How to Use Gemini in 2026: Google's AI Step by Step

Here's how it usually goes. Someone has had a Gmail address for ten years. Their pictures sit in Google Photos, their work spreadsheets in Google Sheets. For the past year a colorful little star labeled Gemini has been blinking in the corner of every one of those products. They see it every day and have never clicked it. They don't know what it is, whether it costs money, or how it differs from ChatGPT, which everyone around them keeps talking about.

Behind that star sits one of the three strongest AI models in the world. It can watch an hour-long webinar recording and hand you notes with timestamps, read ten PDFs at once, draw an illustration for your post, and put together a market overview from dozens of websites. By the way, the cover images on the article cards on this blog's homepage are drawn by Gemini.

There's one caveat I'll name up front. Google rolls features out unevenly: by country, by age, and by plan. Something your colleague already has may reach your account a month later. I mark those spots as we go, so you know when a missing button is Google's schedule and nothing you did wrong.

Inside you'll find a step-by-step first chat, 22 ready-to-use prompts for writing, files, images, and research, a table of plans, and a comparison with ChatGPT and Claude. Read it in order or jump through the table of contents straight to the section you need.

Section 01What Gemini is in 2026

Gemini is a chat assistant made by Google. You type a task in plain words, and it answers with text, an image, a table, or a finished document. It belongs to the same class of tools as ChatGPT and Claude, only built by Google and wired into Google's products.

A little history so the names don't trip you up. In March 2023 Google released a chatbot called Bard. In February 2024 Bard was renamed Gemini, and since then that's the name of both the chat and the family of models inside it. If you run into an old tutorial about Bard, it's the same assistant at a younger age.

Several models run inside Gemini, and you choose between them in the dropdown above the input box:

• Gemini 3.6 Flash – the fast everyday model. It came out on July 21, 2026 and is open to everyone, including free users. Google describes it as an all-rounder: daily tasks, document analysis, multi-step projects.
• Gemini 3.1 Pro – the model for complex reasoning, released on February 19, 2026. Free access to it is limited, and subscribers get higher limits.
• Gemini 3.8 Flash – the newest version of the fast model. AI Pro and AI Ultra subscribers received it in early September 2026.
• Deep Think – an extended-thinking mode for the hardest problems. An answer takes several minutes to prepare, and the mode is available to AI Ultra subscribers.

Which model should a beginner choose? Leave the default. The fast Flash model handles posts, emails, notes, and ideas. Switch to Pro when a task needs calculations, a comparison of many conditions, or a long chain of steps, and the Flash answer didn't satisfy you.

Media models work separately from the text models. Images are created and edited by a model with a funny name, Nano Banana. The current versions are Nano Banana 2 and Nano Banana Pro. Video is made by Gemini Omni, which arrived in the app on May 19, 2026 for subscribers.

Section 02Where Gemini is available

I put this section near the top on purpose. It's frustrating to read a guide for twenty minutes and then find out a feature isn't in your account yet.

Here are the facts as of October 1, 2026.

• The list is long. Google's help pages say the Gemini web app is available in more than 230 countries and territories and in more than 70 languages.
• The feature set varies. Which tools you see and when they arrive depends on your country, your age, and your plan. Video creation, Skills, and some app features roll out in stages.
• Travel can interrupt access. If you take a trip to a country outside Google's list, Gemini may refuse to open there, on the website and in the app alike.
• A VPN is an unreliable fix. Google works out your location from a combination of signals, including account history, location in Maps, and payment details. Changing your IP address alone often changes nothing.

What follows from this in practice? If you live in a country on the list, everything works as designed: local internet, a Google account, and you're in. Before a long trip, check Google's list of supported countries and save anything you'll need offline: export the important chats to Google Docs and download the files.

From here on I explain how Gemini is built and how to work with it. Follow the steps in order. And if a feature hasn't reached your account yet, read the prompt sections anyway: the prompts work in any AI chat.

Section 03Where to open Gemini: 5 ways in

Gemini has several doors, and different ones suit different tasks.

Which door should you start with? I'd begin with the website on a computer. Copying text, uploading files, and reading a long answer in full are all easier there. Add the phone as your second step, for the voice mode.

Section 04What Gemini can do: 12 features in one table

Below is everything Gemini can do as of early October 2026 and how to put it to work. The last column shows the plan where each feature starts.

