Google Gemini is active across almost every service you use, from Gmail and Docs to Google Search results. Yet most users only scratch the surface, treating it like a standard search engine instead of an agent to delegate tasks to.
By tweaking a few hidden settings and understanding the underlying models, you can turn Gemini from a basic text box into a tailored productivity assistant.
Key Takeaways
- LLMs are built on conversation roleplay; steering the persona of your prompts yields vastly better answers.
- Custom Instructions (under Personal Intelligence) apply global formatting and sourcing rules across all future chats.
- The 'Gems' feature allows you to build custom chatbots with specialized rules for recurring tasks.
- Scheduled Actions automate daily summaries, calendar briefs, and context reviews sent straight to your interface.
- Temporary Chats keep search history clean and prevent Google from using private conversations for future model training.
The Chaos Scenario: A Real-World Workflow
Imagine it’s 4:30 PM on a Friday. You are exhausted, your eyes are blurry, and you need to compile a checklist of action items from 20 different emails sent by your boss over the last 24 hours.
Instead of copy-pasting text for an hour and risking a missed deadline, you type a single instruction into Gemini: “List all action items from emails sent by my boss in the last 24 hours.”
Within seconds, the model scans your inbox, extracts the requirements, and outputs a formatted list. You close your laptop, start your weekend, and skip the manual sorting. That’s what happens when you make the AI work for you.
Audit Code and Large Files in the Context Window
With Gemini 1.5 and 2.0 Pro’s large context windows (supporting 1 million to 2 million tokens), you can upload entire codebase repositories or massive text documents.
To get the best analysis out of large files, wrap them in XML tags when uploading:
<source_code>
// Paste your index.ts file contents here
</source_code>
<styling_guide>
// Paste your eslint configurations here
</styling_guide>Then prompt: “Analyze the source code against the styling guide and list any styling violations.” This structured approach helps the model find details across hundreds of files without losing track of your instructions.
Build Specialized Custom Bots with Gems
If you want custom instructions for specific tasks but don’t want those rules active in every conversation, use Gems. Gems are custom versions of Gemini built for single roles.
You can configure separate Gems for tasks like:
- The Boss Editor: Rewrites email drafts to fit your manager’s short-sentence communication style.
- The Code Auditor: Loaded with your project’s styling guide, used solely to check files for bugs.
- The Language Tutor: Converses with you in a target language and corrects your grammar.
Keep Chats Clean and Private with Temporary Chats
By default, Google saves your chat history to train future iterations of its models. If you want to ask questions without having them logged or used for training, toggle the Temporary Chats icon in the top-right corner (represented by a pencil icon surrounded by a dotted line).
In this mode:
- Chats are kept for 72 hours for safety checks, then deleted permanently.
- Chats do not appear in your history sidebar.
- Google’s systems do not use these inputs for model training.
Even in temporary mode, avoid pasting sensitive files like database passwords, medical records, or government IDs. Keep private credentials offline.
Frame Your Requests with Expert Roleplay
Large language models predict the most likely words to follow a conversation template. Every chat begins with a system prompt that defines the model’s behavior.
If you ask a generic question like “what is sleep apnea,” Gemini relies on its general training data, which can occasionally lead to errors.
Instead, direct the model’s source targets by typing:
Who are the leading medical experts on sleep apnea, and what do they say about treatment options?By forcing the model to identify and report what specific experts say, you steer it away from making guesses and direct it toward summarizing legitimate research.
Enable Connected Apps for Personal Context
To make Gemini useful for daily logistics, enable Connected Apps in settings under Personal Intelligence.
This lets the assistant pull context from:
- Gmail: Summarize lengthy email chains.
- Google Calendar: Identify scheduling conflicts.
- Google Docs: Draft content based on research documents.
- YouTube: Locate specific tutorials and extract their step-by-step instructions.
Extract Text Instantly with Multi-Modal Image Input
One of Gemini’s most useful features is image transcription. If you have a screenshot of a design, a code block from a video, or an image that needs alt text:
- Upload the image directly into the chat.
- Type: “Transcribe the text in this image exactly, maintaining formatting.”
The OCR on Gemini 1.5/2.0 Pro is accurate and handles hand-written notes, complex code files, and tabular data.
Specify Trustworthy Source Citations
Gemini’s web-crawling capabilities are powerful, but the sources it pulls from aren’t always reliable. You can force the model to prioritize high-quality pages.
