How to Actually Get Useful Work Out of Gemini

1. Set your custom instructions ONCE, before you do anything else

This is the five-minute fix that nine out of ten people skip. Custom instructions are the global memory that runs behind every chat: your role, your industry, how detailed you want answers, what tone works for you. If you don’t set them, you’re re-introducing yourself to Gemini at the start of every single conversation. That’s the productivity tax.

Go to Settings → Instructions for Gemini → click Add. This is where you tell Gemini who you are, what you do, and how you like responses. Think of it like setting preferences once so you don’t have to repeat yourself in every conversation. Tell it your role, your industry, how detailed you want answers, what tone works for you. Gemini applies this across all your chats.

Keep it short. Under 500 words. Focus on the 20% of preferences that cover 80% of your use cases. If you can’t read them in 90 seconds, they’re too long. Name your role, name your industry, name the format you want answers in (bullets vs. prose, length, tone), and name what you don’t want (no hedging, no preamble, no “as an AI…”). Save. Done. Every chat after this one is sharper by default.

2. Build a Gem for anything you do more than twice a week

If custom instructions are your global preferences, Gems are your specialists. They’re saved configurations, a name, a system prompt, optional knowledge files, and a default tool, that you open like a separate assistant. Gemini Gems are Google’s answer to ChatGPT’s custom instructions, letting you create personalized AI assistants tailored to specific tasks, communication styles, and contexts. Unlike ChatGPT’s single set of global instructions, Gems function as discrete AI personas you can switch between depending on your needs.

The reason to build one: any task you repeat, a weekly status email, a brand-voice editor, an interview-prep coach, a meeting-summary template, is a task you’re currently re-explaining to the model every time. Bake the explanation into a Gem once and never type it again. Free on every Gemini plan. Verified May 2026. There’s no excuse not to.

How to actually build one that works

Open the profile menu, click Gems, click New Gem. Fill in the builder, optionally let Gemini rewrite your Instructions via the wand icon, pick a Default tool (No default tool, Create image, Canvas, Deep Research, Create music, Guided Learning), attach Knowledge files, and Save.

A few rules that separate a Gem that works from a Gem you’ll abandon in a week:

• Don’t name it “Helper.” Name it for the job (Weekly Status Email Writer, Python Code Reviewer, Brand Voice Editor). You’re going to have five of these eventually, and you want to find them at a glance.

• Write the instructions like you’re briefing a contractor, not like you’re tagging keywords. Role Definition: Start with who the Gem is. Communication Style: How it should respond. Knowledge Focus: What it specializes in. Response Format: Preferred output structure. Constraints: What to avoid or emphasize.

• Pick the right default tool. A research Gem should default to Deep Research, a thumbnail Gem should default to Create image, a writing Gem should default to Canvas. The default tool is the one your Gem launches with on every new chat. Set it once and stop toggling. Default tool sets which Gemini tool the Gem launches with on every new chat. Options: No default tool, Create image, Canvas, Deep Research, Create music, Guided Learning. Pick the one your Gem usually uses (for example, a research Gem can default to Deep Research) so users do not have to enable it manually for every conversation.

• Steal from the premades first. Start with the Writing Editor or Brainstormer premade Gem. Click Make a copy, rewrite the instructions to match one repetitive task you do at least twice a week (a meeting summary template, a brand-voice editor, an interview-prep coach), then save. Skip the from-scratch build until you’ve copied and tweaked at least one premade Gem.

3. Use Deep Research instead of asking the chat box a complicated question

This is the single biggest leverage point in Gemini and the feature most people don’t even know exists. Regular chat is fast: you ask, you get an answer in seconds. Deep Research is the opposite. You ask, it builds a research plan, browses dozens of pages, and hands you back a structured, cited report. Think of Deep Research as a research assistant that works while you grab coffee. You give it a question, it creates a plan, browses dozens (sometimes hundreds) of websites, and writes up its findings in a proper report—complete with sections, analysis, and links to sources.

