Google Gemini 使用教學：Gems、Canvas、Deep Research 到 Veo，一篇把這顆 AI 大腦用滿

Last week, I helped a friend who runs an e-commerce business to check his backend system. He complained that AI tools had ordered a bunch of things that weren't being used. I casually opened Gemini on his phone and found that he only used it to ask questions like "Help me write a product description" and then closed it after getting the answer. I told him that Gemini can directly help him generate a competitor research report and even produce a 10-second promotional video. He was surprised - this is probably a reflection of many people: using Gemini as a "Google version of ChatGPT", asking a few questions and then leaving, without touching the valuable features hidden inside.

This article will thoroughly explain Gemini, from the first use to Gems, Canvas, Deep Research, and Veo video, and guide Taiwanese users through the key points they should know.

What is Gemini

To put it simply, Gemini is Google's flagship AI assistant, backed by its own Gemini series models. The biggest difference between Gemini and general chat AI is that it is integrated with the entire Google ecosystem - Gmail, Google Docs, Google Drive, YouTube, Maps, and more. You can ask it "What was the main point of the quote email my boss sent me last week", and it can actually go through your inbox to find the answer. This ability to "access your data" is what sets it apart from external AI tools.

Gemini has both free and paid versions, including Google AI Pro and Ultra. The free version can be used for many things, while the paid version unlocks more powerful models, longer context (allowing for larger files to be processed), and advanced features like Veo video generation. My suggestion is to start with the free version and upgrade when you reach the limit.

First Use: Ask the Right Questions

1. Go to gemini.google.com and log in with your Google account. There is also a standalone app for mobile devices.
2. Type your needs in the input box. The key is to "speak completely" - instead of asking "Write a description", you can say "Help me write an Instagram post selling hand-drip coffee filters, with a casual tone, for office workers, emphasizing that it can be ready in 30 seconds", and the output will be vastly different.
3. After Gemini responds, don't rush to leave. You can usually follow up with more questions. You can say "Too formal, can you make it more conversational" or "Help me generate three more versions", and it will remember the previous context and respond accordingly.
4. To process files, simply drag and drop PDFs, images, or spreadsheets into the input box and ask questions about the content.

To be honest, just developing the habit of "speaking clearly and following up with questions" can make you use Gemini better than most people.

Gems: Make it Your Personal Assistant

This is a feature that I think is highly underrated. Gems are "customized small assistants" that you can set up in advance to perform specific tasks. You can define the role, tone, and rules for a Gem, and then use it to perform tasks without having to repeat the same instructions every time.

Google provides pre-made templates, such as "writing coach" and "brainstorming partner", and you can also create your own using plain language. For example, I created a "Traditional Chinese editor" Gem that follows the rules "always use Taiwanese terminology, never use simplified characters, avoid using AI-like phrases, and use a tone like a senior editor". This saves me a lot of time when editing articles. If you have repetitive tasks, such as responding to customer inquiries, writing weekly reports, or editing resumes, it's worth creating a Gem for them.

Canvas: Create Web Prototypes with Conversations

Canvas is a side-by-side workspace that's particularly suitable for long content and programming tasks that require repeated modifications. You can converse with Gemini on the left side, and the results will be displayed on the right side in real-time. You can modify any section, and the changes will be reflected immediately without having to redo the entire thing.

What's more interesting is that Canvas can now turn your text descriptions into interactive web prototypes without requiring design or programming knowledge. You can say "Create a simple to-do list webpage that can be checked and deleted", and it will generate the prototype for you to preview immediately. This lowers the barrier for people who want to quickly test ideas or create small tools. If you want to develop more complex apps, you can also check out other website building tools we've introduced.

Deep Research: Let it Generate a Report for You

This is a paid feature that I use frequently. Unlike general Q&A, which is "one question, one answer", Deep Research allows you to give Gemini a topic, and it will break it down into sub-questions, search the web, and then compile the results into a structured report with sources.

To use it, select the Deep Research mode, type "Organize the current situation and pain points of Taiwanese SMEs adopting generative AI in 2026", and Gemini will first provide a research plan for you to confirm the direction. After you agree, it will start running the research, and a few minutes later, it will produce a report with chapters. You can now upload your own files as reference sources, and after generating the report, you can also convert it into interactive charts or quizzes with one click. This feature is really convenient for market research and writing proposal backgrounds. However, remember that having sources attached doesn't mean the information is entirely accurate, and important numbers should always be verified by clicking on the original links.

Veo: Generate a Video with One Sentence

Paid subscribers can use Veo video generation. You can type a text description or upload an image, and Veo will generate a short video with synchronized audio. Those AI-generated short videos that are popular on social media, which look "not quite real but very interesting", are often made with this type of tool. For social media editors or people who want to create dynamic content without a budget for filming, this is a useful tool. However, be aware of the copyright and authenticity of generated videos, and don't use them to create misleading content.

Practical Reminders for Taiwanese Users

• Traditional Chinese is supported, but pay attention to terminology: Gemini's Chinese is very fluent, but occasionally, it may use terms like "video" and "software" instead of the preferred "movie" and "application". For important documents, remember to scan through and modify them yourself.
• Integration with Google services is its strong suit: If you're already a heavy user of Gmail and Google Docs, Gemini's integration advantages are more obvious than other AI tools. Don't waste this advantage.
• Be mindful of privacy: While it's convenient that Gemini can access your emails and files, you should consider whether to feed highly confidential company information into it.
• Combine it with other AI tools: I personally use Perplexity to verify sources and NotebookLM to organize knowledge bases, while Gemini is responsible for tasks integrated with the Google ecosystem. Don't feel like you need to use only one tool; choose the best one for each task.

For more scenario-based usage, you can check out our AI Task Guide, which helps you find the right tool for each situation.

TheAI Academy Summary and Evaluation

The most regrettable thing about Gemini is that many people only use a fraction of its capabilities. Its true value lies not in being "another chat window", but in the personalization of Gems, the research capabilities of Deep Research, and its deep integration with the Google ecosystem - these are what set it apart from other tools.

Don't treat Gemini as a chat toy; treat it as a work partner that can access your data, run research, and create prototypes. It's only worth your time when used in this way.

Our specific suggestion for Taiwanese readers is to spend 10 minutes creating a Gem for your most frequently used tasks and use Deep Research to run a report on a topic you're currently struggling with. Once you've used these two features, you'll probably never go back to just asking a few questions.

Sources

• Gemini Apps Official Release Notes (Google)
• Google AI Subscription Plan Description

(This article is compiled based on publicly available information, and features and plans are subject to Google's latest official announcements.)