Google doesn't publish exact numbers of prompts per day. The help pages only say that limits are "standard" without a subscription and grow on paid plans. You can see what you have left in the app settings.

Section 05Your first chat, step by step

Let's walk through a first conversation with Gemini, from an empty screen to a result you can take into your work. The whole thing takes about fifteen minutes.

Here's roughly what that conversation looks like:

Notice the order: context first, then a list of ideas, then one idea goes into production. In three messages you get a post written about your client.

The working formula for a prompt is simple, and it's the same for every AI: who you are + who the result is for + what to do + in what format + an example or a constraint. I broke it down in detail in how to write prompts for AI: a 5-part formula. Every prompt below is built on it. Copy them and change the details in square brackets to fit your work.

Section 06Writing for work: 8 ready-to-use prompts

1. The introduction that saves an hour on every task

Remember this context for this chat. I'm a [nutritionist] and I work with [women aged 35–50 who are tired of diets]. I sell [an 8-week coaching program for $300]. I write warmly, in plain language, with no medical jargon and no pressure. When I ask for a text, ask me up to three clarifying questions if you're missing information. Reply with one word: "Got it."

Why this works: from here on you can write short messages in this chat. You type "write a post about late-night snacking," and Gemini already knows who it's for and in what voice.

2. A post for your newsletter or social feed

Write a post about [why it's hard to get back to workouts after a vacation]. Length: 200–250 words. Open with a scene the reader will recognize from their own life, give 3 practical steps, and end with a question for the comments. Short paragraphs, no corporate language, and no emoji in every sentence.

If the text comes out smooth and faceless, add: "Add one specific detail from real life and cut the general phrases that would fit any topic."

3. A reply to "that's too expensive"

A client sent me a DM: "[I like the program, but $300 is a lot for me right now]." Write 2 versions of a reply: calm, with no discount and no pressure. Show what the person gets for the money and offer an easy next step. Each version up to 60 words.

4. A service description for your website or profile

Help me describe my service: [a résumé review and interview prep]. Who it's for: [professionals who haven't landed an offer in six months]. The client's result: [a résumé that gets replies and a confident story about themselves]. Format: [2 one-hour calls + edits in a shared document]. Write a description of 100–130 words: what the person comes in with, what they get, how the work goes, and what to do to sign up.

5. The email that's awkward to write yourself

Write an email to a client who [has postponed payment for the third time]. The tone is respectful and firm. Remind them of our agreement, name the deadline [by Friday], and offer two options: pay now or move their start to the next group. Up to 110 words, no apologies and no threats.

6. Ten headlines in place of one

Here is the text of my post: [paste the text]. Suggest 10 headlines using different approaches: a number, a question, a story, an unexpected fact, a promised result. Mark the three that will grab [parents of school-age kids] the hardest and explain each pick in one line.

7. Cutting a text without losing the point

Cut this text in half: [paste the text]. Keep all the facts, the numbers, and the call to action at the end. Remove repetition, filler words, and explanations of the obvious. After the text, list exactly what you removed.

The last line is the important one here. You see right away whether Gemini threw out something you need, and you bring it back with one sentence.

8. A review of your text through a stranger's eyes

Read my text through the eyes of [a business owner who is hearing about me for the first time]: [paste the text]. Name 5 places where the reader gets bored, gets confused, or wants to leave. For each place, suggest a specific replacement. Do not rewrite the whole text.

Section 07Files, video, and audio: 4 prompts

This is where Gemini is especially strong. It was built from the start to understand text, images, sound, and video together. You upload a recording, and the AI works with it directly.

The upload rules from Google's help pages:

• you can add up to 10 files to one prompt;
• a document, spreadsheet, or image can be up to 100 MB per file;
• a video file can be up to 2 GB;
• without a subscription Gemini handles video up to 5 minutes long and audio up to 10 minutes;
• on AI Pro and AI Ultra the limit grows to 1 hour of video and 3 hours of audio;
• you can connect a code folder or a GitHub repository: up to 5,000 files and up to 100 MB.

You add files with the plus or paperclip button next to the input box. Google Drive connects from the same place, so you don't have to download the document to your computer first.