The right mental model: “Think of it as helping you go from zero to understanding a subject deeply,” Mukund agrees. If your question requires a fast, immediate answer then you probably don’t need Deep Research.

When to actually reach for it

Use Deep Research when the alternative would be twenty browser tabs and a half-finished doc. Comparing five vendors. Mapping a competitor landscape. Researching a regulation. Picking a summer camp. Anything where the work is the synthesis, not the lookup.

To launch it: On your computer, go to gemini.google.com. In the text box, click Add Files Deep Research. Write a real brief, not a one-liner. The most common mistake is treating Gemini Deep Research like a regular chatbot. Typing “tell me about the CRM market” produces a generic overview. Good research inputs are specific and scoped.

Three habits that turn Deep Research from “neat” to “indispensable”

• Edit the plan before it runs. Gemini shows you a structured outline of what it plans to investigate before it goes. Read it. Cross out the sub-topics that aren’t useful, add the ones it missed. Two minutes here saves twenty minutes of re-asking.
• Point it at YOUR data, not just the open web. By default, Gemini includes Google Search as a source for your research. You can change or add other sources, like your personal Gmail or Drive, for your research. You can also upload files and add NotebookLM notebooks. Add your own docs, your Gmail thread, your NotebookLM notebook, and watch the report stop being generic.
• Ask follow-ups instead of starting over. You can also use the space to ask Deep Research to add something new to the report after it’s been generated and it will adjust the report in real time. For example, you can ask something like “add camp cost details to my report” and Deep Research will add the information it finds.

4. Turn the report into something usable with Canvas

A 4,000-word Deep Research report is a starting point, not a deliverable. Canvas is the surface that turns it into one. Your interactive workspace inside Gemini. It opens a side-by-side editor where you can work on something and refine it with AI, all in one view.

The combo nobody talks about: run Deep Research, then have Canvas reformat the output. How do I turn Deep Research into a web page, infographic, quiz or audio overview in Canvas? Select Deep Research below the prompt bar. Your Deep Research report will be generated in a new Canvas. Once the research is complete, you will see a “Create” button at the top-right of the Canvas. Click on “Create,” and a drop-down menu will offer options for creating a web page, infographic, quiz, or audio overview. Simply select one of these options and watch Canvas bring it to life.

Translation: a research report becomes a one-page infographic, a quiz to test your team’s recall, a shareable web page, or a podcast you can listen to on a walk. That’s the actual workflow. Research, format, ship.

A trick worth learning inside Canvas itself: don’t type “rewrite this paragraph to be punchier.” At the bottom right of the Canvas panel, click the Change length and Change tone sliders for global adjustments to tone and length, or provide detailed chat feedback for more specific edits. To refine a specific section of text, highlight it and enter your instructions in one of two ways. You can type directly into the Ask Gemini box that appears next to your selection, or use the main prompt box to tell Gemini how you want that section rewritten. Highlight, instruct, move on.

5. Stop generating audio reports manually, let Audio Overviews do it

If you commute, walk, or wash dishes, this is the most underrated feature on the list. Audio Overviews turn any document, report, or Deep Research output into a podcast-style conversation between two AI hosts. Ever wish you could listen to a document or report while you’re on the go instead of reading it? We’re introducing the popular Audio Overviews feature to the Gemini app, similar to the capability you know and love in NotebookLM. Simply select “Create Audio Overview” after you upload documents or slides, or generate one directly from inside a Deep Research report you create with Gemini. These downloadable podcast-style audio discussions can offer new insights on your content and make it easier to learn on the go.

Use it for the dense PDF you’ll never sit down and read. The Deep Research report you ran last week and forgot to open. The board pack before a meeting. Pop in earbuds, hit go, and dead time turns into prep time.

To make one from a report you’ve already run: Go to gemini.google.com. If your chats are hidden, at the top, click Menu. Under “Recent,” find your research chat. Then On the right in the Canvas panel, click Create Audio Overview.

