Gemini

2.1 Complete a Gemini research task

The model selector at the top of the Gemini chat window is where you pick which model answers your prompt — 3.5 Flash is the free-tier default for everyday questions, and 3.1 Pro is the heavier option you switch to mid-conversation when a task needs deeper reasoning.

Open Gemini with a Google account, use the free tier and run a real research prompt

Just type it into the prompt bar at gemini.google.com — no setup, no API key. Gemini answers in whatever language you write in.

1. Open gemini.google.com and sign in with any Google account
2. Enter a real research prompt into the conversation input box
3. Select 3.1 Pro from the model selector when deeper reasoning is required

• You'll see A plain‑language answer that includes a structured response and three discussion questions
• Takeaway The free tier can handle whole tasks in natural language; switch to 3.1 Pro only for hard reasoning
• Check Which model did you switch to partway through the conversation, and why did you wait to switch instead of starting with it?
• Cost Free. The Gemini free tier includes 3.5 Flash, limited 3.1 Pro, and the major features (Deep Research, Live, Canvas, Gems). Paid Google AI plans raise usage limits and unlock the newest models — covered in later lessons.

2.2 Craft targeted Gemini responses

The prompt itself is the control surface here: giving Gemini a role, an audience, and an output format in the same message shapes the answer before it is generated, and refining in the same thread lets you correct one part without re-explaining the whole task.

Do this first Complete a Gemini research task

Generate sharper answers by assigning a role, audience and output format then refining within the same conversation

Give it a role (grant reviewer), the material (your summary), the task (critique + rewrite), and a format constraint (under 150 words).

1. SET the role and audience for Gemini, e.g., “You are a grant reviewer” or “Explain this to a first‑year master’s student”
2. SPECIFY the exact output format you need, such as a table, five bullet points, a 150‑word abstract, or a numbered protocol
3. ITERATE in the same thread by asking for refinements like “make point 2 more specific” or “turn this into a table”

• You'll see The reply includes both a generic improvement and a reviewer‑grade critique formatted exactly as requested
• Takeaway Define who Gemini is, what it sees and how to answer, then refine without restarting
• Check What three things did you specify in your prompt that turned Gemini's answer from generic advice into a reviewer-grade critique?
• Cost Free. Prompt-craft costs nothing and works identically on every plan — it's the single biggest lever on answer quality.

2.3 Ask Gemini questions using your own files and Google Workspace

The Upload button (the paperclip in the prompt bar) lets Gemini read a file directly rather than you pasting its contents into the chat, while the separate Connect Google Workspace step is what lets it search your own Gmail, Drive and Docs instead of only the file you attached.

Do this first Craft targeted Gemini responses

Upload a PDF, image or other file and ask Gemini questions that draw on your Gmail, Drive and Docs

Click the upload button (the + / paperclip in the prompt bar) and attach the file before sending. Gemini reads the document itself — you don't paste its contents.

1. Click Upload and select the PDF, image or file you want Gemini to read
2. Press Connect Google Workspace and authorise access to Gmail, Drive and Docs
3. Ask Gemini a question and add “and tell me which page/section it’s on” to receive source locations

• You'll see A concise summary of the uploaded document together with a ranked list of its most‑cited references and links to the exact locations in your files
• Takeaway Upload the real artifact and connect Workspace so Gemini answers from your own documents and email rather than just from memory
• Check What phrase did you add to your prompt to make Gemini report which page or section a finding came from, and why does that matter for checking its work?
• Cost Free tier supports file upload and Workspace/Photos connections. Higher usage limits and larger document handling come with paid Google AI plans.

2.4 Get web‑sourced answers from Gemini

The arrow chip at the end of a grounded Gemini answer opens the "Related content and sources" panel — it is the mechanism that turns a claim into something traceable back to an actual web page rather than the model's own paraphrase.

Do this first Ask Gemini questions using your own files and Google Workspace

Receive up‑to‑date facts with clickable source links you can verify yourself

Ask for recent, factual information and explicitly request sources. Some Gemini responses are grounded on live Google Search results.

1. Ask Gemini for current information and explicitly cite your sources in the prompt
2. Click the arrow chip at the end of the response to open the “Related content and sources” panel
3. Open each listed link in a new tab to read the original page
4. Paste a specific claim into a fresh Gemini query and request verification with the exact source sentence

• You'll see An answer that includes live‑web citations and an arrow chip revealing a related content panel
• Takeaway Validate the information by checking the cited pages rather than trusting the model’s summary
• Check After opening a source card in the related-content panel, what should you check on the original page before trusting the claim Gemini made about it?
• Cost Free. Search grounding is available on the free tier; it's about how you check answers, not which plan you're on.

2.5 Create and analyse images with Gemini

The Describe your image box on the image-generation screen turns a text prompt into a picture, and the same screen accepts an uploaded file for Gemini to analyse instead — the Create image chip is what tells it to render an image rather than reply with text.

Do this first Get web‑sourced answers from Gemini

Generate a diagram from text and interpret an uploaded picture using Gemini

Type the request; Gemini generates the image inline. Image generation is available on the free tier.

