Quick answer: Google Gemini is Google’s AI assistant that understands text, images, PDFs, audio and video. To use it well, give it the file or context, say exactly what you want back, and pick a format, such as a table, a list or JSON. Below I share the workflows, prompt formats and examples I actually use.
I’m Raj. For a long time I used Gemini like a search box. Type a question, read the answer, close the tab. That works, but it misses most of what the tool can do. The real value shows up when you start giving it your own stuff: a PDF you don’t have time to read, a screenshot of an error, a photo of a bill, a long meeting recording.
I came across a Gemini cheat sheet on Pinterest that sorted all of this by input type and prompt format. It got me thinking, so I tested the ideas myself. This post is my practical version, with my own examples, and checked against Google’s own pages.
What Gemini is (in plain words)
Gemini is the name of Google’s AI models and also the app you use at gemini.google. The key word you’ll hear is “multimodal.” It just means Gemini can take in more than text. You can upload images, documents, audio and video, and ask questions about them in one chat.
Developers can use the same models through the Gemini API. Google’s prompt design guide is written for developers, but honestly the advice works for normal users too: be clear, give examples, add context, and say what format you want.
What you can give Gemini (and what to ask for)
Here’s the simplest way I think about it. Match the input to the job.
Text
Notes, emails, articles, chat logs. Best for summaries, rewrites, plans and extracting details.
Example: Paste a messy 40-message email chain and ask: “List every decision and every open question. Put open questions first.”
Images
Screenshots, photos of documents, charts, product photos. Best for reading text from images, explaining a chart, or comparing two designs.
Example: Upload a photo of your electricity bill. Ask: “What am I being charged for? Which line went up compared to normal bills?”
PDFs and documents
Reports, research papers, contracts, user manuals. Best for summaries, Q&A and pulling data into tables.
Example: Upload a 30-page rental agreement. Ask: “List every clause about deposit, notice period and penalties, with page numbers.”
Audio
Voice notes, podcast clips, recorded calls. Best for transcripts, key points and action items.
Video
Tutorials, recorded webinars, product demos. Best for summaries and finding where something happens.
Example: “Summarise this 20-minute tutorial into 7 steps I can follow without watching.”
One honest note: exact file limits and which features you get depend on your plan and the app version. Always double-check important details, especially numbers from contracts or reports.
8 workflows I actually use
1. Quick research
I ask Gemini to explain a topic at three levels: simple, medium and expert. Then I pick the level I need. It’s great before writing a blog post on something new.
Prompt: “Explain what an index fund is in 3 levels: for a 12-year-old, for a college student, and for someone who already invests.”
2. Compare options
Comparisons work best when you ask for a table and give the criteria yourself.
Prompt: “Compare these 3 phones in a table: price, battery, camera, software updates. Then tell me which one fits a student who mostly watches videos.”
3. Extract data
This is my favourite time-saver. Upload invoices, receipts or a long PDF and pull out exactly what you need.
Input: 5 invoice screenshots.
Output asked for: “A table with vendor, date, amount and GST number. Write ‘not shown’ if any field is missing.”
That “not shown” line stops the AI from guessing. I add it to almost every extraction prompt.
4. Summaries that are useful
“Summarise this” gives you a bland paragraph. Instead, say who the summary is for.
Prompt: “Summarise this report for my boss who has 2 minutes. 5 bullets. Start with the biggest risk.”
5. Plans
Prompt: “I have 6 weeks and 1 hour a day to learn basic Excel for a job interview. Make a week-by-week plan with one practice task per day.”
6. Code help
Even if you’re not a developer, Gemini can explain a formula or write a small script. Paste the error message or screenshot and ask what went wrong in simple words.
7. Images
Gemini can create and edit images too. I wrote about this in my post on Gemini AI photo prompts, if you want ready-made ideas.
8. Turn a long video or recording into notes
Upload a recorded class or webinar and ask for: main points, examples used, and questions to review later.
Prompt formats that work (with examples)
This part of the cheat sheet was the most useful for me. Same question, different format, very different results.