9. Notes from a webinar or a call

This is a recording of [my webinar on choosing a niche]. Make notes: 1) the main idea in two sentences; 2) 7–10 key points with timestamps; 3) every question from the audience and my answers; 4) three post ideas based on the webinar. Write in plain language, with no commentary of your own.

Why this works: you're asking for four different results from one recording. A single webinar turns into notes for attendees, answers to common questions, and topics for the week ahead. I covered other ways to turn a recording into text in AI transcription: a summary of your webinar in 5 minutes.

10. The short version of a long document

Read this document: [a 14-page lease agreement]. Summarize it in 10 points. In a separate list write out: every amount and deadline, my obligations, the penalties, and the termination terms. If any wording is ambiguous, quote it and explain both readings.

11. Comparing several files

I've uploaded [three proposals from contractors]. Put them in a table: price, timeline, what's included, what's billed separately, guarantees. Under the table, write the questions I should ask each contractor before signing.

12. A table from a photo or a screenshot

The photo shows [a notebook page with a week of client records]. Move the data into a table: date, name, service, amount. If you're unsure how a word reads, put a question mark next to it. Add up the total for the week.

The request for a question mark protects you from quiet mistakes. Without it the AI confidently "reads" an illegible word by guessing, and you find out too late.

Section 08Images and video: 5 prompts

Images in Gemini are created and edited by the Nano Banana model. It's built right into the chat: you describe in words what you want to see, or you upload your own photo and ask for changes. Nano Banana's strength is editing. It does a good job of keeping the face and the object in a photo while you change the background, the clothes, or the lighting.

The formula for an image: what's in the foreground + where it happens + the light + the style + the aspect ratio.

13. An illustration for a post

Draw an illustration for a post about [the fear of a first call with a client]. In the foreground: [a laptop on a kitchen table and a cup of tea gone cold]. Morning light from a window, calm warm tones, realistic photo style. No people and no text. Horizontal frame, 16:9.

14. A new background for a product photo or a portrait

This is a photo of me [at my desk]. Keep my face, hair, and clothes unchanged. Replace the background with [a bright office with a bookshelf] and add soft light from the left. The result should look like an ordinary photograph.

You need the phrase "keep my face unchanged" in every prompt that uses your portrait. Without it the model may slightly "improve" your features, and the picture ends up showing someone who only resembles you.

15. A cover with text on it

Make a cover for [a guide called "7 Résumé Mistakes"]. Large text "7 Résumé Mistakes" in the center, even letters, easy to read on a phone. A solid [dark green] background and one simple object: [a sheet of paper with a paperclip]. Vertical frame, 4:5.

Always check the text letter by letter. The model handles short phrases well, and typos show up in long lines. If a letter slipped, ask: "Fix the text. It has to read exactly like this: …"

16. A series in one style

Draw 4 illustrations for a carousel about [morning habits]: [a glass of water, sneakers by the door, an open notebook, an alarm clock]. One style for all of them: flat illustration, three colors [beige, navy, mustard], lots of white space. Square, 1:1.

17. A short clip in Gemini Omni

Make an 8-second clip for Stories: [a cup of coffee on a table, steam rising, the camera slowly moves in, rain starts outside the window]. Vertical 9:16 format, a calm mood, no text on screen.

Video creation is available to subscribers starting with the AI Plus plan. Gemini Omni can build a clip from text, photos, and your own videos, edit it right in the chat, and create an avatar. The feature set depends on your country.

Google puts an invisible digital watermark called SynthID on every image and video. You can't see it, but Google's products use it to tell that an image was made by AI. For more on choosing an image tool, read AI image generators in 2026: which to choose, and for clips, AI video from text.

Section 09Deep Research and Gemini Notebook

A regular chat answers from what the model learned and from a quick search. Deep Research works differently. You ask a question, and Gemini drafts a research plan and shows it to you. After you say yes, it visits dozens of websites on its own, reads them, and a few minutes later brings back a multi-page report with links to the sources. You can open the report in Google Docs with one click.

You turn the mode on in the tools menu under the input box. The free version has it with a small monthly limit, and paid plans give you more reports.

18. A market overview for your niche

Run a research project: the market for [online SAT math prep courses] in the US in 2026. I need: average prices and formats, 10 notable players with prices and links, how they differ from each other, and what parents complain about in reviews. End with 5 takeaways for [a private tutor who wants to launch a small group].

