. The model sounds confident when it is wrong, same as every other one.
Images and video. Gemini is the front door to Google's image and video models. The Nano Banana guide covers stills and edits. The Veo guide and the Gemini Omni guide cover clips.
Five prompt patterns that get a usable first draft
Each pattern is short enough to type from memory. Pick one per job.
1. The brief. Role, context, task, format, constraints, in that order.
You are a B2B copywriter. Context: I sell an onboarding tool to HR teams at
companies with 50 to 500 staff. Task: write a cold email for a first touch.
Format: under 120 words, one link, one question at the end. Constraints:
no buzzwords, no exclamation marks, plain English.
2. Ask me first. Add one line to any prompt: "Before you write, ask me the five questions whose answers would most change the result." The answers you give become the context you forgot to include.
3. File first, question second. Upload the document, then ask a narrow question, then a wider one. "What does clause 7 say about termination?" before "What in this contract would worry a small supplier?" The narrow question proves it read the file.
4. Match this voice. Paste two paragraphs you wrote and say: "Write the following in the same voice as the sample above. Match sentence length, vocabulary, punctuation and paragraph length." Then paste the draft. Gemini imitates well when it has a sample and goes generic when it does not.
5. Put it in a table. Any comparison gets clearer as columns. "Compare these four tools in a table with columns for price, free tier, best use and the reason I would say no." A table forces the model to fill every cell, and an empty cell shows you what it does not know.
The mistakes that make Gemini generic
Four, and I have made each of them this month.
One-line prompts. "Write a blog post about onboarding" gets you the average of every onboarding post ever written. Give it a reader, an angle, a length and one example.
No format. If you leave out "bullet points, under 200 words" you get prose, and if you leave out "prose" you get bullet points. Say which.
Asking for "the best". The best headline, the best plan. Ask for five and pick. The fifth is often the one.
Accepting the first draft. The first output is a draft, same as from a junior colleague. Reply with what is wrong with it. The second draft is where the value is.
Gemini vs ChatGPT for common jobs
I use both. The gap is small for everyday jobs and large for a few, and the few are about where your data lives.
| Job |
Where I start |
Why |
| Writing inside Gmail or Docs |
Gemini |
The panel is in the document already |
| Long file analysis |
Gemini |
Large context in one conversation |
| Saved specialist setups |
Either |
Gemini's saved setups and custom GPTs do the same job |
| Image generation |
Either |
Nano Banana and ChatGPT images both hold up; each has its own guide here |
| Video |
Gemini |
Veo and Omni sit inside Google's tools |
| A second opinion on a draft |
The other one |
Whichever model wrote the draft, the other one catches different problems |
For a longer list of starting points, the Gemini prompts in the library are sorted by job.
What is inside the Gemini guide
The free download has four parts.
- Gemini Prompt Engineering Guide: the patterns above in full, with worked examples for each.
- Veo 3 Guide plus Prompts: how to get video out of Google's video model, with prompts you can paste.
- Gemini Prompt Engineering Mini-Course: short lessons that turn the patterns into a habit.
- Lifetime access and updates: every revision as Gemini changes.
Drop your email in the form and it is yours.
