Using AI Right Now: A Quick Guide for 2025
Date Published

Which AI to use, and how to use it best by Ethan Mollick (summary and rewrite)
TL;DR
In 2025, the AI world is dominated by ChatGPT, Claude, and Gemini. Choosing the right one isn't just about the smartest model, but the full experience: features, workflows, and how well they fit your needs.
1. Top AI Systems to Choose From
If you're serious about using AI, focus on:
ChatGPT (OpenAI)
Claude (Anthropic)
Gemini (Google)
Each offers:
Fast & powerful models
Voice chat
Image & document understanding
Coding capabilities
Mobile apps
Image/video generation (ChatGPT & Gemini)
Deep Research functionality
💡 Best Value: Paid plans at ~$20/month give you full feature access.
Other Mentions:
Perplexity: Search-focused
Manus: Autonomous agent
Microsoft Copilot: Integrated with Windows
DeepSeek: Free & capable, but lacking in key areas
2. How to Choose a Model
Think of models like vehicles:
Task TypeChatGPTClaudeGemini
Fast model (casual)
GPT-4o
Sonnet
Gemini Flash
Workhorse model (deep work)
o3 / GPT-4.5
Opus
Gemini Pro 2.5
Ultra advanced (rare use)
o3-pro
-
-
🧠 Tip: Most free versions default to the casual model. Switch manually to the stronger model if available.
3. Privacy & Memory
Claude: Doesn’t train on your data ✅
ChatGPT: Turn off training under settings ✅
Gemini: More limited privacy options ❗
ChatGPT's memory can be useful but is still inconsistent.
4. Deep Research: Secret Power Tool
This isn’t just Q&A—Deep Research is your mini analyst. Examples:
Gift ideas for a chess-loving 11-year-old
Travel guide focused on cheese in Wisconsin
Legal or medical second opinions
🎯 Accuracy is higher with Deep Research than basic queries.
5. Voice Mode: The Underrated Superpower
Voice Mode lets you:
Converse hands-free
Share your screen or camera
Get real-time visual help (math, cooking, foreign signs)
📱 Best on: ChatGPT & Gemini
6. Making Things (Images, Code, Docs)
ChatGPT: Most controllable image generation (via DALL-E)
Gemini: Two generators (Imagen + Multimodal)
Claude: Text-focused, no image output
Gemini Veo: Impressive video generation 🔥
Use "Canvas" in ChatGPT or Gemini to generate:
Documents
Code
Interactive apps
Simulations
7. Prompting: Simpler Than You Think
Good news: Complex prompting is mostly unnecessary for daily use.
✅ Tips:
Give rich context (upload docs/images)
Be clear about your goal
Give step-by-step direction if needed
Ask for quantity: 30 ideas instead of 3
Use branching to explore alternative paths
8. Common Pitfalls to Avoid
Hallucinations still happen—double-check outputs
AI doesn’t explain logic well—use “Show Thinking” when available
Don’t expect miracles—it’s not magic, it’s smart computation
9. Your First Hour with AI
Try this:
Switch to the powerful model and tackle a real work challenge
Run Deep Research on a real-life problem
Use voice mode while walking, cooking, or commuting
Final Thoughts
Don’t use AI like it’s just a better Google. These tools are thinking partners, capable of:
Generating insights
Validating your ideas
Challenging assumptions
The difference between a casual user and a power user is not complex prompts, but knowing the tools—and actually using them.