19. A competitor review from a list

Study the websites and public social profiles of these specialists: [link 1, link 2, link 3]. Put it in a table: what they sell, at what price, the main promise at the top of the page, and what proof they show. Separately, write what none of them has.

20. A check on a bold claim

Check this claim: "[80% of people quit online courses in the second week]." Find the original source of the number, the year, and the research method. If there is no reliable source, say so directly and give data I can trust, with links.

This prompt is handy for cleaning up your own texts before you publish. Half of the "well-known" numbers on the internet turn out to have no source once you check.

Gemini Notebook solves the opposite problem. You put your own sources into a notebook: lesson transcripts, workbooks, articles, PDFs. Gemini answers only from them and shows where each answer came from. A separate feature, Audio Overviews, turns the materials into an audio summary you can listen to like a podcast on the road. The notebook is in the free version. AI Plus gives you more audio overviews, and AI Pro gives you five times more.

How do you use it? Collect all the materials from your course in a notebook and ask questions as a client would: "where do I start if I have no audience?" You'll quickly see which questions your materials answer and where there's a hole.

Section 10Gemini Live, Canvas, Gems, and Skills

Gemini Live: a conversation out loud

The phone app has a Live button. You tap it and talk to Gemini out loud, like on a phone call. You can interrupt it mid-sentence, ask it to keep things shorter, or go back to an earlier thought. You can turn on the camera and show what's in front of you, or share your phone screen. The mode is free. Since May 2026 you can switch between voice and text inside one chat.

21. A voice brainstorm

Let's brainstorm. I'm a [therapist] and I want to launch [a mini course on anxiety before public speaking]. Ask me one question at a time: who my client is, what hurts, what they've already tried. After ten questions, turn my answers into an outline for a 5-lesson course.

This prompt is easy to say out loud on a walk. You answer the questions by voice, and at home the same chat holds a finished outline that only needs polishing.

Canvas: a window for a big document

When a text is long, editing it in the chat feed is awkward. Canvas opens a separate window next to the chat with your document in it. You select a paragraph and ask for a rewrite of that part only, change the tone and length with buttons, and the chat stays on the left for discussion. In Canvas Gemini can also build simple web pages, mini apps, and prototypes: you describe a price calculator for your service in words and see it working right away.

Gems and Skills: helpers for your tasks

A Gem is a saved version of Gemini with your instructions. You describe once who you are and how to write, attach examples, and from then on you open that helper in one click with no need to repeat the background.

On September 30, 2026 Google announced the launch of Skills. A Skill is a reusable set of instructions for a task you repeat. The release notes say Skills will replace Gems. The feature is gradually reaching users over 18, including free ones. If you only see Gems for now, use them: the logic is the same.

22. Instructions for your own helper

You are the editor of my posts. I'm a [career consultant] and I write for [professionals aged 28–40 who want to change jobs]. My style: direct "you," short paragraphs, examples from my practice, no corporate language. When I send a draft: 1) fix the mistakes; 2) suggest a stronger first sentence; 3) mark the places where the reader gets bored; 4) return the full text. Keep my wording and do not add new facts.

Paste this text into the instructions field when you create a Gem or a Skill and attach three to five of your best posts as samples. Now every draft goes through the same edit in a minute.

Section 11Free or paid: 4 plans

Google has four tiers. The prices below are for the US as of October 1, 2026. In Europe they differ slightly: AI Pro, for example, costs €21.99.

What does "chat memory" mean in this table? It's the amount of text Gemini keeps in its head within one conversation. 32K tokens is a few dozen pages. Google equates a million tokens to roughly 1,500 pages. On the free plan, in a long chat with big files, the AI starts to forget how the conversation began. On AI Pro you can upload an entire book and discuss it chapter by chapter.

How do you choose?

• Free – for getting acquainted and for short daily tasks: posts, emails, ideas, images. It's enough to find out whether this is your tool.
• AI Plus – if you keep hitting the limits and want Gemini inside your inbox. For the price of a cup of coffee you also get 400 GB in Drive.
• AI Pro – if you work with hour-long recordings, large documents, and long projects. This is the main working plan.
• AI Ultra – for people who sit in Gemini all day and run into the AI Pro limits. A beginner doesn't need it.