Plain text
Good for quick, simple asks.
“Give me 10 blog title ideas about saving money as a student.”
Structured (JSON)
Useful when you want to paste the result into a sheet or app.
“Extract name, phone and city from these 20 contact cards. Return JSON with keys name, phone, city.”
You’ll get something like {"name": "Amit", "phone": "98xxxxxx10", "city": "Pune"} for each card.
With context
The more Gemini knows about your situation, the better the answer.
“I run a small bakery in Jaipur. Most customers order on WhatsApp. Write 5 short Diwali offer messages under 40 words each.”
With examples
Google’s own guide recommends giving examples, because the model copies the pattern. Show it one or two good outputs.
“Write product descriptions like this example: ‘Soft cotton kurta. Breathes in summer. Looks sharp at office.’ Now do the same for a linen shirt and a denim jacket.”
System role
Tell Gemini who to be and how to behave for the whole chat.
“For this chat, act as a strict editor. Point out weak sentences, but don’t rewrite my whole post.”
If you use Gems in the Gemini app, you can save these instructions so you don’t repeat them every time. It’s a similar idea to the custom skills I wrote about in my Claude skills ideas post.
Multimodal
Mix a file and clear text instructions.
[Upload a photo of a handwritten to-do list] “Type this out, sort tasks by urgency, and estimate time for each.”
Real use cases by type of person
- Students: Upload lecture slides, ask for a summary and 10 quiz questions.
- Bloggers: Research a topic, build an outline, and check that the draft answers common questions.
- Small business owners: Pull data from invoices, draft customer replies, plan a festival offer.
- Job seekers: Paste a job description and your CV, ask which skills are missing.
- Parents: Snap a photo of a homework question and ask Gemini to explain the method, not just the answer.
Mistakes I made early on
- Being vague. “Help me with marketing” gets you a generic essay. “Give me 5 Instagram post ideas for a home tuition centre in Lucknow” gets you something usable.
- Not asking for a format. Always say: table, bullets, steps, JSON or word limit.
- Trusting every number. AI can still get facts wrong. Check anything important against the original source.
- Uploading private stuff without thinking. Don’t upload documents with passwords, bank details or other people’s private data unless you understand the privacy settings.
If you write with Gemini, also keep your own voice. My guide on making AI content sound human has simple editing rules that work with any AI tool.
A full example from start to finish
Last month a friend sent me a 45-page government scheme PDF and asked if her small shop qualified. Here’s exactly what I did. First, I uploaded the PDF and asked: “Who is eligible for this scheme? List the conditions as a checklist.” Then I gave context: “The shop is in Maharashtra, run by one woman, yearly turnover around ₹12 lakh.” Then I asked: “Based on the checklist, which conditions does she meet, which does she not, and which are unclear? Quote the page for each.” In ten minutes we had a clear answer, plus two questions to ask the bank. I still checked those pages myself, but it saved a whole evening of reading.
FAQ
Is Gemini free?
There’s a free version of the Gemini app. Paid Google AI plans add access to more advanced models and higher limits. Check Google’s current plan page, as features change often.
Can Gemini read PDFs and images?
Yes. You can upload documents and images and ask questions about them. Accuracy depends on file quality, so blurry scans give weaker results.
What’s the best prompt format for beginners?
Start with “context + task + format.” Example: “I’m a teacher (context). Make a 10-question quiz on fractions (task). Use multiple choice with answers at the end (format).”
Is Gemini better than ChatGPT or Claude?
Each has strengths. Gemini fits nicely if you already use Google apps a lot. I’d suggest trying the same task on each and keeping what works for you.
Can I use Gemini output on my blog?
Yes, but edit it. Add your own experience and check facts. Google cares about helpful, people-first content, not whether a tool helped you write it.
Final thoughts
Gemini isn’t magic, but it’s a very good assistant once you stop treating it like a search box. Give it your files, give it context, and tell it what shape you want the answer in. Start with one workflow this week, maybe extracting data from receipts or summarising a long PDF, and build from there.