Add a rule to your prompts (or save it in your Custom Instructions):
Provide links to evidence. Prioritize primary research papers, Wikipedia articles, or official government documentation. Do not cite generic affiliate blogs or lifestyle websites.This forces the model to ignore SEO-optimized blog spam and look for legitimate primary documentation.
Let Scheduled Actions Automate Your Routine
If you run the same prompts every morning, you can automate them. Under the settings menu, look for Scheduled Actions.
This tool lets you schedule recurring prompts to run at specific times. Useful configurations include:
- Morning Briefs: Extracting key emails and calendar entries at 8:00 AM.
- Daily Weather Fits: Checking the local forecast and recommending what to wear.
- Competitor Tracking: Running a daily web search for news about specific companies.
Scheduled results are typically delivered within an hour of the scheduled time. For critical, time-sensitive alerts, manual prompts are still recommended.
Automate File Changes with Gemini Canvas
If you need to make edits to a long document or block of code, don’t let Gemini rewrite the entire file from scratch. Use Gemini Canvas.
Canvas provides a split-screen workspace where your text or code sits on one side, and the chat box sits on the other. You can highlight specific paragraphs or functions and ask Gemini to refactor only that selection, preventing the model from losing context or generating unnecessary boilerplate code.
Save Time with Custom Instructions
Tired of typing the same rules at the start of every chat? In Gemini’s settings, under Personal Intelligence, you’ll find Instructions for Gemini (previously called Saved Info).
This setting is a permanent system prompt overlay. You can use it to establish:
- Your job role and target writing style.
- Preferred code formatting rules (e.g., “Always write TypeScript with ES6 syntax”).
- Formatting constraints (e.g., “Never write paragraphs longer than three sentences”).
You can also use this area to create keyboard shortcuts. For example, add an instruction like: “When I type ‘H2s’, format the preceding text to use markdown H2 subheadings.”
Summary Checklist
- Audit Large Files: Use XML tags to structure code and docs in the context window.
- Use Gems for Siloed Work: Create custom chatbots for email drafts or code checks.
- Use Temporary Chats: Keep personal searches private and off training logs.
- Use Expert Roleplay: Structure prompts to ask what leading experts recommend.
- Link Connected Apps: Connect Gmail and Calendar for workspace context.
- Transcribe Images: Upload screenshots and extract text with multi-modal input.
- Command High-Quality Citations: Force links to primary research and official docs.
- Schedule Daily Tasks: Automate briefings and forecasts.
- Use Canvas: Edit code and text side-by-side without generating boilerplate.
- Save Custom Instructions: Set rules for formatting, code styles, and length constraints.
What to Read Next
- Google Keep: 15 Features for Power Users — Go beyond simple notes with checklists and reminders.
- Apple Notes: Master Your Workspace — Deep-dive tips for managing notes on macOS and iOS.
- VS Code Alternatives: Code Editors in 2026 — Find the best AI-integrated text editor for your development stack.
Frequently Asked Questions
Does Google train its AI on my personal emails?
Only if you have Connected Apps enabled and are using a personal account. If you are using a Google Workspace account for work or school, your data is kept private and is not used to train Google’s models.
How do I access Gemini Gems?
Gems are available to users subscribed to Gemini Advanced. You can find them by clicking on the Gemini settings menu and selecting “Create a Gem.”
What happens to Temporary Chats after 72 hours?
They are deleted from Google’s active database. The messages are temporarily cached for 3 days to monitor for system abuse, but they are never saved to your chat history or used to train Google’s AI.
Can Gemini transcribe handwriting?
Yes. As long as the handwriting is relatively clear, Gemini’s OCR engine can transcribe handwritten notes, whiteboards, and print scans into editable text files.
Related Articles
Deepen your understanding with these curated continuations.
15 Apple Notes Tips That Will Change How You Work
Discover the most useful Apple Notes features for iPhone, iPad, and Mac. From Quick Notes and Apple Intelligence to hidden tricks, boost your productivity with these tips.
15 Google Keep Tips That Will Change How You Work
Discover the most useful Google Keep features. From AI-powered lists and location reminders to OCR and Google Workspace integration, maximize your productivity with these tips.
Agent Skills with Google Gemini: Function Calling Guide
Complete guide to Gemini function calling — define tools, handle function_call responses, return results, and compare syntax with Claude and OpenAI. Node.js.