Before you pay, live on the free version for at least two weeks. If you never hit a limit in that time, you have no use for a subscription yet. Other free options are collected in free AI tools 2026: what works without paying.

Section 12Gemini, ChatGPT, and Claude: which to pick

The three leading AI assistants of 2026 solve the same tasks, and each has a place where it beats the others. Here's a comparison from the point of view of someone who sells their own services and runs a blog.

What's the takeaway? I'd reach for Gemini for three jobs: working through video and audio, working inside Google's products, and a low-cost way into paid features. For long texts and big projects I open Claude more often. ChatGPT remains the most all-purpose option, with a huge number of tutorials online.

The detailed breakdowns are in neighboring articles: Claude 5, GPT-5, and Gemini 3: what changed and which one fits you, Claude vs ChatGPT, and the big roundup best AI tools 2026 for 12 real work tasks. If you're still unsure, the 5-minute scheme for choosing an AI tool will narrow it down.

Section 13Limits and privacy

Gemini is a strong tool, and it has weak spots you're better off knowing in advance.

• It makes up facts. Like any AI, Gemini can confidently name a study that doesn't exist or mix up a number or a date. Check everything that goes to clients or into print.
• It plays it safe. On health, legal, and financial topics Gemini sometimes answers in generalities or declines. A clarification helps: "I understand this doesn't replace a professional. I need an overview so I can prepare my questions."
• Free memory is short. 32K tokens run out fast when big files sit in the chat. The sign is simple: Gemini starts mixing up what you agreed on at the beginning.
• Features arrive unevenly. Google rolls out new things gradually and country by country. What a friend already has may reach you a month later.
• Access depends on your country. I covered the details in section 2.

Now about your data. In its privacy notice for Gemini, Google directly asks you not to enter confidential information you wouldn't want a reviewer to see: people read some conversations to improve the service. By default chats are saved in your account.

What to do about it:

• open Gemini's activity settings and decide whether to keep your chat history and whether it can be used to train the models;
• for one-off sensitive questions turn on a temporary chat: it doesn't go into your history;
• don't upload ID documents, passwords, medical records, or clients' personal information without their consent;
• before you upload a contract or a message thread, swap the names and amounts for placeholders: Gemini will get the meaning, and nothing extra lands in the service.

Section 14When something goes wrong

Section 15Your first-week checklist

To make Gemini a habit, give it one task a day. The plan below takes 15–20 minutes a day.

• Day 1. Open gemini.google.com, introduce yourself with prompt 1, and ask for 10 post ideas.
• Day 2. Take one idea, write a post with prompt 2, and make three edits in the same chat.
• Day 3. Upload a recording of a call or a webinar, or a long document, and get notes with prompt 9 or 10.
• Day 4. Create an illustration for the post with prompt 13 and try changing the background on your own photo with prompt 14.
• Day 5. Run Deep Research on your niche with prompt 18 and open three sources from the report.
• Day 6. Build your own editor helper in Gems or Skills with prompt 22.
• Day 7. Hold a voice brainstorm in Gemini Live with prompt 21 and put together a plan for the month.

A week from now you'll have finished posts, a set of notes, images, a niche overview, and a personal helper. Most important, you'll know from your own tasks where Gemini helps you and where another tool is the better fit.

Section 16The whole path in a minute

• Gemini is Google's AI, formerly Bard. It runs at gemini.google.com and in the apps for phone, Mac, and Windows.
• It's available in 230+ countries and territories and in 70+ languages. The feature set depends on your country, age, and plan.
• The free plan includes the 3.6 Flash model, images, Deep Research, Live, Canvas, Gems, and Notebook.
• Write prompts with the formula: who you are, who it's for, what to do, the format, an example.
• Upload up to 10 files per prompt. Video up to 5 minutes and audio up to 10 minutes for free, up to 1 hour and 3 hours on AI Pro.
• Turn on Deep Research for market overviews, Gemini Notebook for your own materials, and Gemini Live for ideas on the go.
• Pack the tasks you repeat into Gems or Skills.
• Plans: $0, $4.99, $19.99, and from $99.99. The main difference is the limits and the chat memory.
• Check facts against the sources and don't upload your clients' personal data.

Open Gemini today and hand it the first task from the checklist. Fifteen minutes will give you ten post ideas for the weeks ahead.