1. Enter a description in the Describe your image prompt box and click Create image to generate a picture
2. Drag or select an image file onto the upload area to provide Gemini with a picture to analyse
3. Ask Gemini to describe the uploaded image, read any embedded text, or flag anomalies
4. Specify a local edit (e.g., “relabel the third arrow ‘translation’ and make the background white”) and click Create image again to apply the modification

• You'll see A labelled diagram produced from one sentence and a textual description of the image you uploaded
• Takeaway Gemini can both synthesise visual content and read real‑world images while preserving existing elements during local edits
• Check Why should a local edit request name only the one detail that's wrong, rather than asking Gemini to regenerate the whole diagram from scratch?
• Cost Image generation and editing are on the free tier (downloads at 1K resolution). Google AI plans add 2K-resolution downloads and the higher-detail Nano Banana Pro regeneration for text-heavy images and infographics.

2.6 Create a reusable assistant and edit drafts side‑by‑side

A Gem is a saved persona with standing instructions (and optionally reference files) that you reopen instead of retyping a prompt; Canvas is a separate side-by-side editor for drafting and revising a document section by section rather than through chat replies.

Do this first Create and analyse images with Gemini

Build a Gem for a recurring task and use Canvas to write and revise longer pieces in a live editor

Open the Gems area, choose 'create your own', paste these instructions, and optionally upload reference files. Then chat with the Gem whenever you need that job done.

1. Open the + button in the sidebar to launch the Gem manager
2. Select a premade Gem such as Brainstormer, then click Create new Gem to set standing instructions and upload reference files
3. Choose Canvas from the main menu to open the side‑by‑side editor
4. Paste or import your draft in Canvas and press Revise section to have Gemini edit the selected part live

• You'll see A Methods Editor Gem you can reopen at any time and a Canvas document that Gemini updates section by section
• Takeaway Encode a repeated job once as a Gem and treat Canvas as a full‑featured editor for anything beyond a chat message
• Check What problem does saving a recurring task as a Gem solve that simply keeping a saved chat prompt does not?
• Cost Gems and Canvas are both available on the free tier. Some advanced capabilities (e.g. Canvas with the 3.x model and a 1-million-token context for large projects) are reserved for Google AI Pro and Ultra subscribers.

2.7 Generate a written overview and bar chart from CSV data in Sheets

The Gemini sidebar is the same assistant, docked inside a Google Workspace app instead of its own tab. Inside Sheets it can read the range you have selected — so the selection, not the prompt, is what tells it which numbers to describe, and what a chart it drafts is drawn from.

Do this first Create a reusable assistant and edit drafts side‑by‑side

Create a text summary and an editable bar chart directly from a CSV file opened in Google Sheets

1. Open the CSV file in Google Sheets
2. Select the range containing the data you want summarised
3. Type a prompt into the Gemini sidebar requesting a description of table trends and a draft bar chart
4. Press Enter to submit the request
5. Click Insert chart on the generated result to add the chart image to the sheet

• You'll see A paragraph summarising the data patterns appears alongside an inserted bar‑chart that matches those trends
• Takeaway Gemini can turn tabular data into combined textual insight and visual draft
• Check What do you have to select in the sheet before typing your prompt into the Gemini sidebar, and why does that selection matter for the summary it writes?

2.8 Run a saved Gemini assistant with a single click

A Gem is a saved prompt with a name. Everything you would otherwise retype — the instructions, the format, the role — is stored once and run from an icon, which is what turns a good one-off prompt into something you use every week.

Do this first Generate a written overview and bar chart from CSV data in Sheets

Execute your recurring Gemini workflow without re‑typing the prompt

1. Open the Gemini sidebar in any Google Workspace app by selecting Gemini from the right‑hand panel
2. Enter your full prompt for the recurring task in the input box
3. Save the prompt as a custom assistant by clicking Save as Gem
4. Provide a descriptive name in the Gem name field and confirm with Save
5. Run the workflow later by clicking the saved Gem icon

• You'll see The assistant runs automatically and displays its result, such as a side‑by‑side comparison table, after one click
• Takeaway Custom Gemini assistants turn complex prompts into reusable tools that streamline repetitive workflows across Workspace
• Check What are the two steps that turn a one-off Gemini prompt into a Gem you can rerun from the Workspace sidebar with a single click?

2.9 Generate a research report and reshape it in Canvas

Deep Research is a Gemini mode that first drafts a multi-step research plan for your approval, then browses and synthesises sources on its own — it behaves less like a chat reply and more like an analyst you brief, watch work, and then check.

Do this first Run a saved Gemini assistant with a single click

A complete, sourced research report is created, approved, and exported for further formatting

Select 'Deep Research' in the prompt bar, then enter the query. Deep Research is available on the free tier.

12 Enlarge

1. 1 button approve personalized research plan
2. 2 tab view web search results live

Best viewed on desktop — tap Enlarge to read the numbered controls.

1. Open Deep Research, review the personalised research plan, refine the scope, and approve it
2. Watch the agent perform web, Gmail, Drive and Chat searches and analysis in real time
3. Send the finished report to Canvas and convert it into a slide‑ready infographic, interactive quiz or one‑page brief

• You'll see An approved plan, a live view of the agent searching and analysing, and a multi‑section report ready to be turned into a brief or infographic in Canvas
• Takeaway Direct the analyst‑like Deep Research by approving its plan, letting it browse, then transform the output into a usable deliverable
• Check Why is reviewing and amending the research plan before Deep Research starts searching the most effective point to control the quality of the final report?
• Cost Deep Research is available on the free tier; paid Google AI plans give higher usage limits (more reports, more often) and early access to advanced reasoning like Deep Think on Ultra.