Let's be honest: most people open ChatGPT, type "hello," get back "Hi, how can I help you today?" — and then have no idea what to do next. So they close the tab.
That's not a skill issue. It's just that nobody told them what these tools are actually good for.
This isn't a technical deep-dive. It's a practical answer to: what exists, what it does, and how to start using it today.
First: What Kinds of AI Tools Are Out There?
The "AI" label gets slapped on everything, but most tools fall into a few categories:
| Type | Examples | What It Does |
|---|---|---|
| Conversational AI | ChatGPT, Claude, Gemini, DeepSeek, Doubao | Writing, Q&A, analysis, translation, coding |
| Image Generation | Midjourney, DALL-E, Kling | Create and edit images from text |
| Video Generation | Sora, Kling, Vidu | Turn text or images into video |
| Code Assistants | Cursor, GitHub Copilot | Help write and explain code |
| Music Generation | Suno, Udio | Generate songs and instrumentals |
| AI Search | Perplexity, ChatGPT Search | Answers with cited sources |
For most people, start with conversational AI. Everything else can wait.
The Conversational AI Players
ChatGPT (OpenAI)
The one everyone's heard of. Free accounts get GPT-4o — you can upload images, analyze files, and ask questions. There are daily usage limits on the free tier, but they're enough for casual use. ChatGPT Plus is $20/month, removes limits, and unlocks the latest models.
Good for: drafting emails, summarizing articles, translation, brainstorming, explaining concepts, debugging code.
Requires a VPN in mainland China. Payment needs a foreign credit card or gift card.
Claude (Anthropic)
For writing and analysis, this is often better than ChatGPT. The responses feel less templated, more like a thoughtful person actually engaging with what you said. The free tier includes Claude Sonnet; Pro is $20/month.
What makes it stand out: the context window is huge. Paste in an entire long document and it won't lose track of earlier details. If you're doing anything writing-heavy, start here.
Also needs a VPN from China.
Gemini (Google)
Google's AI, tied to your Google account and able to pull real-time web results. The free tier (Gemini 2.0 Flash) is genuinely capable. Advanced is $20/month.
Best angle: deep integration with Google Docs, Drive, and Gmail. If you already live in that ecosystem, the AI assistance shows up right where you're working. No VPN needed when accessed through google.com or as part of Google Workspace.
DeepSeek
Made by a Chinese research lab, DeepSeek caused a real stir internationally when it launched. Particularly strong at math, logic, and structured reasoning. The R1 model shows its thinking step by step — interesting to watch it work through a hard problem.
Big upside: completely free, no VPN required, excellent Chinese support.
For users in China, this is one of the first stops. chat.deepseek.com.
Doubao (ByteDance)
ByteDance's AI assistant — probably the most widely used in China. It has a good mobile app, voice conversation, image understanding, and live web search. The onboarding is smooth, the interface doesn't feel intimidating.
Free, no setup friction. If you want something that just works without any configuration, this is it.
Kimi (Moonshot AI)
Strong on long document processing. Upload a PDF, drop in a long report, ask it to summarize or answer questions about specific sections — it handles this well. Good at Chinese and English. Accessible in China at kimi.moonshot.cn.
Qwen / 通义千问 (Alibaba)
Alibaba's model family. The QwQ thinking variant does well on technical and coding tasks. Free to use at qwen.ai, no VPN needed in China.
Image Generation
Midjourney
Still the benchmark for image quality and aesthetic. Runs inside Discord (a chat app). Needs a VPN from China. Basic plan is $10/month with limited generations; Standard is $30/month for heavier use.
How it works: type /imagine followed by a description, and it generates four options. You can upscale or vary them from there.
Kling / 可灵 (Kuaishou) and Jimeng / 即梦 (ByteDance)
Chinese-made image generation tools — no VPN, some free credits. Kling specializes in high-quality realistic images; Jimeng integrates with the Douyin/CapCut ecosystem. If you're in China and don't want to mess with VPNs, start with these.
Video Generation
Kling (Kuaishou)
One of the most polished video generation tools available in China. Text-to-video and image-to-video both work. Visual quality is consistently solid. Free credits available at kling.kuaishou.com.
Vidu
Chinese-made, supports 16:9 output, worth trying at vidu.studio. Has a free tier.
Sora (OpenAI)
Impressive results. Requires ChatGPT Plus or Pro, so the barrier is higher for users in China.
Code Assistants
Skip this section if you don't write code.
Cursor
A VS Code fork with deep AI integration. You can chat with your codebase, ask it to explain functions, make changes, or write new features. The most widely used AI coding tool in the developer community right now. Free tier for getting started; Pro is $20/month.
GitHub Copilot
Microsoft and GitHub's offering — integrates into most major editors and autocompletes as you type. Free for students; $10/month for everyone else.
AI Search
Perplexity
Combines search and AI answers with cited sources. You ask a question, it searches the web and synthesizes a clear answer with links. Much better than sifting through a list of blue links, without the hallucination risk of a pure chatbot. Free tier is solid; Pro is $20/month.
Needs a VPN from China.
Tips That Actually Make a Difference
1. Give it context
"Write me an email" is much weaker than "Write an email politely declining an unreasonable client request. I'm a product manager at a software company. The client wants a feature that's out of scope and wasn't in the contract." More context = better output.
2. Push back when the result is off
If the first answer isn't right, say so. "This is too formal, rewrite it casually" or "The third paragraph doesn't apply to my situation — I'm dealing with X, not Y." AI doesn't have feelings. Iterating is the whole point.
3. Assign a role
"You're a senior UX designer reviewing this feature from a user perspective" works better than just asking the same question cold. A well-defined role focuses the response.
4. Feed it your actual content
Most people treat AI like a search engine. It's actually better at processing content you give it. Paste in the full article you need summarized, the contract clause you don't understand, the codebase function that's broken. The AI can work with your specific material.
5. Specify the output format
"Give me this as a table," "format this as bullet points," "give me three different versions" — all work. You don't have to accept whatever structure it defaults to.
What to Start With
| Situation | Try This |
|---|---|
| Never used AI before | Doubao (app) or DeepSeek (web) |
| Writing, content, editing | Claude or ChatGPT |
| Math, logic, reasoning | DeepSeek R1 |
| Analyzing a document | Kimi or Claude |
| Generating images (in China) | Kling or Jimeng |
| Generating images (top quality) | Midjourney |
| Writing code | Cursor |
The Bottom Line
AI tools have developed fast, but the basic principle hasn't changed: you need a specific task, and you need to describe it clearly. Think of these tools as a capable assistant who's available 24/7 and knows a little about everything — not an oracle that magically produces answers.
Pick one tool, get comfortable with it, then expand from there. You don't need to try everything. Being genuinely good at using one tool puts you ahead of most people already.