6. Use the Workspace sidebar, that’s where Gemini’s real advantage lives

If you live in Google Workspace and you’re still copying text out of Gmail to paste into Gemini’s chat box, you’re working harder than the tool requires. The sidebar in Gmail, Docs, Sheets, Slides, Drive, and Chat is where Gemini has context on what you’re actually doing. Gemini is Google’s flagship AI model, now integrated natively across the entire Google Workspace ecosystem. It is not a chatbot in a separate tab. It lives inside your existing apps and has direct context access to your emails, documents, calendars, and meetings. Think of it as the difference between a consultant you have to brief from scratch every time versus a team member who was in every meeting, read every document, and is already up to speed. The Gemini side panel in Gmail, Docs, Sheets, Slides, Drive, and Chat is the clearest expression of this philosophy: one click, instant AI assistance, full context from wherever you’re working.

The honest take: this is the part of Gemini that ChatGPT can’t replicate. Skip Gems entirely if your workflow lives outside Google Workspace and you already pay for ChatGPT Plus. Inside Workspace, Gems beat GPTs because they show up in the Docs and Gmail sidebars without leaving the app you’re already in.

A few of the highest-ROI moves inside the sidebar:

• Ask across your whole Drive. The “Ask Gemini” feature in Drive lets you query across your entire file library in natural language. Compare two vendor contracts and highlight the cost differences. Find everything related to a specific client project. Get a summary of a document you’ve never opened.

• Add events from Gmail in one click. It’s easy for you to add an event to your calendar right from your inbox with help from Gemini. Gemini detects emails that contain details about events and will display an “Add to Calendar” button. Just tap it to add the event to your Google Calendar.

• Summarize busy Chat threads. In busy team channels, Gemini summarizes long conversation threads so you can catch up without scrolling through hundreds of messages. Smart replies in Chat go beyond generic suggestions - Gemini reads the actual conversation and generates replies that match the content and context of what’s being discussed.

7. Set Scheduled Actions so Gemini works while you sleep

This is the feature that turns Gemini from “tool I open” into “assistant that just delivers.” Scheduled Actions are recurring prompts that run on a timer, without you sitting in the chat. Gemini can do things for you on a schedule, without you being in the chat. Scheduled Actions are recurring prompts that run automatically at a set time.

A short list of what’s worth scheduling:

• A Monday-morning industry-news briefing in your beat, delivered at 7:45 a.m.
• A Friday-afternoon retrospective: “Summarize this week’s meeting notes and flag anything I owe a follow-up on.”
• A pre-meeting prep pack the morning of every recurring call: pull the latest emails from the attendees, summarize the last meeting, suggest three questions to open with.
• A weekly competitor-watch run that re-runs a small Deep Research brief on a competitor and emails you only the diffs.

Pair this with a Gem (so the recurring prompt already has the right voice and format baked in) and you’ve got a system that produces real work product on its own.

The bonus habit that ties it all together

Stop treating Gemini’s tools as separate. The reason this guide is seven habits and not “one prompt to rule them all” is that Gemini’s value in 2026 is the chain: a Gem with the right system prompt, defaulting to Deep Research, pulling context from your Drive and Gmail, dumping the result into Canvas, exporting that to Slides, and reading you the executive summary as an Audio Overview on the drive home.

That’s the workflow. That’s what Google has been quietly building underneath you while you’ve been typing “summarize this” into the chat box. Start using it that way and Gemini stops feeling like the assistant you tolerate and starts feeling like the one you’d actually miss.

The one habit that beats all six others: assume Gemini already has a feature for whatever you’re about to do manually. Before you copy-paste, before you re-type your role, before you open another tab to research something, check whether there’s a Gem, a tool, a sidebar, or a scheduled action that does it for you. There almost always is. The people getting useful work out of Gemini aren’t smarter prompters. They just stopped using the chat box like it was 2023.